Spiele im Vergleich

Complete three bridges using six possible words. Each chosen word is used once; checks and hints help you find the whole matching set before the explanations appear.

Vollständig deutsche, vorbereitete Beispiele ohne Live-Datenzugriff. Die ursprüngliche englische Ausgabe bleibt separat erreichbar. Fortschritt bleibt bei verfügbarem Browserspeicher auf diesem Gerät.

English

Alle 10 Rätsel sind über die Auswahl direkt erreichbar. Wiederholung nutzt dieselben Wörter und Zuordnungen; Hinweise und Prüfen helfen beim Einstieg.

Zwölf vorbereitete Sätze mit je drei gekoppelten Brücken

Keine eigene Tagesrotation. Der Tageslink führt zur separaten Einzelbrücken-Ausgabe.

Alle 12 vorhandenen Aufgaben sind ohne Tageswartezeit zugänglich; Wiederholen erzeugt keine neue Aufgabe.

12 authored practice sets

Solo + shareable result

Limited beta ready

Kandidat für eine überschaubare Rätselsammlung; langfristiges Interesse ist noch nicht belegt.

Twelve practice-only sets (six Easy/six Medium); uniqueness is only within authored matches, not every defensible English reading. Difficulty/enjoyment unmeasured; no daily schedule; browser-local saves.

Resolved and locally verified 11 October 2026: completed and unfinished attempts reconcile across tabs; delayed copy feedback cannot cross an attempt boundary; a successful later save clears its warning; All games opens the full catalog. Final replays established no further defect. Independent collection review, 15 October 2026: Unknown saved results survive read outages; full result receipts are visible and peer updates preserve valid focus. Fresh60 client checks,24 browser cases,60 native transitions and actual saved-state/peer/Copy checks across both editions passed. Accepted for invited human playtesting; human feedback remains pending.

Authored three-bridge matching sets and six words, accepted two-sided readings and unique complete assignment validation.

No live Termpath, Contextoid or LLM scoring call.

Spring/JTE + browser client; packaged authored content and browser-local progress with coordinated cross-tab attempts and explicit practice replay; no cross-device account sync.

4 Verbesserungen umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-22. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-retained-leaders-20261022/evidence/trio/ledger.json

Die Sortierung „Weitere Prüfrunden“ zählt alle nachfolgenden Prüfungen zusammen. Die ersten fünf Runden bleiben separat.

2 von höchstens 10 Runden am 2026-10-20: Deutsche Bedienhilfen und Zuordnungen geprüft. Nach den belegten Korrekturen war keine weitere Änderung begründet; sprachliche Fairness braucht menschliche Rückmeldung.

• conf/tasks/games-nonviral-further-rounds-20261020/evidence/words/ledger.json

10 von höchstens 10 Runden am 2026-10-20: 2 neue Übungsaufgaben in zehn weiteren Runden geprüft. Bisherige Aufgaben und Tagesauswahl bleiben erhalten. Sprachliche Fairness und gefühlte Schwierigkeit brauchen menschliche Rückmeldung.

• conf/tasks/games-nonviral-ten-rounds-20261020/evidence/words/ledger.json

4 frühere Prüfeinträge · 1 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json
• conf/tasks/games-nonviral-five-rounds-20261020/evidence/words/ledger.json

Alte und neue Aufgaben mit deutschen Testpersonen vergleichen: Regelverständnis, strittige Verbindungen oder Antworten, genutzte Hilfe und freiwillige nächste Runde beobachten. Ohne Englischkenntnisse testen; noch keine menschliche Bestätigung von Schwierigkeit oder Bindung.

Audit rejected alternative arrangements with players and extend the difficulty-balanced set bank.

Reassessed 11 October 2026 after two completed source-informed local play/review/improvement/replay cycles and final passing 22 client checks plus 4 desktop/phone browser cases. Retains Limited beta ready: the authored matching loop and repaired recovery were exercised, while human alternative-reading fairness, difficulty and enjoyment remain unverified. No public-release or production-readiness promotion.

Maturity gap: Twelve sets provide limited difficulty range; Garden/Software and Clothing/Software can invite plausible alternative interpretations.

Two source-informed local cycles exercised complete matching sets, deliberate partial matches, swap correction, hints, explanations, sharing and reload. A stale-tab overwrite was reproduced before repair and the completed result and newer unfinished progress were retained afterward. The second cycle verified phone play, copy feedback, save-warning recovery and catalog navigation. Final local checks: 22 client tests and 4 browser cases across desktop and phone. These checks do not establish unprimed human enjoyment, broad device accessibility or production readiness.

• modules/games/src/main/java/io/kgard/web/slice/games/game85/trio/BridgeHunterTrioController.java
• modules/games/src/main/resources/games/bridge-hunter-trio-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game85/trio/BridgeHunterTrioRenderedTest.java
• modules/games/docs/games-maturity-20261005.json
• modules/games/src/main/java/io/kgard/web/slice/games/game85/trio
• conf/tasks/evidence/games-maturity-20261005/root-browser-baseline.json
• conf/tasks/evidence/games-maturity-20261005/root-play-followup.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/99-trio/round-1
• conf/tasks/games-two-round-review-20261011/evidence/99-trio/round-2
• modules/games/src/main/resources/static/js/games/bridge-hunter-trio.js
• modules/games/src/test/js/bridge-hunter-trio.test.cjs
• modules/games/src/test/js/bridge-hunter-trio-runtime.test.cjs
• modules/web-visual/playful-games-tests/bridge-hunter-trio.spec.ts
• conf/tasks/games-quality-20261013/results/batch-066-public-trio-repair.md
• conf/tasks/games-quality-20261013/results/verification-batch-066-public-trio.md

Build a word route between endpoints, deliberately switching an ambiguous hinge word’s meaning. Eight explained journeys score route length against the shortest legal solution.

Oberfläche, Regeln und vorbereitete Aufgaben sind auf Deutsch.

German

Alle 13 Reisen sind ohne Tageswartezeit direkt zugänglich. Wiederholen erzeugt keine neue Aufgabe; Hinweise und Zurückgehen helfen beim Einstieg.

8 German word networks/journeys; unlimited undo; hints reported separately; word-step route scoring.

Solo practice

Limited beta ready

Eigenständige Wegregel innerhalb einer möglichen gemeinsamen Sammlung; neue Wege brauchen neue Entscheidungen.

Eight journeys only; arbitrary free associations are not evaluated; valid everyday connections may fall outside graph; local-only progress. Old README says real-browser QA open, but 5 October module assessment records actual completion/reload/phone baseline. Storage notification/preflight is best effort, not atomic cross-tab synchronization. Strict history remains bounded at 2,000 entries; excess forward actions require explicit Undo or a new attempt. Oversized live state or whole saves continue only in warned memory until a genuinely smaller save fits. Controlled local desktop/emulated-phone and native laptop play do not establish human learning benefit, physical-phone behavior or production durability.

No additional unresolved reproducible defect remains in the reviewed isolated local journeys after repairing unreadable-save overwrite and stale peer actions, and bounding forward history so completed attempts recover. See unchanged dated maturity gaps and limitations. Independently accepted for invited human playtesting on 13 October 2026: complete pivot meanings, native result/reset arrivals and actual peer receipts remain visible. Verification 043 covers 61 clients, 14 browser cases, 80 native frames, genuine saved histories, read-only outages and interruption guards. Human feedback and production validation remain pending.

Eight authored finite word graphs with ambiguous hinge words, explicit sense-switching and explained edges; shortest legal path solver.

None. No live KGARD Contextoid, Termpath, large lexicon, or model API connection.

Browser JavaScript; browser-local saved progress.

3 Verbesserungen umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-22. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-retained-leaders-20261022/evidence/wortbruecke/ledger.json

Die Sortierung „Weitere Prüfrunden“ zählt alle nachfolgenden Prüfungen zusammen. Die ersten fünf Runden bleiben separat.

2 von höchstens 10 Runden am 2026-10-20: Bezahlte Rückweg-Hinweise bleiben nach Neuladen erhalten; neue Vorwärts-Hinweise werden weiter gezählt. In den geprüften Abläufen war keine weitere Änderung begründet.

• conf/tasks/games-nonviral-further-rounds-20261020/evidence/cloud/ledger.json

10 von höchstens 10 Runden am 2026-10-20: Zwei neue Bedeutungsreisen: Linse mit fünf und Ton mit drei optimalen Wortschritten. Zehn Reisen über die normale Auswahl und Weiter-Navigation vollständig gespielt; Sackgassen, Hinweise, Neuladen und 320-Pixel-Ansicht geprüft. Die bisherigen Reisen bleiben erhalten; es gibt keine Tagesrotation. Sprachliche Fairness und gefühlte Schwierigkeit brauchen menschliche Rückmeldung.

• conf/tasks/games-nonviral-next-ten-rounds-20261020/evidence/words/ledger.json

4 frühere Prüfeinträge · 1 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json
• conf/tasks/games-nonviral-five-rounds-20261020/evidence/cloud/ledger.json

Alte und neue Aufgaben mit deutschen Testpersonen vergleichen: Regelverständnis, strittige Verbindungen oder Antworten, genutzte Hilfe und freiwillige nächste Runde beobachten. Ohne Englischkenntnisse testen; noch keine menschliche Bestätigung von Schwierigkeit oder Bindung.

Collect uncoached human feedback on meaning switches, route alternatives, hint/Undo use and phone comfort.

Latest recorded source-informed local play/review; carried forward, not a fresh human or production playtest. Independently accepted for invited human playtesting on 13 October 2026: complete pivot meanings, native result/reset arrivals and actual peer receipts remain visible. Verification 043 covers 61 clients, 14 browser cases, 80 native frames, genuine saved histories, read-only outages and interruption guards. Human feedback and production validation remain pending.

Maturity gap: Eight prepared journeys will run out quickly; the accepted graph still decides which otherwise plausible associations count.

Static browser JavaScript; progress in this device’s localStorage; no account or cross-device synchronization. Free hinge rotations change active meaning; legal replay validates recovered paths. No AI service or external fonts needed. Two ordered isolated local play/review/improvement/replay cycles protected cold thrown reads and peer progress, retained one paid recovery Hint, and reserved history for every remaining simple continuation. Final 64 owning/client checks, 14 R1 and 4 R2 actual desktop/emulated-phone browser cases passed; genuine older 2,000-entry saves loaded unchanged and completed after explicit Undo. All eight canonical journeys and finite authored meanings/reasons were reviewed; strict reader, authored content, word-step/par/Hint costs and earned best-record policy remain unchanged.

• modules/games/cloud-games/wortbruecke/src/puzzles.json
• modules/games/cloud-games/wortbruecke/src/engine.js
• modules/games/cloud-games/wortbruecke/README.md
• modules/games/cloud-games/build-static.mjs
• modules/games/docs/games-maturity-20261005.json
• modules/games/src/main/resources/static/games/cloud/wortbruecke/index.html
• conf/tasks/evidence/games-maturity-20261005/root-browser-baseline.json
• conf/tasks/evidence/games-maturity-20261005/root-play-observations.json
• conf/tasks/games-two-round-review-20261011/partition-8-1-20261012/evidence/3-wortbruecke/final-review.json
• conf/tasks/games-two-round-review-20261011/partition-8-1-20261012/evidence/3-wortbruecke/changed-files.json
• conf/tasks/games-two-round-review-20261011/partition-8-1-20261012/evidence/3-wortbruecke/round1-browser.log
• conf/tasks/games-two-round-review-20261011/partition-8-1-20261012/evidence/3-wortbruecke/round2-final-browser.log
• conf/tasks/games-two-round-review-20261011/partition-8-1-20261012/evidence/3-wortbruecke/round2-final-all-clients.log
• conf/tasks/games-two-round-review-20261011/partition-8-1-20261012/evidence/3-wortbruecke/final-child-verification.json
• conf/tasks/games-two-round-review-20261011/partition-8-1-20261012/evidence/3-wortbruecke/actual-original-legacy-save-manifest.json
• conf/tasks/games-quality-20261013/results/batch-043-wortbruecke-arrivals.md
• conf/tasks/games-quality-20261013/results/verification-batch-043-wortbruecke-arrivals.md

Open a mystery dossier. Find the hidden concept using as little evidence as you can.

Deutsche Bedienung, Aufgaben, Rückmeldungen und Ergebnislinks im Starterpaket. Die ursprüngliche englische Ausgabe bleibt unter ihrer bisherigen Adresse erhalten; ihre Aufgaben und Spielstände sind getrennt.

English

Alle 9 Aufgaben sind heute als Übung erreichbar. Davon 5 außerhalb der unveränderten Tagesrotation; 3 davon wurden in diesem Durchgang ergänzt. Jede Aufgabe zählt einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards: eight rotating UTC daily + four separate practice. Six suspects, three evidence cards and three unique guesses per dossier.

Solo + friend challenge

Limited beta ready

Kandidat für eine überschaubare Rätselsammlung; langfristiges Interesse ist noch nicht belegt.

Finite six-suspect deduction; exact authored answer, no open concept inference. Friend challenge shares the same puzzle, not a live room. Finite repeating content; human difficulty, enjoyment and retention remain unverified. Twelve English dossiers repeat8daily+4practice without calibrated difficulty progression. Local saves remain best effort without globally atomic or cross-device/account guarantees; local checks do not establish production or human fairness.

No unresolved gameplay defect reproduced in the bounded repaired local checks. The rd02/03/06 feedback corrections change their content fingerprints; an actual older rd02 save shows the honest edition notice and starts fresh rather than resuming that earlier edition. Independently accepted for invited human playtesting on 13 October 2026. Full twelve-case editorial review, unfamiliar daily/practice rounds, native completion/reload/Replay/Reveal and closed-help controls pass. Guarded restored-result focus preserves newer state and trusted input. Human feedback and deployment remain pending.

Authored dossiers with three progressive evidence cards and six candidate suspects.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

2 Verbesserungen umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-22. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-retained-leaders-20261022/evidence/redacted/ledger.json

Die Sortierung „Weitere Prüfrunden“ zählt alle nachfolgenden Prüfungen zusammen. Die ersten fünf Runden bleiben separat.

10 von höchstens 10 Runden am 2026-10-20: Zwei neue Übungsfälle zu Zeitmessung und kleinen Schriftzeichen. Frühere Begründungen zu falschen Kandidaten bleiben im deutschen Verlauf sichtbar. Sechs Fälle vollständig gespielt; Hinweise, Neuladen, Tastatur und Wiederholung geprüft. Bisherige Aufgaben und Tagesauswahl bleiben erhalten. Sprachliche Fairness und gefühlte Schwierigkeit brauchen menschliche Rückmeldung.

• conf/tasks/games-nonviral-next-ten-rounds-20261020/evidence/words/ledger.json

4 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json
• conf/tasks/games-nonviral-five-rounds-20261020/evidence/words/ledger.json

Alte und neue Aufgaben mit deutschen Testpersonen vergleichen: Regelverständnis, strittige Verbindungen oder Antworten, genutzte Hilfe und freiwillige nächste Runde beobachten. Ohne Englischkenntnisse testen; noch keine menschliche Bestätigung von Schwierigkeit oder Bindung.

Collect unprimed human feedback on deduction, clue value, difficulty and replay interest.

Latest recorded source-informed local play/review; carried forward, not a fresh human or production playtest. Independently accepted for invited human playtesting on 13 October 2026. Full twelve-case editorial review, unfamiliar daily/practice rounds, native completion/reload/Replay/Reveal and closed-help controls pass. Guarded restored-result focus preserves newer state and trusted input. Human feedback and deployment remain pending.

Maturity gap: The first clue deliberately fits several suspects; fairness depends on players understanding when extra evidence is necessary, and only eight daily plus four practice cases exist.

2026-10-13: two ordered local play/review/improvement/replay/review rounds completed. Original34 clients29PASS5FAIL plus separate2completecold failures and8CLI0PASS8FAIL/10late soft assertions are preserved with reached/masked boundaries. FIRST readable/cold initialURLReveal0 Copy was already truthful before late preservation guards; no FIRST scoring/Copy defect claimed. Original20freshJava=8authored+5controllercontracts+7historical provider units; no HTTP/live-provider credit. R1 repairs actual-state preflight/adoption, accepted warned-memory/cold-prior preservation, rejected feedback/phase focus and catalog link; exact2cold declarations append without changing originals. R1fresh36PASS/8CLI PASS; no freshR1Java rerun. R2 only moves SAME existing chronological guess ol between active/result wrappers with Your guesses/emptyhistory hidden, corrects three plural decoyfeedback leaves(Tweezers,Boots,Pliers does→do), and appends unchanged full-history body/newthree dailybrowser bodies. Final37PASS=16pure+21DOM,8freshauthoredJavaPASS(no finalcontroller/provider rerun),26distinct finalbrowser checks=18inventory(8R1+4R2+6newdaily)+8historiccompatibility;before4/new6 are separate repeats. Final32directPNG=22inventory+10historic; rootphonepixel reviews are CLI, not nativephone. Actual native originaldesktoprd09earned60(noReplay),R1desktopownReplay60; R2beforedailyrd02desktop80 and after same daily desktop80 with correctedfeedback/visible2guesshistory/selectedMagnetactive reload/full3literalcards+explanation/actualOSClipboard byte-identicalbefore/terminalreload. Onlyrd02edition reset was observed natively; rd03/06 fingerprints also change,other9boards exact. No nativephone/cold/peer/loss/Reveal/fallback claim. All34originalbackups and both R1snapshot sets preserved; final30originals unchanged4changed plus newbrowserfile. Public route uses fixed authored dossiers, not the historical Contextoid provider generator. Dated maturity4/date/human fairness gap/nextStep retained; no human, live provider, production or deployment claim.

• modules/games/src/main/java/io/kgard/web/slice/games/game5/v2/RedactedContextoidV2Controller.java
• modules/games/src/main/resources/games/redacted-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game5/v2/RedactedContextoidV2RoundServiceTest.java
• modules/games/src/test/java/io/kgard/web/slice/games/game5/v2/RedactedContextoidV2ControllerTest.java
• modules/games/docs/games-maturity-20261005.json
• modules/games/src/main/java/io/kgard/web/slice/games/game5/v2
• conf/tasks/evidence/games-maturity-20261005/root-browser-baseline.json
• conf/tasks/evidence/games-maturity-20261005/root-play-observations.json
• conf/tasks/games-two-round-review-20261011/evidence/13-redacted/original-baseline-review.json
• conf/tasks/games-two-round-review-20261011/evidence/13-redacted/independent-original-and-source-gate13.json
• conf/tasks/games-two-round-review-20261011/evidence/13-redacted/round1-final-review.json
• conf/tasks/games-two-round-review-20261011/evidence/13-redacted/independent-round1-review13.json
• conf/tasks/games-two-round-review-20261011/evidence/13-redacted/round2-before-review.json
• conf/tasks/games-two-round-review-20261011/evidence/13-redacted/independent-round2-before-and-source-gate13.json
• conf/tasks/games-two-round-review-20261011/evidence/13-redacted/round2-apply-decision.json
• conf/tasks/games-two-round-review-20261011/evidence/13-redacted/final-round2-review.json
• conf/tasks/games-two-round-review-20261011/evidence/13-redacted/final-source-snapshots.json
• conf/tasks/games-two-round-review-20261011/evidence/13-redacted/final-browser-capture-manifest.json
• conf/tasks/games-two-round-review-20261011/evidence/13-redacted/independent-final-review13.json
• conf/tasks/games-quality-20261013/results/verification-batch-020-terminal-focus.md

Navigate a branched word course with visible stroke costs, rough landings and one long club. Every shot moves you along a real link; aim for par.

Deutsche Bedienung, Aufgaben, Rückmeldungen und Ergebnislinks im Starterpaket. Die ursprüngliche englische Ausgabe bleibt unter ihrer bisherigen Adresse erhalten; ihre Aufgaben und Spielstände sind getrennt.

English

Alle 9 Aufgaben sind heute als Übung erreichbar. Davon 5 außerhalb der unveränderten Tagesrotation; 3 davon wurden in diesem Durchgang ergänzt. Jede Aufgabe zählt einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards (8 repeating daily + 4 practice)

Solo

Limited beta ready

Eigenständige Wegregel innerhalb einer möglichen gemeinsamen Sammlung; neue Wege brauchen neue Entscheidungen.

Twelve finite English authored courses comprise eight repeating daily boards and four practice boards. Directed word associations, stroke costs and rough penalties are game design choices, not measured semantic distance. A long club skips one intermediate landing once per attempt; Retake returns to the previous station but preserves spent strokes and the used club and consumes another shot. Par comes from the declared course and club rules. Ten committed shots, including Retakes, can end a genuine unfinished loss; Reveal is unearned. Selected shot previews are intentionally local and unsaved, so active reload preserves committed progress but clears the preview. Progress is stored only in this browser, with a visible memory-only warning if saving fails; there is no account or cross-device synchronization. Local save checks are non-atomic. The mapped default route is authored only and has no live/custom data mode. Human judgments of the word links, onboarding, fairness, difficulty, enjoyment and replay interest remain unmeasured; live-provider and production readiness remain unverified.

Resolved and locally verified 11 October 2026: adopting a peer station invalidates the old unsaved shot preview before describing its edges, preventing the reproduced undefined-reading exception. Free previews and refused actions check the current save before their effects; accepted-state identity gates local action feedback and focus independently of durable save success. Same-station shot buttons retain native focus, while hidden or removed in-game focus is rescued to the actual visible course or result. Replay effects require acceptance and rejected Replay focuses the adopted state. The owning All games link now opens the catalog; one-stroke status and return-control spacing are corrected. Original paid Hint focus, startup handoff, inside-save shot rejection, failed-storage memory play and rejected Replay save/URL protection already passed and remain compatibility evidence. Final replays established no further defect. Browser-local preflight remains non-atomic across simultaneous writers. Independently accepted for invited human playtesting on 13 October 2026. Purchased guidance follows the current station, remaining club and shots; impossible new hints are free refusals. Existing paid histories remain valid. Advice appears beside its station and accepted arrivals remain readable. Independent31 journeys/93 strict immediate and settled visibility frames; actual native Tab purchase, paid advice after spent club and Retake, refusal, real90/60, native Replay, delayed peers and getItem-only outage. Prior2112-state oracle and full12-course review attributed. Human feedback and deployment remain pending.

Authored directed word courses, stroke costs, rough and one-use long-club resource.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

3 Verbesserungen umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-22. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-retained-leaders-20261022/evidence/golf/ledger.json

Die Sortierung „Weitere Prüfrunden“ zählt alle nachfolgenden Prüfungen zusammen. Die ersten fünf Runden bleiben separat.

2 von höchstens 10 Runden am 2026-10-20: Schlagkosten und Aktionsbudget sind getrennt erklärt. Umwege, Rücknahme und Ergebnisse geprüft; keine weitere Änderung begründet.

• conf/tasks/games-nonviral-further-rounds-20261020/evidence/words/ledger.json

10 von höchstens 10 Runden am 2026-10-20: Zwei zusätzliche deutsche Parcours mit unterschiedlichen Kosten, Abzweigen und Sackgassen. Die Vorschau benennt einen einzelnen Strafschlag jetzt grammatisch richtig. Bisherige Aufgaben und Tagesauswahl bleiben erhalten. Sprachliche Fairness und gefühlte Schwierigkeit brauchen menschliche Rückmeldung.

• conf/tasks/games-nonviral-next-ten-rounds-20261020/evidence/routes/ledger.json

6 frühere Prüfeinträge · 2 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json
• conf/tasks/games-nonviral-five-rounds-20261020/evidence/words/ledger.json

Alte und neue Aufgaben mit deutschen Testpersonen vergleichen: Regelverständnis, strittige Verbindungen oder Antworten, genutzte Hilfe und freiwillige nächste Runde beobachten. Ohne Englischkenntnisse testen; noch keine menschliche Bestätigung von Schwierigkeit oder Bindung.

Collect unprimed human feedback on route decisions, guidance usefulness, challenge and replay interest.

Reassessed 11 October 2026 after two completed local play, review, improvement and replay cycles. Actual Sand shelter 100 was repeated before and after repair; phone Cotton 50 included paid Hint, discarded long shot, Retake and rough landings, followed by genuine ten-shot 0 and two fresh efficient 100 routes. All nine map readings per course, committed histories, par examples, canonical Copy and active/terminal recovery were reviewed. Stale preview invalidation, accepted-action effects and native focus were repaired with the pure engine, strict five-field save, deck and scoring unchanged. Four status-caption expectations changed only after the preserved original baseline. Final verification passed 24 client checks, thirteen fresh Java checks and twelve distinct desktop/phone browser journeys, with independent source review and bounded desktop/phone visuals. The second review required no further source change. Retains Limited beta ready, 4, and the historical need for human judgments of authored links, difficulty, fairness and replay interest. These local results do not establish atomic storage, human enjoyment, live-provider quality, deployment or public readiness. Independently accepted for invited human playtesting on 13 October 2026. Purchased guidance follows the current station, remaining club and shots; impossible new hints are free refusals. Existing paid histories remain valid. Advice appears beside its station and accepted arrivals remain readable. Independent31 journeys/93 strict immediate and settled visibility frames; actual native Tab purchase, paid advice after spent club and Retake, refusal, real90/60, native Replay, delayed peers and getItem-only outage. Prior2112-state oracle and full12-course review attributed. Human feedback and deployment remain pending.

Maturity gap: Authored word links and landing costs need human difficulty/replay feedback. Human beta feedback is still needed on onboarding, fairness, difficulty and replay interest.

The mapped player is TermpathGolfController.playableTermpathGolf, its authored round service, termpath_golf.jte, termpath-golf-deck.json and termpath-golf.js. The inventory identity remains game-32 and the persisted runtime identity remains game32. The authored-v1 envelope uses golf-1 and exactly five state fields: boardId, events, phase, replay and version. Recovery replays committed legal shots, Hint, Retake and Reveal. The selected shot preview is deliberately ephemeral; there is no notebook sidecar. The retained unannotated language/source/target method delegates to the authored player and does not expose a live-provider branch. The first actual desktop cycle used tg-09 Sand shelter. Root read all nine directed map links and rough penalties, compared free long previews to Window and Room, then committed normal Glass. A staged Room preview cleared after active reload while the paid Glass station, one stroke, one shot and available long club remained. The long shot through Room to Home earned 100 at two strokes, par two and two shots. Root read both chronological semantic shot records and the par example, copied the canonical result, recovered the terminal result and inspected a bounded readable desktop view. The original focused client run passed fifteen of 24 checks and failed nine with zero skips in 604.089209 milliseconds. All eight pure checks passed. Paid Hint focus, the inside-save Silica branch ending at 90, warned failed-storage memory play and calibrated startup already passed compatibility checks. Some nested checks stopped at an earlier failure, so nine failing cases are not nine independently observed defects. The original anchored browser run passed two of eight desktop/phone journeys and failed six with zero skips in 23.3 seconds. Every targeted case completed full 100 or 90 outcomes and finite learning evidence before final soft verdicts. Ordinary peer adoption reproduced an uncaught undefined-reading error when an old Sand preview was described from Glass. Native controls detached and hidden Copy focus was not rescued. A delayed free preview stayed at the old station until a subsequent guarded shot reconciled it. Rejected Replay already preserved the newer saved 90 and URL, but focused the wrong control. Normal 100 also exposed the owning All games link to Daily; actual inspection found one-stroke plural wording and tightly spaced desktop return controls. The runtime repair invalidates local previews before rendering adopted main state, preflights free controls and refusals, and separates actual action acceptance from successful durable storage. It preserves memory-only progress and malformed saved bytes, aligns observations after initialization and accepted saves, retains native choices for the same station and club availability, and rescues only invalid in-game focus while preserving useful map, help and outside focus. Replay requires acceptance before ancillary effects; Copy preflights the main save and retains the shared asynchronous result fences. The owning template now uses the existing catalog route and existing route-actions styling. The pure engine/export prefix, strict five-field schema, authored deck, rough and club rules, paid Retake, ten-shot limit, par computation and scoring are unchanged. After the original baseline was closed, root corrected exactly four status-caption literals from “1 strokes” to “1 stroke”: two in the DOM fixture and one in each browser file. All behavioral assertions and case counts remained unchanged, and original fixture backups were retained. Original results are not claimed against the revised wording. Independent read-only review verified the full source diff, all sixteen DOM cases, byte-identical pure prefix and exact four wording-only fixture changes. The repaired actual desktop replay again earned Sand shelter 100 after normal Glass and the long Room-to-Home shot. All nine map explanations, both paid receipts, free preview behavior, active recovery, par plan, canonical Copy and terminal reload were checked. The bounded desktop view showed separated return controls and the correct catalog destination. Root fresh verification passed all 24 client checks with zero skips in 565.959458 milliseconds. Thirteen fresh Java checks passed at 06:55:01 UTC and the isolated build succeeded in eleven seconds; the immutable package, served runtime and packaged inventory were checked. Eight anchored first-cycle browser journeys passed on desktop and phone in 10.5 seconds, with no skips. The second actual cycle used tg-10 Cotton on a phone. Root read all nine directed arrows and rough penalties, bought the Hint, used the long club from Cotton through Fibre to Yarn, then paid to Retake to Cotton with the club still spent. Normal Fibre and Blanket shots each incurred their landing rough. Active reload after staging Comfort restored six strokes and four shots while clearing the unsaved preview. Reaching Comfort earned 50 at seven strokes, par three and five shots. Root read all six chronological entries, including the discarded long shot and paid Retake, the par example, canonical Copy and terminal recovery. The review justified no further repair. A fresh actual phone Replay used five normal Fabric shots and five Retakes to exhaust ten shots at Cotton for an honest loss of 0. Every one of its ten committed records, canonical unfinished result and terminal recovery was checked. Another fresh Replay used normal Fibre and the long shot through Blanket to Comfort to earn 100 at three strokes, par three and two shots; both receipts and all map readings were reviewed. An additional fresh route used normal Fabric, a long shot through Cushion to Sofa and normal Comfort to earn 100 at three strokes, par three and three shots. Active Sofa recovery, all three receipts, canonical Copy, terminal reload and bounded phone views were inspected. Four anchored second-cycle browser journeys passed on desktop and phone in 7.4 seconds, with no skips. The rounds provide twelve distinct passing browser journeys, eight first-cycle and four second-cycle. No source changed during the second cycle. The final runtime SHA-256 is 9f7dafbb3b4d88a829413ceae57fc27bb62b3b76a708da243416b9567ae29e97 and template SHA-256 is 53a332bde57c4ba73afc636145e0d8d1794ecf1059fb0eccfe6d9b01fe4b17b4. This is local agent gameplay, automated verification and source review; it does not establish atomic storage, human semantic agreement or enjoyment, live-provider quality, deployment or public readiness.

• modules/games/src/main/java/io/kgard/web/slice/games/game32/TermpathGolfController.java
• modules/games/src/main/resources/games/termpath-golf-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game32/TermpathGolfPlayableTest.java
• modules/games/docs/games-maturity-20261010-next-prototype-beta.json
• conf/tasks/evidence/next-ten-games-20261010/route
• conf/tasks/evidence/next-ten-games-20261010/final-verification.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/58-termpath-golf/verification.json
• conf/tasks/games-two-round-review-20261011/evidence/58-termpath-golf/client-repair.json
• conf/tasks/games-two-round-review-20261011/evidence/58-termpath-golf/client-repair.diff
• conf/tasks/games-two-round-review-20261011/evidence/58-termpath-golf/round-1-before
• conf/tasks/games-two-round-review-20261011/evidence/58-termpath-golf/round-1-after
• conf/tasks/games-two-round-review-20261011/evidence/58-termpath-golf/round-2-final
• modules/games/src/main/java/io/kgard/web/slice/games/game32/termpath_golf.jte
• modules/games/src/main/resources/static/js/games/termpath-golf.js
• modules/games/src/test/js/termpath-golf.test.cjs
• modules/games/src/test/js/termpath-golf-ui.test.cjs
• modules/web-visual/playful-games-tests/termpath-golf.spec.ts
• modules/web-visual/playful-games-tests/termpath-golf-round2.spec.ts
• conf/tasks/games-quality-20261013/results/verification-batch-025-advice-arrival.md

Assemble a line from a surplus station bank. Choose among supported routes while meeting service requirements and managing fare.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben, Rückmeldungen und Ergebnislinks. Die englische Ausgabe und ihre Aufgabenkennungen bleiben separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle 7 Aufgaben sind heute ohne Wartezeit erreichbar. Davon 3 Übungen außerhalb der Tagesrotation; jede Aufgabe zählt nur einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards (8 repeating daily + 4 practice)

Solo

Limited beta ready

Als spielbare Variante erhalten; derzeit keine eigene Ausbau- oder Wachstumslinie vorgesehen.

Twelve finite English authored boards comprise eight repeating daily boards and four separate practice boards. Directed associations, station fares and service requirements are assigned game rules, not measured semantic distances. Only declared directed links are traversable, and each complete line must cover both required services within five intermediate tiles. A win earns 100 minus three points per fare above the cheapest valid line, fifteen per earlier distinct failed inspection and ten for the optional Hint; three distinct failed inspections end an honest unfinished loss of 0. Empty and repeated failed inspections and editing are free; Reveal is unearned. The 2,000-event compatibility boundary compacts only redundant free edits when needed for another valid action, preserving paid evidence and the current line. Progress is browser-local, with a visible warning and memory-only continuation if writing fails; no account or cross-device synchronization exists. Preflight cannot make local storage atomic. The mapped default route is authored only, and the retained language/source/target method delegates to the same authored player rather than a live/custom graph. Human semantic agreement, onboarding, fairness, difficulty, enjoyment and replay interest remain unmeasured; provider, production and public-release readiness remain unverified.

Resolved and locally verified 11 October 2026: a legal 2,000-event active save could previously load but no longer accept Add, Inspect or other progress. New valid actions at that limit now compact redundant free editing while retaining each paid inspection and Hint in order, its exact evaluated line, the final draft and replay status. Strict recovery still accepts original legal histories through 2,000 unchanged and rejects oversized or malformed histories. Owning preflight and accepted-state checks prevent obsolete free-refusal or failure feedback after peer adoption; Replay and Copy use current saves and accepted reset outcomes. Native bank buttons and per-ID line rows remain stable where valid, with visible Hint/Reset and adopted-state focus. The catalog link, service-specific help, one-failed-inspection share wording and return spacing are corrected. The intermediate candidate replaced an existing line row when a peer added a tile; this was caught and repaired before final verification. Startup, noncheapest routes and warned memory play already passed original compatibility checks. Final replays established no further defect. Local storage checks remain non-atomic across simultaneous writers. Independently accepted for invited human playtesting on 13 October 2026. Restored result headings and full scores remain visible in twelve native reload journeys and 66 samples; fourteen input/state guard controls passed. All twenty-four combined board identities remain unchanged. Human feedback and deployment validation remain pending.

Authored directed word graph, station bank, service requirements and fares.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

2 Verbesserungen umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-22. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-four-candidates-20261022/evidence/subway/ledger.json
• conf/tasks/games-four-candidates-20261022/verification.json

Die Sortierung „Weitere Prüfrunden“ zählt alle nachfolgenden Prüfungen zusammen. Die ersten fünf Runden bleiben separat.

10 von höchstens 10 Runden am 2026-10-20: Deutsche Ausgabe mit vier Karten: Fahrpreis und zwei Stationsarten abwägen. Regeln, Verbindungsprüfung, Wiederholung und Ergebnisse sind auf Deutsch. Die englische Ausgabe bleibt erhalten. Sprachliche Fairness und gefühlte Schwierigkeit brauchen menschliche Rückmeldung.

• conf/tasks/games-nonviral-fourth-ten-rounds-20261020/evidence/words/ledger.json

6 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json
• conf/tasks/games-nonviral-five-rounds-20261020/evidence/routes/ledger.json

Neue und bisherige Aufgaben mit deutschsprachigen Ersttestern ohne vorausgesetzte Englischkenntnisse vergleichen: Regelverständnis, strittige Bedeutungen, Hilfen und freiwillige weitere Runde beobachten. Danach gezielt nachbessern.

Collect unprimed human feedback on onboarding, fairness, difficulty and replay interest; live deployment remains separately unverified.

Reassessed 11 October 2026 after two completed local play, review, improvement and replay cycles. Actual Sun 75 before and after repair, phone Ore 85, genuine three-distinct-inspection 0 and fresh cheapest 100 were completed with every finite map, link, service and alternative-line reading, canonical Copy and active/terminal recovery. The reachable 2,000-event completion trap, accepted-action effects and native focus were repaired while preserving strict five-field recovery, paid chronology, fare/service rules, scoring and the authored deck. Exactly one singular-word DOM expectation changed after the retained original baseline; existing pure and browser fixtures were unchanged. Final verification passed 23 client checks, twelve fresh Java checks and twelve distinct desktop/phone browser journeys, with independent source review and bounded desktop/phone visuals. The second review justified no further source change. Retains Limited beta ready, 4, and the historical need for human judgments of authored connections, onboarding, fairness, difficulty and replay interest. No human, live-provider, atomic-storage, deployment or public-readiness claim follows. Independently accepted for invited human playtesting on 13 October 2026. Restored result headings and full scores remain visible in twelve native reload journeys and 66 samples; fourteen input/state guard controls passed. All twenty-four combined board identities remain unchanged. Human feedback and deployment validation remain pending.

Maturity gap: Finite authored connections and fares are not measured graph distances. Human beta feedback is still needed on onboarding, fairness, difficulty and replay interest.

The mapped player is SemanticSubwayTilesController.playableSemanticSubwayTiles, SemanticSubwayTilesRoundService, semantic_subway_tiles.jte, semantic-subway-tiles-deck.json and semantic-subway-tiles.js. The inventory identity remains game-30 and the persisted runtime identity remains game30. The authored-v1 envelope uses subway-tiles-1 and exactly five state fields: boardId, events, phase, replay and version. Recovery replays legal add, remove, move, reset, Hint, Check and Reveal events. There is no notebook sidecar. The retained unannotated language/source/target method delegates to the authored player and does not provide a live-data branch. The first actual desktop cycle used st-09 Sun. Root read the help, both service goals, all ten directed map readings and six bank tiles with their fares and services. Empty Inspect was free; Sunburn committed a failed inspection costing fifteen, and its repeat was free. Root returned Sunburn, bought the ten-point Hint, added Solar panel then Light, and moved Light Earlier. Active reload retained the exact Light/Solar panel line, failed inspection and Hint. Inspect earned 75 at fare four. Root read all three directed-link audits, both service audits and all four valid alternatives with their fares, copied the canonical result, restored the terminal result and inspected a bounded desktop view. An initially guessed invalid practice identifier returned 404 and was corrected from the mapped fixture; this was not a product defect. The original root client run passed twelve of 23 checks and failed eleven, with zero skips in 2023.531375 milliseconds. All seven existing pure checks passed. Completed 75, 85, genuine 0, 100, noncheapest-route, warned-memory and calibrated startup behavior supplied compatibility evidence. The original eight anchored desktop/phone browser journeys all failed at protected boundary verdicts; the full log, including its initially truncated middle, was reviewed. The first three journeys completed real 75, 100 or 90 outcomes before later boundary verdicts. The capacity journey first completed 100, then generated 2,000 genuine free-edit and Hint events through controls and remained trapped after Add/Inspect. It did not complete the requested 90 in the original client, and no such credit is claimed. Observed failures included replaced native controls, hidden focus, obsolete guidance and rejected-action effects. Actual inspection also found the owning All games link led to Daily, help mentioned nonexistent rough/closure costs, and Copy pluralized one failed inspection. The repair compacts redundant free add/remove/move/reset runs only when a new valid action reaches the 2,000-event boundary. It preserves every paid Check and Hint in order with the exact line evaluated at that point, plus the final line and replay flag. Original valid histories up to 2,000 remain recoverable without normalization; 2,001-event and malformed extra-key histories still reject. The legal route, fare, service, penalty and score rules and authored deck are unchanged. Owning current-save preflight now covers free or refused actions and Copy. Actual state identity gates feedback and focus independently of whether an accepted in-memory write was durable. Guarded observation aligns after initialization and reconciliation. Peer adoption clears obsolete local messages, keeps valid native bank and per-ID line controls, and rescues hidden or disconnected in-game focus without stealing outside focus. Replay ancillary effects require actual acceptance; rejected resets preserve newer bytes and URL. Shared asynchronous Copy fences remain in use. The initial candidate passed 21 of 23 client checks and failed two, with zero skips in 3086.361375 milliseconds. One was the expected singular-word correction in the full-75 Copy assertion. The other was a real existing-row replacement when a peer added a tile. Per-ID row caching and minimal reordering repaired that focus issue. After the original baseline had closed, exactly one owning DOM regex word changed from “1 failed inspections” to “1 failed inspection”; its normalized remaining bytes were identical, and no behavioral boundary or case count was weakened. The seven pure checks, authored deck and both browser files remained byte-identical. Independent source review caught an undefined catalog constant in the draft template, which root replaced with the existing literal /games/catalog before any build. The final template also corrects authored-service help, makes the exposed Hint focusable and uses existing route-actions spacing for result navigation. The repaired actual desktop replay again earned Sun 75 using the paid Sunburn mistake, free repeat, Hint, reversed two-tile draft and Earlier control. Native Hint focus, all ten active map readings, exact active recovery, all three link audits, both services, four alternatives, canonical singular Copy and terminal reload were checked. The actual next-board link opened Ore st-10. Root fresh verification passed all 23 client checks with zero skips in 2902.299125 milliseconds. Twelve fresh Java checks passed at 07:35:41 UTC and the build succeeded in fourteen seconds. The immutable package, served runtime and packaged 43-review inventory were verified. All eight anchored first-cycle desktop/phone browser journeys passed with zero skips in 16.3 seconds. The real long-edit journey completed 100, performed 999 Add/Return pairs, bought Hint and added Light to reach 2,000, then successfully added Solar panel and inspected for 90 with bounded recovery. The independent literal-2,000 fixture retained an earlier failure and Hint and completed 75 with their original order. Independent final source review passed. The second actual cycle used st-10 Ore on a phone. Root read all ten directed map readings, both service goals and six fare/service bank tiles. Hammer, Ingot and Steel were added in reverse; Steel moved Earlier twice, Ingot moved Earlier once, and Hammer was returned and re-added. Active reload retained the exact Steel/Ingot/Hammer line at fare nine with both services. Inspect earned 85: 100 minus fifteen fare excess, with no failed inspections or Hint. All four link audits, both service audits and all four alternative lines/fares were reviewed. Canonical Copy, terminal reload and a bounded readable phone view completed the baseline. This second review justified no additional source change. An actual fresh phone Replay inspected three distinct one-tile lines, Steel, Copper and Mine, with free Reset between them. The third committed failure exhausted the attempt and honestly earned 0. Root reviewed the final missing-link and missing-service audits, all ten map readings and all four valid alternatives, then copied and recovered the unsolved result. Another fresh Replay added Steel, reloaded the active one-tile line and its covered metal service, then added Hammer and inspected the cheapest fare-four line for 100 with no misses or Hint. All ten map readings, three directed-link audits, both service audits and four alternatives, canonical Copy, terminal recovery and bounded phone rendering were reviewed. All four anchored second-cycle desktop/phone browser journeys passed with zero skips in 8.0 seconds. These rounds supply twelve distinct passing browser journeys, eight first-cycle and four second-cycle, with full logs read. No source changed in the second cycle. The final runtime SHA-256 is 235b0882abb1bd0204a90342e26c30974f1d68b4e570213681ddbf07f51ae82a and template SHA-256 is 5f5e7ecc37c0eb1746e7da7788c0a412af27b233f21cb890ccaae20936cfcb72. This evidence is local agent gameplay, source review and automated verification. It does not establish atomic storage, human semantic fairness or enjoyment, successful live-provider behavior, deployment or public readiness.

• modules/games/src/main/java/io/kgard/web/slice/games/game30/SemanticSubwayTilesController.java
• modules/games/src/main/resources/games/semantic-subway-tiles-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game30/SemanticSubwayTilesPlayableTest.java
• modules/games/docs/games-maturity-20261010-next-prototype-beta.json
• conf/tasks/evidence/next-ten-games-20261010/route
• conf/tasks/evidence/next-ten-games-20261010/final-verification.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/56-semantic-subway-tiles/verification.json
• conf/tasks/games-two-round-review-20261011/evidence/56-semantic-subway-tiles/round-1-before
• conf/tasks/games-two-round-review-20261011/evidence/56-semantic-subway-tiles/round-1-final
• conf/tasks/games-two-round-review-20261011/evidence/56-semantic-subway-tiles/round-2-final
• modules/games/src/main/java/io/kgard/web/slice/games/game30/semantic_subway_tiles.jte
• modules/games/src/main/resources/static/js/games/semantic-subway-tiles.js
• modules/games/src/test/js/semantic-subway-tiles.test.cjs
• modules/games/src/test/js/semantic-subway-tiles-ui.test.cjs
• modules/web-visual/playful-games-tests/semantic-subway-tiles.spec.ts
• modules/web-visual/playful-games-tests/semantic-subway-tiles-round2.spec.ts
• conf/tasks/games-quality-20261013/results/verification-batch-018-restored-results.md

Place nine contextual stars among three cores or a shared-star wormhole. Balance meaning, counts and energy, with several valid arrangements and a low-energy objective.

Deutsche Bedienung, Aufgaben, Rückmeldungen und Ergebnislinks im Starterpaket. Die ursprüngliche englische Ausgabe bleibt unter ihrer bisherigen Adresse erhalten; ihre Aufgaben und Spielstände sind getrennt.

English

Alle 9 Aufgaben sind heute ohne Wartezeit erreichbar. Davon 5 Übungen außerhalb der Tagesrotation; jede Aufgabe zählt nur einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards: eight rotating UTC daily + four separate practice. Nine stars, three cores and one shared-star wormhole.

Solo

Limited beta ready

Planung und Wiederholungswert müssen sich im tatsächlichen Spielen bewähren.

Twelve repeating authored English boards; meaning membership and energy are declared constraints, not live Contextoid graph results or physics. Device-local saves remain non-atomic without account sync. Human difficulty, enjoyment, retention, provider/XR quality and production readiness remain unverified.

The13 October independent review accepted Semantic Supernova after electronic Technology and Buildings and their layout clarified two printed category contracts. All108 star scenes and12 boards were reviewed; old earned100 records remain intact across the two edition changes. Actual native play, alternate85/clean100 results, read-only outage and peer recovery passed. Accepted for invited human playtesting; human feedback remains pending. This is local technical evidence, not deployment or public-release signoff.

Authored word-to-core memberships plus assigned energy, counts and capacities.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

2 Verbesserungen umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-22. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-four-candidates-20261022/evidence/supernova/ledger.json
• conf/tasks/games-four-candidates-20261022/verification.json

Die Sortierung „Weitere Prüfrunden“ zählt alle nachfolgenden Prüfungen zusammen. Die ersten fünf Runden bleiben separat.

1 von höchstens 10 Runden am 2026-10-20: Gefahren, öffentliche Lasten, Kosten und Ergebnis geprüft. Kein zusätzlicher Fehler belegt.

• conf/tasks/games-nonviral-further-rounds-20261020/evidence/strategy/ledger.json

10 von höchstens 10 Runden am 2026-10-20: Zwei neue Übungsaufgaben verbinden neue Bedeutungsfelder mit unterschiedlichen Kapazitäten und mehreren gültigen Verteilungen. Kosten, Entwurfswechsel, Wiederholung und Tastaturbedienung wurden gespielt. Bisherige Aufgaben und Tagesauswahl bleiben erhalten. Sprachliche Fairness und gefühlte Schwierigkeit brauchen menschliche Rückmeldung.

• conf/tasks/games-nonviral-third-ten-rounds-20261020/evidence/strategy/ledger.json

4 frühere Prüfeinträge · 2 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json
• conf/tasks/games-nonviral-five-rounds-20261020/evidence/strategy-next/ledger.json

Neue und bisherige Aufgaben mit deutschsprachigen Ersttestern ohne vorausgesetzte Englischkenntnisse vergleichen: Regelverständnis, strittige Bedeutungen, Hilfen und freiwillige weitere Runde beobachten. Danach gezielt nachbessern.

Collect unprimed human feedback on comprehension, fairness, difficulty and replay interest; technical readiness and human feedback remain separate in conf/tasks/games-quality-20261013/game-quality.json.

Latest recorded source-informed local play/review; carried forward, not a fresh human or production playtest.

Maturity gap: Twelve authored scenes constrain replay; meaning judgments are scene-specific. Human beta should test whether capacity planning feels engaging and whether novices understand shared stars.

The mapped /games/game12 player uses SemanticSupernovaAuthoredDeck, semantic_supernova_beta.jte and supernova-beta.js, with runtime identity game12. Twelve English boards provide eight rotating UTC daily puzzles and four practices. Nine contextual stars must satisfy scene membership, core and wormhole counts, and assigned energy capacities. Successful scores start at 100, subtract 15 per earlier failed test and five per excess wormhole energy. Three distinct failed complete arrangements end at zero; incomplete and repeated tests are free, Clear retains attempts, and Reveal is unscored. Retained provider/controller methods remain unmapped. Energy is a designed constraint, not measured semantic strength or physics. Original native Astronomy play completed efficient 100 on desktop and higher-energy 85 on phone, with partial and complete-draft recovery. Reviews read all nine scene/energy records, three core definitions, four ledger rows, nine local loads, every paid issue, nine efficient-placement reasons, the six-arrangement benchmark and score equation. Actual canonical Copy and terminal recovery worked. Desktop measured 1280 by 720, page width 1265; phone measured 390 by 844, page width 375. Native peer completion and Replay left focus on the document body. Original verification passed 33 of 41 clients with eight failures in 749.257375 milliseconds, and two of eight browser journeys with six failures in 29.4 seconds. Full outcomes preceded deferred focus, rejected-feedback and false-save-status verdicts; all twelve baseline captures completed. Before this browser baseline, its blocked-Reveal fixture was calibrated to capture initial status before Copy/Replay and assert it after full memory 100, retaining the final won-status check. Two full-outcome regressions appended after baseline closure both failed against the unchanged original in 397.98925 milliseconds. Cold unreadable storage with working writes overwrote an unknown paid Law 85 during memory 100 and Replay, hid the warning and prevented recovery of that result; memory completion and subsequent durable alternate 100 preceded the aggregate verdict. Durable initial Reveal already supported first Copy: its separate case failed only the accepted-Replay focus expectation. Existing startup, scoring and accepted memory outcomes are not newly discovered defects. The owning repair reconciles changed attempts before menus, Test, Clear, Reveal, Replay and Copy. Accepted state identity controls feedback and follow-up effects separately from durability. Readable-save protection preserves unknown bytes during read outages while allowing warned memory progress; compatible observation aligns initialization, reconciliation and own Reveal. Only accepted Replay applies reset effects, while unchanged free Test results skip redundant revisions. Focus captured before rendering and shared notifications preserves enabled menus and outside controls, rescuing invalid targets to the actual select or result. Neutral terminal wording no longer promises a durable save. Four-field recovery, daily replay restrictions, scoring, chronology, rules, deck, Java, shared helpers and CSS remain unchanged. Independent source review passed. Repaired native Astronomy 100 and 85 repeated the complete proofs, actual Copy and recovery; peer focus now followed the visible result or first menu. Eight intermediate repaired browser journeys passed in 36.7 seconds. The second native phone round used Law. Contract in Law caused energy 14 against capacity 11 and an empty wormhole. Exact repeat was free; moving Contract to the wormhole and reloading the complete draft earned 85 with equation 100 minus 15 minus zero. Three distinct all-in-one-core drafts then genuinely exhausted the tests at zero, retaining paid history through Clear and reload after two failures. All issues, nine reasons, four ledger rows, nine loads, two-arrangement benchmark, actual Copy and terminal recovery were reviewed. The only second-round change corrected local-load singular wording to one star, with matching DOM and browser helper literals; all test bodies and browser journey bodies remained unchanged. Fresh alternate Law 100 on phone and Astronomy 100 on desktop completed full evidence, actual Copy and recovery on final source, with bounded readable layouts. All 43 final clients passed in 948.023292 milliseconds: seventeen owning DOM, thirteen owning pure and thirteen Sheriff compatibility checks. Twelve fresh original Java checks are retained without claiming a final rerun: three mapped authored, three retained unmapped controller and six retained service checks. The final combined browser run passed twelve distinct desktop/phone journeys in 59.0 seconds, with twenty completed captures and none unavailable; earlier runs are rechecks, not extra distinct cases. Final runtime SHA-256 is 3c77ef50b69e088db85bd03fedecbfd0bcba923968ef9a0f1723b58c61821f51. The thirty-five-input audit records four changed files: runtime, two appended DOM cases plus its grammar helper, the initial-Reveal browser calibration, and browser grammar helper; thirty-one inputs are unchanged. Original fifteen DOM bodies remain exact. The immutable local classpath has 596 entries and 17,883,271 copied bytes; source and packed sixty-five-review base inventory agree. This is not a fat archive, deployment or live-provider check. Browser play covers two boards and eighteen distinct scene records; all twelve boards and 108 scene/reason rows were source-reviewed. Maturity 4, its 10 October date, basis and historical gap remain unchanged. Earlier known-issue record, retained for history: The 11 October review repaired stale action feedback, invalid focus, unreadable-save overwrite and false saved-status wording; round two corrected one-star local-load text. Both local cycles passed. Maturity remains 4.

• modules/games/src/main/java/io/kgard/web/slice/games/game12/SemanticSupernovaController.java
• modules/games/src/main/resources/games/semantic-supernova-beta.json
• modules/games/src/test/java/io/kgard/web/slice/games/game12/SemanticSupernovaRoundServiceTest.java
• modules/games/src/test/java/io/kgard/web/slice/games/game12/SemanticSupernovaControllerTest.java
• modules/games/docs/games-maturity-20261010-third-prototype-beta.json
• conf/tasks/evidence/third-ten-games-20261010/root
• conf/tasks/evidence/third-ten-games-20261010/final-verification.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/baseline.json
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/baseline-browser-calibration.json
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/actual-round1.json
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/actual-round2.json
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/original-client.log
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/original-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/original-browser-outcomes.json
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/additional-regressions.json
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/original-additional-regressions.log
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/java-original.json
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/round1-proposed-runtime.diff
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/independent-path-review-round1.json
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/round1-client.log
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/round1-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/round2-local-load.diff
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/round2-client.log
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/final-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/final-browser-outcomes.json
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/final-hashes.json
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/classpath-round2-package.json
• conf/tasks/games-two-round-review-20261011/evidence/34-semantic-supernova/verification.json
• modules/games/src/main/resources/static/js/games/supernova-beta.js
• modules/games/src/main/resources/static/js/games/supernova-beta-rules.js
• modules/games/src/test/js/supernova-ui.test.cjs
• modules/web-visual/playful-games-tests/semantic-supernova.spec.ts
• modules/web-visual/playful-games-tests/semantic-supernova-round2.spec.ts
• modules/web-visual/playful-games-tests/semantic-supernova-support.ts
• conf/tasks/games-quality-20261013/results/verification-batch-014.md

Build directed routes from either endpoint and earn ordered passport stamps. Four authored challenges on two endpoint pairs support hints, checked word removal and shortening to a shortest-route medal.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben und Rückmeldungen. Die englische Ausgabe bleibt separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle 8 Aufgaben sind heute ohne Wartezeit erreichbar. Vier neue Pässe verteilen sich auf zwei zusätzliche Wortnetze; Wiederholen erzeugt keine neue Aufgabe.

Vier Passaufgaben auf zwei Endpunktpaaren

Keine Tagesrotation.

Alle 4 vorhandenen Aufgaben sind ohne Tageswartezeit zugänglich; Wiederholen erzeugt keine neue Aufgabe.

4 English passport challenges; 2 endpoint pairs; shared authored graph has 53 edges and 160 vocabulary entries; three- and four-stamp goals.

Solo practice

Limited beta ready

Neu aus der Reserve für den deutschen Vergleich ausgewählt. Als Variante innerhalb dieser Sammlung prüfen; keine belegte Nachfrage und keine Zusage einer eigenen Inhaltsserie.

4 finite challenges reuse only 2 endpoint pairs and one small authored graph; restricted vocabulary and incoming reverse-side edges need learning; human engagement/learnability unmeasured. Browser progress uses optimistic checks between pages rather than an atomic shared transaction. Verifying an older full history may require 161 additional sequential service checks. A history of 160 paid or meaningful actions requires restarting; controlled local storage/service failures are now covered; physical-device and production behavior remain unverified.

No additional unresolved defect established by this source audit; see limitations and the dated maturity review. Two local rounds repaired stale-page overwrites, unreadable-storage writes, missing manual Copy, refused-history continuation and the arrival-side tool/form mismatch. Final local checks pass; the deliberate 160 paid-or-meaningful-action limit remains. Independently accepted for invited human playtesting on13 October2026. The loaded main heading is persistent; phone word/type labels and removal controls remain readable and contained. Final independent checks cover24 layouts across8 widths,8 real edits and128 native frames. One original owner Review activation remains unclassified; five exact owner/independent repeats passed without code changes. Human feedback and production validation remain pending.

Frozen directed word graph with semantic types and ordered passport/checkpoint constraints; filtered edges and BFS over word/checkpoint state produce legal constrained shortest routes.

None. No live KGARD Contextoid, Termpath, large lexicon, or model API connection.

Browser interface with browser-local route and draft storage; Spring-managed Node 24+ stateless scoring API. Sense Seals has no SQL round store or anonymous-cookie ownership; the shared site's SQLite social challenges belong to Replykin and Word Roast.

4 Verbesserungen umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-22. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-four-candidates-20261022/evidence/wordport/ledger.json
• conf/tasks/games-four-candidates-20261022/verification.json

Die Sortierung „Weitere Prüfrunden“ zählt alle nachfolgenden Prüfungen zusammen. Die ersten fünf Runden bleiben separat.

10 von höchstens 10 Runden am 2026-10-20: Eigene deutsche Ausgabe mit vier Passaufgaben auf einer gemeinsamen Karte und zwei Start-Ziel-Paaren. Erklärte Pfeile, Rückwärtsschritte, Kürzen, Hinweise, Speichern und Teilen auf Deutsch; schmale Handyansicht geprüft. Die englische Ausgabe bleibt erhalten. Sprachliche Fairness und gefühlte Schwierigkeit brauchen menschliche Rückmeldung.

• conf/tasks/games-nonviral-fifth-ten-rounds-20261020/evidence/learning/ledger.json

4 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json

Neue und bisherige Aufgaben mit deutschsprachigen Ersttestern ohne vorausgesetzte Englischkenntnisse vergleichen: Regelverständnis, strittige Bedeutungen, Hilfen und freiwillige weitere Runde beobachten. Danach gezielt nachbessern.

Collect uncoached feedback on directed route construction, ordered stamps, shortening and physical phone controls.

Latest recorded source-informed local play/review; carried forward, not a fresh human or production playtest. Independently accepted for invited human playtesting on13 October2026. The loaded main heading is persistent; phone word/type labels and removal controls remain readable and contained. Final independent checks cover24 layouts across8 widths,8 real edits and128 native frames. One original owner Review activation remains unclassified; five exact owner/independent repeats passed without code changes. Human feedback and production validation remain pending.

Maturity gap: Four finite passport challenges reuse two endpoint pairs and one small authored graph; human learnability, enjoyment and return play are unverified.

Spring-hosted browser client; private Node 24+ rules runtime and durable per-site SQLite; anonymous cookie ownership is separate from KGARD accounts. Original cloud databases/cookies were not migrated. Build from either endpoint, each intermediate occurrence earns at most one stamp, repeat Object stamps require distinct occurrences; hints, legal removal and Silver-to-Gold shortening; history/draft stored locally. The maintained catalog now also states four challenges on two endpoint pairs. This earlier shared-site SQLite and cookie description applies to its social games; Sense Seals itself uses the stateless core API and browser-local progress. Two local review rounds completed on 13 October 2026. Sense Seals uses four finite English passports on two endpoint pairs. Its frozen graph has 160 word profiles and 53 authored edges, of which 52 meet the playable thresholds. Arrival moves follow incoming arrows. Ordered intermediate types, rather than the endpoints, supply the stamps; an occurrence can supply only one stamp. Optimal routes are four, four, five and five hops. Longer valid routes earn Silver, Bronze or Complete, and shortening retains the best route and paid hints. The first round observed stale-tab overwrites, writes over unreadable saved progress, refused-move history blocking valid continuation, and unavailable manual result text. Repairs check actual saved bytes before actions and replies, replay newer routes through the stateless service, distinguish fresh attempts, preserve unreadable or malformed records with an unsaved-practice warning, and provide current selectable result text. Two small test-control calibrations retained the original outcomes and assertions: waiting for a mounted request to finish, and using the actual terminal result when it replaces an old input. Fresh verification passed 54 checks and 10 browser cases. Manual play shortened a both-end Silver route to Gold, copied it, observed the second tab adopting it, and reloaded both pages. The second round found that older full saves containing only free refusals still blocked valid continuation, and a successful arrival-side public tool move left the next form pointing at departure. At overflow, the repair verifies actual saved prefixes before trimming only proved free refusals. Paid hints and meaningful moves remain. A full old record may require up to 161 additional sequential scoring checks. The accepted tool response now selects its actual side. Recovery interrupted by a genuine completed peer route or a failure to read browser storage preserves the correct record and permits later real play. Final verification passed 62 checks: 33 shared engine checks, 19 Sense Seals interface checks, two complete cold-storage journeys, six distinct second-round journeys and two prefix-recovery checks. All ten first manual copies matched the actual earned results. Fourteen desktop and phone-layout browser cases passed, including all four passports, all 14 ordered stamp requirements and all 52 directed neighbor labels. Two of those cases are historical first-stamp checks. Manual replay completed an entirely incoming Gold route through the public tool and ordinary form, then shortened an Aftercare Silver route to Gold, checked a missing Concept stamp, restored it, copied the results and reloaded. These are local agent checks and browser layouts; human learnability, physical phones and production remain unverified. All paid or meaningful 160-action histories still require restarting. Saved-byte checks are optimistic and do not provide an atomic transaction between pages. The finite authored vocabulary, repeated endpoint pairs and human engagement limits remain. The dated maturity rating and next step are retained.

• modules/games/cloud-games/meaning-games/lib/data.mjs
• modules/games/cloud-games/meaning-games/lib/engine.mjs
• modules/games/cloud-games/meaning-games/public/game.js
• modules/games/cloud-games/meaning-games/README.md
• modules/games/src/main/java/io/kgard/web/slice/games/CloudGamesRuntime.java
• modules/games/src/main/resources/cloud-games/server.mjs
• modules/games/docs/games-maturity-followup-20261010.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/wordport/round1-independent-review.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/wordport/round1-correction-review.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/wordport/final-root-transcript.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/wordport/round1-partial-reload.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/wordport/final-city-terminal-reload.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/wordport/final-verification.json
• conf/tasks/games-two-round-review-20261011/evidence/9-wordport/original-baseline-review.json
• conf/tasks/games-two-round-review-20261011/evidence/9-wordport/independent-original-review9.json
• conf/tasks/games-two-round-review-20261011/evidence/9-wordport/round1-review.json
• conf/tasks/games-two-round-review-20261011/evidence/9-wordport/independent-round1-review9.json
• conf/tasks/games-two-round-review-20261011/evidence/9-wordport/round2-before-review.json
• conf/tasks/games-two-round-review-20261011/evidence/9-wordport/independent-round2-before-review9.json
• conf/tasks/games-two-round-review-20261011/evidence/9-wordport/final-round2-review.json
• conf/tasks/games-two-round-review-20261011/evidence/9-wordport/round2-final-clients62.log
• conf/tasks/games-two-round-review-20261011/evidence/9-wordport/round2-final-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/9-wordport/round2-final-native-review.json
• conf/tasks/games-two-round-review-20261011/evidence/9-wordport/round2-final-delivery.json
• conf/tasks/games-two-round-review-20261011/evidence/9-wordport/readme-wordport-adjustment.json
• conf/tasks/games-two-round-review-20261011/evidence/9-wordport/independent-final-review9.json
• conf/tasks/games-two-round-review-20261011/evidence/9-wordport/root-final-acceptance.json
• conf/tasks/games-quality-20261013/results/batch-030-wordport-assessment.md
• conf/tasks/games-quality-20261013/results/verification-batch-030-wordport-layout.md

Manage a chip bank across four sealed-reserve route auctions. Pay for an audit, choose a bid or pass, then compare delivered value with your costs.

Deutsche Bedienung, Aufgaben, Rückmeldungen und Ergebnislinks im Starterpaket. Die ursprüngliche englische Ausgabe bleibt unter ihrer bisherigen Adresse erhalten; ihre Aufgaben und Spielstände sind getrennt.

English

Alle 9 Aufgaben sind heute ohne Wartezeit erreichbar. Davon 5 Übungen außerhalb der Tagesrotation; jede Aufgabe zählt nur einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards (8 repeating daily + 4 practice)

Solo

Limited beta ready

Planung und Wiederholungswert müssen sich im tatsächlichen Spielen bewähren.

Solo planning uses imaginary chips against authored sealed reserves; there are no live bidders, real currency or market simulation. Twelve English boards provide eight repeating daily auctions and four separate practice auctions. Each board has four finite route lots and editorial connection certificates. An audit costs 3 chips, one bid or pass settles a lot, a below-reserve bid spends nothing but consumes that opportunity, and an accepted uncertified route spends chips without delivery. Delivery arrives only at Close and cannot fund later bids. The final bank is unspent chips plus certified delivery; a balance below the initial 100 is a completed financial loss, not a separate lost phase or an earned point score. The displayed benchmark assumes perfect information and no audit costs; it is not another player. Reveal is unearned. Progress and bid drafts are browser-local with no account or cross-device synchronization; changed-save checks are not atomic between simultaneous tabs, and unavailable storage allows warned memory-only play. The mapped player has no live-data or custom-board mode. Retained provider-backed fixture code is unmapped and was not validated by these authored cycles. Human beta feedback is still needed on onboarding, semantic certificates, fairness, difficulty and replay interest; public production readiness remains unverified.

The13 October independent review accepted Path Auction after its private result focus was made immediately visible. All48 lots and16 distinct paths were reviewed; independent subset benchmarks agree. Ordinary daily/practice play,91 loss and164 win, upfront budget guards, true read-only outages and peer restoration passed. Accepted for invited human playtesting; human feedback remains pending. This is local technical evidence, not deployment or public-release signoff.

Authored route lots, connection certificates, sealed reserves, audit costs and delivery values.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

3 Verbesserungen umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-22. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-four-candidates-20261022/evidence/auction/ledger.json
• conf/tasks/games-four-candidates-20261022/verification.json

Die Sortierung „Weitere Prüfrunden“ zählt alle nachfolgenden Prüfungen zusammen. Die ersten fünf Runden bleiben separat.

1 von höchstens 10 Runden am 2026-10-20: Gebote, Abschlussauszahlung, Abbruch und Wiederholung geprüft. Kein zusätzlicher Fehler belegt.

• conf/tasks/games-nonviral-further-rounds-20261020/evidence/strategy/ledger.json

10 von höchstens 10 Runden am 2026-10-20: Zwei zusätzliche Übungen mit unterschiedlichen Kaufentscheidungen: Ein belegter Weg kann Verlust bringen; bezahlte Prüfungen begrenzen das verfügbare Gebotsbudget. Bisherige Aufgaben und Tagesauswahl bleiben erhalten. Sprachliche Fairness und gefühlte Schwierigkeit brauchen menschliche Rückmeldung.

• conf/tasks/games-nonviral-fourth-ten-rounds-20261020/evidence/strategy/ledger.json

3 frühere Prüfeinträge · 2 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json
• conf/tasks/games-nonviral-five-rounds-20261020/evidence/strategy-next/ledger.json

Neue und bisherige Aufgaben mit deutschsprachigen Ersttestern ohne vorausgesetzte Englischkenntnisse vergleichen: Regelverständnis, strittige Bedeutungen, Hilfen und freiwillige weitere Runde beobachten. Danach gezielt nachbessern.

Collect unprimed human feedback on comprehension, fairness, difficulty and replay interest; technical readiness and human feedback remain separate in conf/tasks/games-quality-20261013/game-quality.json.

Reassessed 11 October 2026 after two completed local play, review, improvement and replay cycles. Actual reviewed results included audited 152 before and after repair, a completed financial loss of 37 and fresh audited 148; browser journeys additionally completed actual 161, 164 and 160. Every route certificate, all sixteen connection readings, ledger, bank arithmetic, canonical Copy and active/terminal recovery were inspected. Lot-display/bid ownership, stale-action effects and peer/native focus were repaired without changing the authored economy, normalized saves or rules. Final verification passed 27 client checks, twelve fresh Java checks and twelve distinct desktop/phone browser journeys, with independent source review and bounded desktop/phone visual inspection. The second review required no further source change. Retains Limited beta ready, 4. The historical imaginary-chip, onboarding, fairness, difficulty and replay-interest gap remains open; local correctness does not establish human semantic quality, live market or provider behavior, or public production readiness.

Maturity gap: Solo planning uses imaginary chips; it does not simulate live multiplayer or an economic market. Human beta feedback is still needed on onboarding, fairness, difficulty and replay interest.

The mapped player uses PathAuctionController.playable, PathAuctionRoundService.generatePlayableRound, AuthoredGameDeck, path_auction_beta.jte, path-auction-beta.json, reserve-auction.js and path-auction-beta-rules.js. Inventory ID game-36 and persisted runtime ID game36 remain distinct. The authored-v1 envelope and reserve-auction-1 rules identity are unchanged. Recovery retains the tolerant normalized five-field contract: selected, bid, actions, phase and replay. No new required field, storage identity, scoring rule, authored reserve, route certificate or delivery value was introduced. Four audits and four lot settlements bound paid actions, so no event-cap repair was needed. The retained direct-call pathAuction method and generateRound fixture API can use cached Termpath or fallback routes, but they are unmapped. Language, source and target query parameters do not switch the mapped player to that legacy path. No external provider or custom-board quality was exercised. The first actual desktop cycle used pa-09. Root read the rules, checked that an empty bid was free, paid for all four audits and inspected all sixteen connection readings and four certificates. Lot 1 was passed. The selected Lot 2 bid of 46 survived active reload, then its accepted purchase locked 75 in delivery. The unprofitable certified Lot 3 was passed, and Lot 4 was bought for 42. Close produced an actual final bank of 152: four audits cost 12, accepted bids cost 88, and certified deliveries returned 152. Root reviewed all four ledger rows, bank and benchmark arithmetic, certification summaries and every connection, then copied the canonical spoiler-free result and recovered it after terminal reload. A bounded original desktop result was visually inspected. The pristine original client run passed 18 of 27 checks with nine targeted failures and zero skips. The original anchored desktop/phone browser run passed two of eight journeys, with six targeted failures in about 1.6 minutes; complete 152, 161 or 164 outcomes were reached before final defect verdicts. Ordinary peer Replay could leave terminal Lot 4 on screen while the saved selected lot had reset to Lot 1, so a real bid of 39 bought a different lot from the one displayed. Draft updates detached native lot tabs. Free terminal inspection did not reconcile a newer active ledger, and adopted or rejected actions could apply requested-state focus or leave focus on a hidden, disabled or detached control. Normal accepted bid and Close focus, failed-storage memory play and calibrated normalized legacy/startup recovery already passed, so those checks are compatibility evidence. The owning All games link opened Daily, and a rule sentence joined earns and 0. The repair limits local lot inspection to terminal review and clears it when an active attempt is accepted or adopted. Active display and bid ownership therefore use the saved selected lot. Lot buttons are retained, and unchanged bid values are not rewritten, preserving valid native focus and input position. Drafts, free refusals, terminal inspection, Next and Copy preflight the current save. Requested action effects use actual state acceptance, independently of whether durable storage succeeds, so warned memory play remains usable. Rejected actions adopt the latest state without claiming the rejected bid or audit. Focus reconciliation follows the actual active or terminal state and captures in-game focus before hidden manual Copy is cleared. Accepted practice Replay uses tryReplay; rejected Replay preserves current saved bytes and URL. Startup observation is aligned after shared initialization, while the original calibrated startup case remains a passing compatibility check. The template changes only the owning catalog link and earns 0 spacing. The normalized five fields, rules, authored deck and economy are unchanged. The original saved 152 survived the build. Root explicitly started a fresh repaired desktop Replay, paid all four audits, reviewed all sixteen readings, passed Lot 1, restored the raw Lot 2 bid of 46 after active reload, bought it, passed Lot 3 and bought Lot 4 for 42. Close again returned 152. All certificates, connection readings, four ledger rows, bank and benchmark were reviewed, copied and recovered, with a bounded final desktop visual inspection. Final focused verification passed all 27 client checks with zero skips in 707.1 milliseconds and all twelve fresh Java checks: five owning checks and seven inventory checks. All eight repaired anchored first-cycle browser journeys passed with zero skips in about 1.3 minutes, including complete audited 152, ordinary-peer 161, fresh no-audit 164, delayed action boundaries and warned memory/manual Copy. Independent review passed on the exact frozen runtime and template. Packaged and served sources matched the frozen runtime, with the then-current 37-review inventory included. The second actual cycle used pa-10 on a phone. Root paid four audits and inspected all sixteen connection readings and four certificates. Buying uncertified Lot 2 for 51 spent chips without delivery. A Lot 1 bid of 43 exceeded the remaining bank and was freely refused. A bid of 37 was within the bank but below the sealed reserve; it was declined without chip spend and settled that opportunity. Active reload retained these distinct facts. Passing Lots 3 and 4 and choosing Close yielded a genuine completed financial loss with 37 chips remaining. Root reviewed all four ledger rows, bank and benchmark, certificates and connection readings, then copied and recovered the result. This review established no additional defect and justified no further source, rule or economy change. Root then explicitly started a fresh phone Replay, bought all four audits and inspected every connection. A Lot 1 bid of 44 survived active reload. Buying certified Lots 1 and 3 for 44 and 40, and passing Lots 2 and 4, produced 148 at Close: 4 unspent chips plus 144 certified delivery. Every connection reading, all four certificates, four ledger rows, bank formula and perfect-information benchmark were reviewed, copied and recovered. The bounded phone result remained readable. The final four anchored second-cycle desktop/phone browser journeys passed with zero skips in 28.9 seconds, completing the 37-chip loss, recovered audited 148 and an actual fresh no-audit 160. The 160 is an executed final balance as well as the authored benchmark; the 148 actual audited balance is not relabelled as 160. The two frozen suites therefore provide twelve distinct passing owning browser journeys across both rounds. No source changed during the second cycle, so the final 27 client checks, twelve fresh Java checks and independent source review remain applicable without counting a repeated combined browser run. Final runtime SHA-256 is b02ee31ca6780dcd1722b35302fd4e29d1cbf96fe595d961da3872b2bb350fb2, and template SHA-256 is fde81edf95665118167df877e0faf8cc2e27d445d8804df16e8cc500c4bf626f. Rules, deck and frozen DOM/browser suites remained unchanged. The evidence is local agent gameplay, automated verification and source review. It does not establish human enjoyment, certificate fairness, live/custom-provider behavior, real-market behavior, deployment or public production readiness. Earlier known-issue record, retained for history: Resolved and locally verified 11 October 2026: terminal lot inspection is cleared when an active peer attempt is adopted, keeping the displayed lot aligned with the lot charged by a real bid. Native lot tabs remain stable during draft updates. Drafts, free refusals, terminal inspection, Next and Copy reconcile changed saves before acting; requested effects and Replay cleanup require actual acceptance. Invalid in-game focus after adoption returns to the current lot or completed result, including when manual Copy becomes hidden. Startup observation aligns with the recovered attempt; calibrated original startup already passed and is compatibility evidence, not an established original startup defect. The owning All games link now opens the catalog, and the rule spacing is corrected. Final replays established no further defect. Browser-local save reconciliation remains non-atomic.

• modules/games/src/main/java/io/kgard/web/slice/games/game36/PathAuctionController.java
• modules/games/src/main/resources/games/path-auction-beta.json
• modules/games/src/test/java/io/kgard/web/slice/games/game36/PathAuctionAuthoredDeckTest.java
• modules/games/docs/games-maturity-20261010-next-prototype-beta.json
• conf/tasks/evidence/next-ten-games-20261010/tactical
• conf/tasks/evidence/next-ten-games-20261010/final-verification.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/62-path-auction/verification.json
• conf/tasks/games-two-round-review-20261011/evidence/62-path-auction/round-1-before
• conf/tasks/games-two-round-review-20261011/evidence/62-path-auction/round-1-after
• conf/tasks/games-two-round-review-20261011/evidence/62-path-auction/round-2-final
• modules/games/src/main/java/io/kgard/web/slice/games/game36/PathAuctionRoundService.java
• modules/games/src/main/java/io/kgard/web/slice/games/game36/path_auction_beta.jte
• modules/games/src/main/resources/static/js/games/reserve-auction.js
• modules/games/src/main/resources/static/js/games/path-auction-beta-rules.js
• modules/games/src/test/js/tactical-beta-rules.test.cjs
• modules/games/src/test/js/reserve-auction-ui.test.cjs
• modules/web-visual/playful-games-tests/path-auction.spec.ts
• modules/web-visual/playful-games-tests/path-auction-round2.spec.ts
• conf/tasks/games-quality-20261013/results/verification-batch-012.md

Count shortest hops on a finite directed map, then commit a prediction and a chip stake. Three markets share one bank, with explicit returns and charged level hints.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben und Rückmeldungen. Die englische Ausgabe bleibt separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle vier Aufgaben sind heute als Übung erreichbar. Tagesaufgaben und Übungen nutzen denselben Vorrat. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards (8 repeating daily + 4 practice)

Solo

Limited beta ready

Simulated chips and twelve finite authored directed maps, with eight repeating daily and four practice boards. Shortest routes are proven only on the printed board; the arrows and association readings are authored rather than live semantic measurements. Exact counting makes the largest affordable stake dominant, so stake choice expresses uncertainty rather than a random market outcome. Progress is browser-local without account or cross-device synchronization; changed-save checks remain non-atomic preflights rather than simultaneous-write transactions. The default player route has no live-data mode; the retained older provider service was not exercised by these cycles. Unprimed human onboarding, counting difficulty, fairness, enjoyment, learning, retention and production behavior remain unverified.

Resolved and locally verified 11 October 2026: changed-save preflight and accepted-state checks prevent stale bets from reading nonexistent receipts or publishing settlement guidance for rejected actions. Peer adoption clears obsolete payout messages and repairs hidden, disabled or detached game focus while retaining valid visible focus. Accepted Hint moves focus to the persistent bought-levels container after its own button becomes disabled. Unchanged receipts and nested proofs retain their exact nodes, disclosure choices and focus; changed receipts replace only their audit and restore equivalent focused summaries. Replay cleanup and focus require acceptance, and queued terminal focus checks the current visible attempt. Changed malformed bytes are preserved, deleted saves remain playable, and failed writes retain warned in-memory progress for later recovery. Legacy loading still derives balances and receipts from legal events and preserves untouched arrival bytes. The final calculation now subtracts negative net returns explicitly. Engines, deck, keys, recovery contract and chip economy remain unchanged. Final local replays established no additional material defect. Independently accepted for invited human playtesting on13 October2026. Complete two-level counts already published by a prior same-source proof or paid hint are free to review. Newly played sb04 earns157 without a duplicate charge; genuine historical154/77 receipts retain their exact charges.64 client and20 browser checks plus independent information, native viewport and storage checks pass. Human feedback and production validation remain pending.

Authored finite directed maps; shortest-hop predictions and simulated chip stakes.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

Eine Verbesserung umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-22. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-eighth-ten-rounds-20261022/evidence/strategy/ledger.json

6 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json

Mit deutschsprachigen Ersttestern Verständnis, strittige Antworten, genutzte Hilfen und eine freiwillige weitere Runde beobachten. Mit anderen Varianten derselben Richtung vergleichen; menschliche Rückmeldung steht aus.

Collect uncoached human feedback on estimates, stakes, shared-bank decisions and replay variety.

Reassessed 11 October 2026 after two completed local play, review, improvement and replay cycles. Full three-market files finished at 160, 117, 74 and zero chips, covering exact counting, rounded refunds, paid hints, passing, exhaustion, shortest-route proofs, bank audits, public Copy and active/terminal recovery. Reproduced stale guidance, focus loss, collapsing proofs and a rejected-bet receipt error were repaired; the final net-return wording was clarified without changing scoring. Thirty-four client checks passed, fourteen Java checks were fresh in cycle one, eight first-cycle and four final second-cycle browser cases passed, and packaging succeeded. Independent review found no blocker. Retains Limited beta ready: twelve finite authored maps and the dominant maximum stake under perfect counting still need unprimed human difficulty, fairness and repeat-interest evidence. Browser-local non-atomic saves and local checks do not establish live semantic measurements or production readiness. Independently accepted for invited human playtesting on13 October2026. Complete two-level counts already published by a prior same-source proof or paid hint are free to review. Newly played sb04 earns157 without a duplicate charge; genuine historical154/77 receipts retain their exact charges.64 client and20 browser checks plus independent information, native viewport and storage checks pass. Human feedback and production validation remain pending.

Maturity gap: Perfect counting makes the maximum stake dominant; the confidence choice is meaningful for uncertain players, not a random market simulation. The puzzle proves a shortest route only on its finite printed map. Human beta feedback is still needed on unprimed onboarding, difficulty, perceived fairness and repeat interest. The finite authored bank and device-local progress are stated limits; production is not verified.

The first completed local cycle played practice sb-09, Ocean connections. All ten printed directed-arrow readings were inspected. The two-hop, twenty-chip draft survived active reload; exact settlement returned sixty for a net forty and bank eighty. Three-hop and six-hop predictions at twenty chips then settled banks of 120 and 160. All three shortest witnesses, breadth levels and bank audits, canonical public Copy and terminal reload were reviewed. Actual peer play reproduced Copy focus falling to the page body after Replay. After the first exact settlement, a peer hop selection collapsed an opened nested receipt proof and detached its focused summary. Buying the second-market Hint charged three chips, reducing eighty to seventy-seven, and disabled its focused button without a handoff. A peer Pass kept seventy-seven and advanced to the third market, but the receiving tab retained obsolete Exact plus forty guidance. The final six-hop exact settlement completed that secondary file for 117, with all proofs, readings and the eighty minus three hint minus zero stake plus zero return audit inspected. The repair adds game-scoped changed-save checks, accepted-state effects, Replay gating, current guidance and focus, and guarded delayed result focus. Receipt nodes are retained when unchanged; changed audits are replaced in place while preserving both disclosure levels and equivalent summary focus. Existing derived recovery still recomputes forged summary values from legal chronology and leaves compatible legacy arrival bytes untouched. Independent rules review found no economy or reachable history-cap defect: free choices add no events, and at most three hints plus three settlements fit the recovery bound. Controlled complete browser journeys reproduced a stale later-market Bet adopting a peer Replay and then trying to read an absent receipt; acceptance gating now prevents that error. The actual repaired replay preserved the focused open nested proof across a peer hop change, focused bought levels after Hint, and moved to the visible third-market heading when peer Pass hid the hint. Old payout guidance cleared. The complete file again earned 117 and retained its proofs, audits, Copy and terminal result. Thirty-four focused client checks, comprising fifteen engine and nineteen UI cases, fourteen fresh Java checks, packaging and a design scan with no findings passed. Eight complete desktop/phone first-cycle browser cases passed after the repair, including full 160, 120 and 100 results, delayed Bet into a fresh Replay followed by a complete 160 result, saved-byte and URL preservation, and no browser errors. The second completed cycle played sb-10, Star connections, after inspecting all ten readings. Four hops and a five-chip stake were selected before buying a three-chip Hint, leaving thirty-seven. Active reload retained the affordable draft and bought levels. The actual shortest route was three hops; the one-away prediction returned two, rounded down from half of five, and the bank became thirty-four. Pass kept thirty-four, then the exact six-hop twenty-chip final bet earned forty net and finished at seventy-four. All three shortest witnesses, breadth levels, bank audits, public Copy and terminal recovery were inspected. Actual phone Replay then used two eight-hop, twenty-chip misses to reduce forty to twenty to zero, followed by a free Pass to complete a zero-bank loss. The final calculation initially printed a plus sign followed by a negative forty; a narrow wording change now displays explicit subtraction with the absolute amount, leaving arithmetic unchanged. The final actual phone replay repeated the complete zero-bank file and verified the readable calculation, all proofs and audits, Copy and reload. All thirty-four client checks remained green, and four final second-cycle desktop/phone browser cases passed after that wording change, giving twelve distinct cases across both cycles. The fourteen Java checks were fresh in cycle one. Independent review found no blocker in the runtime repair or final wording. These are source-informed local agent plays and automated checks, not human, live-provider or production validation.

• modules/games/src/main/java/io/kgard/web/slice/games/game69/StepBetsController.java
• modules/games/src/main/resources/games/play/step-bets.json
• modules/games/src/test/java/io/kgard/web/slice/games/game69/StepBetsAuthoredDeckTest.java
• modules/games/docs/games-maturity-20261010-third-prototype-beta.json
• conf/tasks/evidence/third-ten-games-20261010/prediction
• conf/tasks/evidence/third-ten-games-20261010/final-verification.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/82-step-bets/round-1
• conf/tasks/games-two-round-review-20261011/evidence/82-step-bets/round-2
• modules/games/src/main/resources/static/js/games/step-bets-beta.js
• modules/games/src/main/resources/static/js/games/step-bets-beta-rules.js
• modules/games/src/test/js/prediction-beta-rules.test.cjs
• modules/games/src/test/js/step-bets-beta-ui.test.cjs
• modules/web-visual/playful-games-tests/step-bets.spec.ts
• modules/web-visual/playful-games-tests/step-bets-round2.spec.ts
• conf/tasks/games-quality-20261013/results/batch-035-step-bets-paid-count.md
• conf/tasks/games-quality-20261013/results/verification-batch-035-step-bets-paid-count.md

Leite zwei Wortquellen durch begrenzte Stationen. Liefere WELLE und MUSIK, bevor Zeit oder Energie ausgehen.

Ohne Englischkenntnisse spielbar.

Deutsche Oberfläche und Wörter

Eigene und geteilte Runden werden unabhängig gespeichert.

Versioniertes, endliches redaktionelles Spielpaket mit erklärten Wortbeziehungen und eigenen Spielregeln.

Eigene Partie und teilbare Ausgangslage

Playable prototype

Endlicher Wortschatz, keine Live-Contextoid-Abfrage. Browseridentitäten sind keine verifizierten Personen; Tempovergleiche hängen auch von Gerät und Verbindung ab.

Agenten-Spieltests belegen keine menschliche Beliebtheit oder Verbreitung.

Versioniertes, endliches redaktionelles Spielpaket mit erklärten Wortbeziehungen und eigenen Spielregeln.

Teilbare Ausgangslagen und eigene Spielentscheidungen.

Spring prüft Züge; PostgreSQL speichert Partien, Beiträge, Ergebnisse und Wiederholungsbelege.

Menschliche Rückmeldungen zu Regeln, Wortschatz und Weitergabe einholen.

Mit unabhängigen Spielern Verständlichkeit, Spielspaß und Wiederholungswert prüfen.

Umsetzung eines eigenständigen Konzepts aus Word Game Ideas; lokale Spielzyklen werden separat belegt.

Maturity gap: Unabhängige menschliche Spieltests zu Verständlichkeit, Spielspaß und Zugänglichkeit stehen aus.

Ergänzung zur historischen 99-Spiele-Prüfung; deren Befunde bleiben erhalten. Am 25. Oktober 2026 wurden 4 neue lokale Browser-Spielzyklen geprüft, mit zwei abwechslungsreichen fehlerfreien Schlussrunden nach der letzten Korrektur. Das ist kein menschlicher Beliebtheitsnachweis; Bereitstellungen haben eigene datierte Nachweise.

• modules/games/docs/remaining-word-games.md
• conf/tasks/remaining-word-games-20261024/source-thread.md
• conf/tasks/remaining-word-games-20261024/assoziationsfabrik/play-review-summary.json
• conf/tasks/word-games-improvement-20261025/strategy/completion.json
• conf/tasks/word-games-improvement-20261025/README.md
• modules/games/docs/releases.md

Verteile Worttürme auf drei Wegen und fange vier angekündigte Wellen ab.

Ohne Englischkenntnisse spielbar.

Deutsche Oberfläche und Wörter

Eigene und geteilte Runden werden unabhängig gespeichert.

Versioniertes, endliches redaktionelles Spielpaket mit erklärten Wortbeziehungen und eigenen Spielregeln.

Eigene Partie und teilbare Ausgangslage

Playable prototype

Endlicher Wortschatz, keine Live-Contextoid-Abfrage. Browseridentitäten sind keine verifizierten Personen; Tempovergleiche hängen auch von Gerät und Verbindung ab.

Agenten-Spieltests belegen keine menschliche Beliebtheit oder Verbreitung.

Versioniertes, endliches redaktionelles Spielpaket mit erklärten Wortbeziehungen und eigenen Spielregeln.

Teilbare Ausgangslagen und eigene Spielentscheidungen.

Spring prüft Züge; PostgreSQL speichert Partien, Beiträge, Ergebnisse und Wiederholungsbelege.

Menschliche Rückmeldungen zu Regeln, Wortschatz und Weitergabe einholen.

Mit unabhängigen Spielern Verständlichkeit, Spielspaß und Wiederholungswert prüfen.

Umsetzung eines eigenständigen Konzepts aus Word Game Ideas; lokale Spielzyklen werden separat belegt.

Maturity gap: Unabhängige menschliche Spieltests zu Verständlichkeit, Spielspaß und Zugänglichkeit stehen aus.

Ergänzung zur historischen 99-Spiele-Prüfung; deren Befunde bleiben erhalten. Am 25. Oktober 2026 wurden 5 neue lokale Browser-Spielzyklen geprüft, mit zwei abwechslungsreichen fehlerfreien Schlussrunden nach der letzten Korrektur. Das ist kein menschlicher Beliebtheitsnachweis; Bereitstellungen haben eigene datierte Nachweise.

• modules/games/docs/remaining-word-games.md
• conf/tasks/remaining-word-games-20261024/source-thread.md
• conf/tasks/remaining-word-games-20261024/assoziationswacht/play-review-summary.json
• conf/tasks/word-games-improvement-20261025/strategy/completion.json
• conf/tasks/word-games-improvement-20261025/README.md
• modules/games/docs/releases.md

Baue ein Wortnetz mit Schaltern. Teste den Signalfluss und veröffentliche eine nachweislich lösbare Schalteraufgabe.

Ohne Englischkenntnisse spielbar.

Deutsche Oberfläche und Wörter

Eigene und geteilte Runden werden unabhängig gespeichert.

Versioniertes, endliches redaktionelles Spielpaket mit erklärten Wortbeziehungen und eigenen Spielregeln.

Eigene Partie und teilbare Ausgangslage

Playable prototype

Endlicher Wortschatz, keine Live-Contextoid-Abfrage. Browseridentitäten sind keine verifizierten Personen; Tempovergleiche hängen auch von Gerät und Verbindung ab.

Agenten-Spieltests belegen keine menschliche Beliebtheit oder Verbreitung.

Versioniertes, endliches redaktionelles Spielpaket mit erklärten Wortbeziehungen und eigenen Spielregeln.

Teilbare Ausgangslagen und eigene Spielentscheidungen.

Spring prüft Züge; PostgreSQL speichert Partien, Beiträge, Ergebnisse und Wiederholungsbelege.

Menschliche Rückmeldungen zu Regeln, Wortschatz und Weitergabe einholen.

Mit unabhängigen Spielern Verständlichkeit, Spielspaß und Wiederholungswert prüfen.

Umsetzung eines eigenständigen Konzepts aus Word Game Ideas; lokale Spielzyklen werden separat belegt.

Maturity gap: Unabhängige menschliche Spieltests zu Verständlichkeit, Spielspaß und Zugänglichkeit stehen aus.

Ergänzung zur historischen 99-Spiele-Prüfung; deren Befunde bleiben erhalten. Am 25. Oktober 2026 wurden 4 neue lokale Browser-Spielzyklen geprüft, mit zwei abwechslungsreichen fehlerfreien Schlussrunden nach der letzten Korrektur. Das ist kein menschlicher Beliebtheitsnachweis; Bereitstellungen haben eigene datierte Nachweise.

• modules/games/docs/remaining-word-games.md
• conf/tasks/remaining-word-games-20261024/source-thread.md
• conf/tasks/remaining-word-games-20261024/bedeutungsmaschine/play-review-summary.json
• conf/tasks/word-games-improvement-20261025/strategy/completion.json
• conf/tasks/word-games-improvement-20261025/README.md
• modules/games/docs/releases.md

Plane neun Bauplätze. Drei verschiedene Wohnorte brauchen passende Nachbarn und ein gemeinsames Budget.

Ohne Englischkenntnisse spielbar.

Deutsche Oberfläche und Wörter

Eigene und geteilte Runden werden unabhängig gespeichert.

Versioniertes, endliches redaktionelles Spielpaket mit erklärten Wortbeziehungen und eigenen Spielregeln.

Eigene Partie und teilbare Ausgangslage

Playable prototype

Endlicher Wortschatz, keine Live-Contextoid-Abfrage. Browseridentitäten sind keine verifizierten Personen; Tempovergleiche hängen auch von Gerät und Verbindung ab.

Agenten-Spieltests belegen keine menschliche Beliebtheit oder Verbreitung.

Versioniertes, endliches redaktionelles Spielpaket mit erklärten Wortbeziehungen und eigenen Spielregeln.

Teilbare Ausgangslagen und eigene Spielentscheidungen.

Spring prüft Züge; PostgreSQL speichert Partien, Beiträge, Ergebnisse und Wiederholungsbelege.

Menschliche Rückmeldungen zu Regeln, Wortschatz und Weitergabe einholen.

Mit unabhängigen Spielern Verständlichkeit, Spielspaß und Wiederholungswert prüfen.

Umsetzung eines eigenständigen Konzepts aus Word Game Ideas; lokale Spielzyklen werden separat belegt.

Maturity gap: Unabhängige menschliche Spieltests zu Verständlichkeit, Spielspaß und Zugänglichkeit stehen aus.

Ergänzung zur historischen 99-Spiele-Prüfung; deren Befunde bleiben erhalten. Am 25. Oktober 2026 wurden 6 neue lokale Browser-Spielzyklen geprüft, mit zwei abwechslungsreichen fehlerfreien Schlussrunden nach der letzten Korrektur. Das ist kein menschlicher Beliebtheitsnachweis; Bereitstellungen haben eigene datierte Nachweise.

• modules/games/docs/remaining-word-games.md
• conf/tasks/remaining-word-games-20261024/source-thread.md
• conf/tasks/remaining-word-games-20261024/bedeutungsstadt/play-review-summary.json
• conf/tasks/word-games-improvement-20261025/strategy/completion.json
• conf/tasks/word-games-improvement-20261025/README.md
• modules/games/docs/releases.md

Kombiniere sechs Grundbegriffe und entdecke die drei Ziele deiner Mission.

Ohne Englischkenntnisse spielbar.

Deutsche Oberfläche und Wörter

Eigene und geteilte Runden werden unabhängig gespeichert.

Versioniertes, endliches redaktionelles Spielpaket mit erklärten Wortbeziehungen und eigenen Spielregeln.

Eigene Partie und teilbare Ausgangslage

Playable prototype

Endlicher Wortschatz, keine Live-Contextoid-Abfrage. Browseridentitäten sind keine verifizierten Personen; Tempovergleiche hängen auch von Gerät und Verbindung ab.

Agenten-Spieltests belegen keine menschliche Beliebtheit oder Verbreitung.

Versioniertes, endliches redaktionelles Spielpaket mit erklärten Wortbeziehungen und eigenen Spielregeln.

Teilbare Ausgangslagen und eigene Spielentscheidungen.

Spring prüft Züge; PostgreSQL speichert Partien, Beiträge, Ergebnisse und Wiederholungsbelege.

Menschliche Rückmeldungen zu Regeln, Wortschatz und Weitergabe einholen.

Mit unabhängigen Spielern Verständlichkeit, Spielspaß und Wiederholungswert prüfen.

Umsetzung eines eigenständigen Konzepts aus Word Game Ideas; lokale Spielzyklen werden separat belegt.

Maturity gap: Unabhängige menschliche Spieltests zu Verständlichkeit, Spielspaß und Zugänglichkeit stehen aus.

Ergänzung zur historischen 99-Spiele-Prüfung; deren Befunde bleiben erhalten. Am 25. Oktober 2026 wurden 7 neue lokale Browser-Spielzyklen geprüft, mit zwei abwechslungsreichen fehlerfreien Schlussrunden nach der letzten Korrektur. Das ist kein menschlicher Beliebtheitsnachweis; Bereitstellungen haben eigene datierte Nachweise.

• modules/games/docs/remaining-word-games.md
• conf/tasks/remaining-word-games-20261024/source-thread.md
• conf/tasks/remaining-word-games-20261024/wortschmiede/play-review-summary.json
• conf/tasks/word-games-improvement-20261025/arcade/completion.json
• conf/tasks/word-games-improvement-20261025/README.md
• modules/games/docs/releases.md

Überstehe vier Tage mit knappen Wortvorräten. Sammeln, Herstellen und tägliche Bedürfnisse konkurrieren um dieselben Mittel.

Ohne Englischkenntnisse spielbar.

Deutsche Oberfläche und Wörter

Eigene und geteilte Runden werden unabhängig gespeichert.

Versioniertes, endliches redaktionelles Spielpaket mit erklärten Wortbeziehungen und eigenen Spielregeln.

Eigene Partie und teilbare Ausgangslage

Playable prototype

Endlicher Wortschatz, keine Live-Contextoid-Abfrage. Browseridentitäten sind keine verifizierten Personen; Tempovergleiche hängen auch von Gerät und Verbindung ab.

Agenten-Spieltests belegen keine menschliche Beliebtheit oder Verbreitung.

Versioniertes, endliches redaktionelles Spielpaket mit erklärten Wortbeziehungen und eigenen Spielregeln.

Teilbare Ausgangslagen und eigene Spielentscheidungen.

Spring prüft Züge; PostgreSQL speichert Partien, Beiträge, Ergebnisse und Wiederholungsbelege.

Menschliche Rückmeldungen zu Regeln, Wortschatz und Weitergabe einholen.

Mit unabhängigen Spielern Verständlichkeit, Spielspaß und Wiederholungswert prüfen.

Umsetzung eines eigenständigen Konzepts aus Word Game Ideas; lokale Spielzyklen werden separat belegt.

Maturity gap: Unabhängige menschliche Spieltests zu Verständlichkeit, Spielspaß und Zugänglichkeit stehen aus.

Ergänzung zur historischen 99-Spiele-Prüfung; deren Befunde bleiben erhalten. Am 25. Oktober 2026 wurden 5 neue lokale Browser-Spielzyklen geprüft, mit zwei abwechslungsreichen fehlerfreien Schlussrunden nach der letzten Korrektur. Das ist kein menschlicher Beliebtheitsnachweis; Bereitstellungen haben eigene datierte Nachweise.

• modules/games/docs/remaining-word-games.md
• conf/tasks/remaining-word-games-20261024/source-thread.md
• conf/tasks/remaining-word-games-20261024/begriffslager/play-review-summary.json
• conf/tasks/word-games-improvement-20261025/worlds/completion.json
• conf/tasks/word-games-improvement-20261025/README.md
• modules/games/docs/releases.md

Translate three German clues in English contexts, choosing exact or lower-scoring context answers from reviewed word banks.

Deutsche Bedienung und Erklärungen; englische Kontexte und Antworten gehören zur Lernaufgabe. Englischkenntnisse sind erforderlich. Dies ist keine rein deutschsprachige Wortübung.

English / German

Alle 8 Aufgaben sind heute als Übung erreichbar. Davon 4 zusätzliche Übungen außerhalb der Tagesrotation. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards (8 repeating daily + 4 practice)

Solo

Limited beta ready

Bounded German→English exercises with twelve finite authored files, eight repeating daily and four practice; each card's six-word bank, exact answer, aliases and lower-scoring English-context connectors are pre-authored. Plausible alternative associations outside that bank receive no credit, and the exercise does not provide arbitrary translation or global semantic correctness. Browser-local progress has no account or cross-device synchronization; changed-save checks remain non-atomic preflights rather than simultaneous-write transactions. These cycles exercised the authored edition, not the retained experimental custom-language/provider behavior. Native-speaker judgment, unprimed human sense discrimination, difficulty, enjoyment, learning and retention, and production behavior remain unverified.

Resolved and locally verified 11 October 2026: adopting another tab's card selection clears obsolete miss guidance while preserving still-enabled control focus. Peer Replay moves focus from hidden Copy/results to the visible current card. Action feedback and focus require acceptance of the requested state, so a rejected correct guess cannot be described as a charged miss. Rejected Replay preserves the newer saved result, audit disclosures and URL instead of applying reset effects; accepted Replay remains usable when storage is unavailable. Game-scoped changed-save checks precede free refusals and controls, retain changed malformed bytes, and handle adopted deletion without repeatedly rejecting subsequent play. Temporary failed writes retain warned in-memory progress and can save the whole attempt after recovery. Delayed Copy cannot restore obsolete feedback. The strict state schema, saved identity, rules, scoring and committed receipts remain unchanged. Final local replays established no additional material defect. Independently accepted for invited human playtesting on13 October2026. Native Replay exposes the full first context card and guess field below the sticky desk; exact and contextual translation grades remain distinct. Independent036 confirms exact content/scoring, genuine saved histories and native input/ending boundaries. Initial reduced-motion reload transients clear within153ms; human feedback and production validation remain pending.

Authored German meaning clues, English endpoint contexts and graded answer banks.

Supplying language selects the retained Termpath experiment: Games Data API cached paths with rank/weight/node dimensions, falling back to practice data.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

5 Verbesserungen umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-20. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-ten-rounds-20261020/evidence/learning/ledger.json

Die Sortierung „Weitere Prüfrunden“ zählt alle nachfolgenden Prüfungen zusammen. Die ersten fünf Runden bleiben separat.

2 von höchstens 10 Runden am 2026-10-20: Antwort, Auflösung und ausgeschöpfte Versuche sind unterscheidbar. Punkte und Hinweise geprüft; keine weitere Änderung begründet.

• conf/tasks/games-nonviral-further-rounds-20261020/evidence/learning/ledger.json

4 frühere Prüfeinträge · 1 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json
• conf/tasks/games-nonviral-five-rounds-20261020/evidence/learning/ledger.json

Neue und bisherige Aufgaben mit deutschsprachigen Ersttestern vergleichen: Verständnis, strittige Antworten, genutzte Hilfen und freiwillige weitere Runde beobachten. Passende Englischkenntnisse sind erforderlich. Die neuen Aufgaben sind noch nicht durch Teilnehmer validiert.

Collect uncoached human feedback on the finite authored content, difficulty and phone comfort.

Reassessed 11 October 2026 after two completed local play, review, improvement and replay cycles. Three-card files earned 2150 and 1250, with exact/context credit, contextual aliases, paid nudges and misses, free repeated guesses, exhaustion, receipts, public Copy and active/terminal recovery. Reproduced stale peer guidance, focus loss and rejected Submit/Replay effects were repaired; changed invalid/deleted saves and failed-write memory were checked without changing the engine or saved identity. Forty-three focused client checks, sixteen focused Java checks, twelve distinct desktop/phone browser cases and packaging passed; cycle two needed no further source change. Retains Limited beta ready: twelve finite German/English banks still need native-speaker and unprimed human evidence for interpretation, difficulty, enjoyment and learning. Browser-local non-atomic recovery, unexercised experimental custom/provider behavior and local checks do not establish production readiness. Independently accepted for invited human playtesting on13 October2026. Native Replay exposes the full first context card and guess field below the sticky desk; exact and contextual translation grades remain distinct. Independent036 confirms exact content/scoring, genuine saved histories and native input/ending boundaries. Initial reduced-motion reload transients clear within153ms; human feedback and production validation remain pending.

Maturity gap: Finite authored content; native-speaker validation, human novice sense discrimination and learning/retention remain unverified

The first completed local cycle played practice bb-09, Context crossings. Sheet was a miss for the Tree/Green sense of das Blatt, costing 50; the English nudge cost 100 without using a guess, and the padded Leaves alias draft recovered on active reload. Leaf then earned 850, Latch was an English-context connector for Door/Security earning 300, and Ladder was the exact Roof/Climb translation earning 1000. All three cards settled for 2150/3000; Copy, terminal reload, the actual costs and all candidate readings were inspected. Actual two-tab review reproduced Copy focus falling to the page body on peer Replay, and Sheet/Tree/Green miss guidance remaining when the peer selected the Door/Security card, although enabled Submit focus was correctly preserved. The repaired actual replay cleared that old guidance, finished Lock, Ladder and Leaf for 2950 with the committed 50-point miss retained, Copied and restored the terminal file, and handed focused phone Copy to the visible Card 1 heading on peer Replay. Controlled delayed-notification browser play also confirmed a stale correct Leaf Submit adopts the peer's committed Sheet miss without adding a guess or falsely reporting Leaf as a charged miss; the adopted file was then completed for 2950. Rejected Replay from an older 3000-point result preserved the newer 2900-point complete file, paid nudge, exact saved bytes and old reveal URL, and focused the visible result. The scoped repair distinguishes accepted state from durable writes, preflights changed saves before free actions, gates Replay effects, clears obsolete peer guidance and repairs only hidden, disabled or detached game focus. A root review found an empty-snapshot observation gap; its regression first failed and the repair now permits a fresh 3000-point file after peer save deletion. Existing receipt node identity, disclosure choices, literal tested readings and the strict seven-key engine remained unchanged. Forty-three focused client checks, comprising ten engine and thirty-three UI cases, sixteen focused Java checks, eight desktop/phone browser cases and packaging passed in cycle one. The second completed cycle played bb-10, Words with two uses: padded Leaves correctly resolved to the Writing/Book Sheet sense for 1000; Bank was a miss for Park/Rest, its canonical repeat was free, and Chair connector credit earned 250. Feather and Map misses on Mattress/Bounce survived active reload with one guess left; Pear exhausted the third card for zero. All three cards settled for 1250/3000. The three audits exposed all eighteen candidate readings and six committed receipts with truthful costs and no negative card points; public Copy and terminal reload retained 1250. Actual 390 px phone Replay settled cards in order 3, 1, 2 and again earned 1250, with usable controls, wrapped public links and no horizontal overflow. Four additional desktop/phone cases passed, including full in-memory play with unavailable storage and a selected manual Copy fallback, giving twelve distinct browser cases across both cycles. No additional source change was warranted in cycle two. These are source-informed local agent plays and automated checks, not native-speaker, human enjoyment, live-provider or production validation.

• modules/games/src/main/java/io/kgard/web/slice/games/game72/BilingualBridgeController.java
• modules/games/src/main/resources/games/bilingual-bridge-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game72/BilingualBridgePlayRouteTest.java
• modules/games/src/test/java/io/kgard/web/slice/games/game72/BilingualBridgeAuthoredDeckTest.java
• modules/games/docs/games-maturity-followup-20261010.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-72/round1-reviewer-source-notes.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-72/round1-independent-review.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-72/final-transcript.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-72/final-play-checks.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-72/final-verification.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/85-bilingual-bridge/round-1
• conf/tasks/games-two-round-review-20261011/evidence/85-bilingual-bridge/round-2
• modules/games/src/main/resources/static/js/games/bilingual-bridge.js
• modules/games/src/test/js/bilingual-bridge.test.cjs
• modules/games/src/test/js/bilingual-bridge-ui.test.cjs
• modules/web-visual/playful-games-tests/bilingual-bridge.spec.ts
• conf/tasks/games-quality-20261013/results/batch-036-authored-arrivals.md
• conf/tasks/games-quality-20261013/results/verification-batch-036-authored-arrivals.md

Ein fehlender Bestandteil verbindet drei Begriffe. Löse ihn und mache einen Außenknoten zum nächsten Rätsel.

Auf Deutsch bedienbar und ohne Englischkenntnisse spielbar.

Deutsche Oberfläche; deutsche und englische Rätsel

Neue und abgeleitete Runden speichern eigene Beiträge und Ergebnisse.

15 vorbereitete deutsche und englische Rätsel mit drei Hinweisen und geprüften Fortsetzungen.

Eigenes Rätsel und teilbare Herausforderung

Playable prototype

Endlicher redaktioneller Spielwortschatz. Gespeicherte Runden benötigen den aktivierten Spring/PostgreSQL-Dienst; Browseridentitäten sind keine verifizierten Personen.

Agenten-Playtests belegen keine menschliche Beliebtheit oder Verbreitung.

15 vorbereitete deutsche und englische Rätsel mit drei Hinweisen und geprüften Fortsetzungen.

Spieler ergänzen, lösen oder verändern die gespeicherte Aufgabe.

Spring prüft alle Züge; PostgreSQL speichert Runden, Beiträge, Ergebnisse und Abstammung.

Menschliche Rückmeldungen zu Regeln, Wortschatz und Weitergabe einholen.

Mit unabhängigen Spielern Verständlichkeit, Spielspaß und Weitergabe prüfen.

Neue Umsetzung des Chats Spielkonzepte für Wortspiele; lokale Prüfung wird separat dokumentiert.

Maturity gap: Unabhängige menschliche Spieltests zu Verständlichkeit, Spielspaß und Zugänglichkeit stehen aus.

Ergänzung zur historischen 99-Spiele-Prüfung; deren Befunde bleiben erhalten. Vierte lokale Verbesserung am 25. Oktober 2026: 4 weitere geprüfte Spielzyklen, insgesamt 27. Menschlicher Spielspaß und Produktionsbetrieb werden dadurch nicht belegt; Bereitstellungen haben eigene datierte Nachweise.

• modules/games/docs/compound-word-games.md
• modules/games/src/main/java/io/kgard/web/slice/games/compound/BlindknotenGame.java
• conf/tasks/compound-word-games-20261024/continuation.md
• conf/tasks/compound-word-games-20261024/round-4/continuation.md
• conf/tasks/compound-word-games-20261024/round-4/iteration-counts.json
• modules/games/docs/releases.md

Connect five authored endpoint pairs using six-word banks, with accepted and top answers explained separately.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben und Rückmeldungen. Die englische Ausgabe bleibt separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle vier Aufgaben sind heute als Übung erreichbar. Tagesaufgaben und Übungen nutzen denselben Vorrat. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards (8 repeating daily + 4 practice)

Solo

Limited beta ready

Top versus merely accepted bridges follows an assigned rubric, not measured public agreement. Each finite authored session contains five endpoint pairs and six candidates per pair; eight repeating daily boards and four practice boards reuse twenty pair files. Both endpoint links are required for acceptance. An accepted candidate’s base score comprises 40 points for each authored endpoint link and 20 for the requested role, with five points deducted per prior distinct miss; a third distinct miss closes that pair at zero. Reveal also settles a pair at zero, while unknown and repeated guesses are free. These explanations establish the authored model rather than unrestricted semantic equivalence or measured live graph paths. Progress remains browser-local without account or cross-device synchronization, and changed-save protection is a non-atomic preflight. Unavailable storage permits warned memory-only play. Supplying language still selects the retained cached-Termpath experiment and its practice fallback; this separate custom branch and live-provider availability were not validated by the authored play cycles. Human fairness, difficulty, learning, unprimed onboarding, replay interest, enjoyment and public production readiness remain unverified.

Resolved and locally verified 11 October 2026: free and paid actions, Copy and Replay reconcile changed saves and apply requested effects only to accepted state. Native pair tabs, candidate banks, paid receipts and audit disclosures remain stable; peer changes clear obsolete terminal inspection and rescue hidden or disabled focus to the actual active pair, settled file or completed result. Deferred alignment and Copy scrolling are guarded against replaced state or inspection. The calibrated original startup check passed; observer alignment preserves that compatibility and is not evidence of an original startup-stranding bug. The second cycle justified no additional source change. Final replays established no further defect. Independently accepted for invited human playtesting on13 October2026 after the private restored-result repair. Full title and receipt clear the sticky desk across320/390/1440 and reduced-motion320; actual earned/revealed reload, native Replay, old receipts, read-only outages and interrupted arrival pass. Initial brief transients remain qualified; human feedback and production/provider validation are pending.

Authored endpoint pairs, six-word banks and two-sided/precision scoring rubrics.

Supplying language selects the retained Termpath experiment: Games Data API cached paths with rank/weight/node dimensions, falling back to practice data.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

Eine Verbesserung umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-20. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-seventh-ten-rounds-20261020/evidence/words/ledger.json

3 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json

Mit deutschsprachigen Ersttestern Verständnis, strittige Antworten, genutzte Hilfen und eine freiwillige weitere Runde beobachten. Mit anderen Varianten derselben Richtung vergleichen; menschliche Rückmeldung steht aus.

Collect uncoached human feedback on this finite authored mechanic, phone use, difficulty and repeat interest.

Reassessed 11 October 2026 after two completed local play, review, improvement and replay cycles. Full 475-point attempts before and after repair, a mixed 255 and fresh 500 were reviewed with all five outcomes, all ninety rubric readings, paid diagnostics, canonical Copy and active/terminal recovery. Current-save acceptance, native focus, inspection and deferred effects were repaired. The calibrated original startup check passed and is compatibility evidence; the second cycle justified no further source edit. Final verification passed 41 client checks, seventeen fresh Java checks and twelve distinct desktop/phone browser journeys, plus packaging, independent source review and two bounded readable visual reviews. Retains Limited beta ready, 4. The historical need for human fairness, difficulty, learning and replay-interest evidence remains open. Local correctness does not establish unrestricted semantic agreement, human enjoyment, custom/live-provider behavior or public production readiness. Independently accepted for invited human playtesting on13 October2026 after the private restored-result repair. Full title and receipt clear the sticky desk across320/390/1440 and reduced-motion320; actual earned/revealed reload, native Replay, old receipts, read-only outages and interrupted arrival pass. Initial brief transients remain qualified; human feedback and production/provider validation are pending.

Maturity gap: Twelve sessions reuse20 finite pair files; requested roles and binary authored links constrain discovery. Human fairness, difficulty, learning and replay interest remain unvalidated.

The mapped default Bridge route uses BridgeDailyController, bridge_daily_play.jte, bridge-daily-deck.json and bridge-daily.js. Its runtime and inventory identity are both game-52; shared authored-v1 storage uses rulesVersion bridge-daily-1 and strict seven-field state recovery. The default game reads finite authored endpoint pairs, candidate banks and left-link, right-link and requested-role rubrics. Supplying language selects the separate retained Termpath experiment, which may use cached paths or practice fallback; day or puzzle cannot be combined with that branch. This branch was preserved, not exercised as live-provider evidence. The first actual cycle explicitly replayed an existing local 495 result before counting a fresh zero-event original baseline. All five prompts, all thirty candidate words and the complete finite scoring contract were read. An unknown word and repeated Lens were free. Lens committed one one-sided miss with three paid diagnostic readings. The raw PHOTO alias and paid receipt survived active reload, then Photograph earned 95. sheep_wool earned 100, Pottery earned an accepted 80 below the top Kiln answer, Umbrella earned 100 and overall_strategy earned 100. All five pairs settled at 475/500. All five outcome rows, all ninety rubric readings, canonical Copy and terminal recovery agreed. The original client run reported 31 passes and ten targeted failures across 41 checks: ten pure checks, seventeen existing DOM checks and fourteen new runtime checks. The calibrated exact-main-key startup check passed on the original single-read runtime by permitting adoption on the first genuine Check; original startup stranding was not demonstrated. An initial browser selection accidentally included a later Pathbridge edition because the filename filter was unanchored. Root stopped and discarded that entire run, assigned no completion credit, and reran only the exact current Bridge specification. The retained original eight desktop/phone journeys all reached full outcomes before verdicts; four passed and four targeted cases failed in 58.4 seconds. The repair checks the current save before free and paid actions, Copy and Replay, and applies requested effects only when the attempted state was accepted. Save acceptance remains separate from persistence success, preserving warned memory-only play. Native pair tabs, per-pair candidate banks and audit nodes are retained. Paid receipt nodes and open disclosures remain stable. Peer adoption clears obsolete terminal inspection and restores useful focus when the former target disappears or becomes disabled. Rejected actions follow the actual active pair, settled file or completed result. Queued alignment and late Copy scrolling are fenced against changed state or inspection. The new observer is aligned after shared recovery as compatibility protection. Pure scoring and exports, strict seven-field recovery, the score economy, paid receipt chronology, all ten original pure and seventeen existing DOM test bodies, and the optional custom-language branch remain unchanged. The fresh repaired baseline repeated the complete 475-point session, including unknown and repeated free guesses, the Lens diagnostic, raw PHOTO recovery, all thirty candidate words, all ninety rubric readings, canonical Copy and terminal recovery. The first bounded desktop review found a readable result without overflow. All 41 client checks and seventeen fresh Java checks passed with no failures or skips. Java coverage comprised four authored-deck, three retained service, three route and seven inventory checks. All eight repaired first-cycle desktop/phone journeys passed in 49.2 seconds. The second actual cycle started a fresh phone workshop at zero events and 0/500. Treble clef was one paid miss with three diagnostics before Beat earned 75; Stamp earned 100. Leaf and Steam locomotive were two distinct Tea misses. The raw Skateboard draft and both paid receipts survived active reload; the third distinct miss closed Tea at zero. Ocean Reveal settled another pair at zero, and Computer earned an accepted 80. The complete result was 255/500: three bridged, one revealed and one closed after misses. All ninety rubric readings, every paid diagnostic, five outcome rows, canonical Copy and terminal recovery agreed. Review found no further reproducible defect, so no second source edit was justified. A fresh explicit phone Replay reset all five pairs to zero events and 0/500. Rhythm and Stamp each earned 100. The raw TEA__KETTLE draft survived active reload, then Tea kettle, LIGHT__HOUSE and Server each earned 100. The completed 500/500 result retained all five top-answer rows, all ninety rubric readings, canonical Copy and terminal recovery. Four second-cycle desktop/phone browser journeys passed in 29.6 seconds, giving twelve distinct final browser journeys across both rounds. The same frozen runtime passed 41 client checks and seventeen fresh Java checks. Independent source review passed, and the static design detector returned no findings. The bounded phone result review found readable score and outcome rows, with the canonical URL wrapping without horizontal overflow; remaining rows were available below the fold. Source, packaged and served runtime hashes agreed, and the package matched the then-current 32-row refreshed canonical inventory. Final runtime SHA-256 is 228697791f8b0fdb29b0f8c38d32295f5e1009d84b261831ac9895b139b9d236. This is local source-informed agent and automated evidence; no human, live-provider or deployment validation is claimed.

• modules/games/src/main/java/io/kgard/web/slice/games/game52/BridgeDailyController.java
• modules/games/src/main/resources/games/bridge-daily-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game52/BridgeDailyAuthoredDeckTest.java
• modules/games/src/test/java/io/kgard/web/slice/games/game52/BridgeDailyPlayRouteTest.java
• modules/games/docs/games-maturity-followup-20261010.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-52/round1-proposed-content-corrections.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-52/round1-content-correction-applied.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-52/round1-independent-review.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-52/round1-correction-review.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-52/unchanged-edition-recovery.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-52/changed-edition-transition.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-52/round1-root-transcript.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-52/final-transcript.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-52/final-play-checks.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-52/final-verification.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/67-bridge-daily/verification.json
• conf/tasks/games-two-round-review-20261011/evidence/67-bridge-daily/round-1-before-exact
• conf/tasks/games-two-round-review-20261011/evidence/67-bridge-daily/round-1
• conf/tasks/games-two-round-review-20261011/evidence/67-bridge-daily/round-2
• modules/games/src/main/java/io/kgard/web/slice/games/game52/bridge_daily_play.jte
• modules/games/src/main/resources/static/js/games/bridge-daily.js
• modules/games/src/test/js/bridge-daily.test.cjs
• modules/games/src/test/js/bridge-daily-ui.test.cjs
• modules/games/src/test/js/bridge-daily-runtime.test.cjs
• modules/web-visual/playful-games-tests/bridge-daily.spec.ts
• modules/web-visual/playful-games-tests/bridge-daily-round2.spec.ts
• conf/tasks/games-quality-20261013/results/batch-029-route-restoration.md
• conf/tasks/games-quality-20261013/results/verification-batch-029-route-restoration.md

Choose two bank ideas and a shared span, then balance support, load and a construction budget. Several stable builds are possible.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben und Rückmeldungen. Die englische Ausgabe bleibt separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle vier Aufgaben sind heute als Übung erreichbar. Tagesaufgaben und Übungen nutzen denselben Vorrat. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards: eight rotating UTC daily + four separate practice. Nine building cards; select two supports and one shared span.

Solo

Limited beta ready

Neu aus den weiteren Spielen für den deutschen Vergleich ausgewählt. Als Variante dieser Richtung prüfen; Nachfrage und eigenständiger Ausbau sind noch nicht belegt.

Twelve repeating authored English boards; support, load and costs are designed constraints, not real term-path search or semantic strength. Saves are device-local and non-atomic, without account sync. Human fairness, difficulty, enjoyment, retention, live-provider quality and production readiness remain unverified.

The13 October independent review accepted Bridgeforge after its bounded scene and focus corrections. Soapdish removes the ordinary parking-permit ambiguity in two authored scenes while genuine prior results remain intact. First completion and accepted Replay keep the actual result/explanation or first native choice visible. The exact native Replay failure now passes at320/390/1440 through six seconds, followed by honest efficient-cost6 completion and exact saved reload. All12 current board identities and production classes are verified. Accepted for invited human playtesting; human feedback remains pending. This is local technical evidence, not deployment or public-release signoff.

Authored contextual left-bank/right-bank/span roles and assigned cost/support/load values.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

Eine Verbesserung umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-20. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-sixth-ten-rounds-20261020/evidence/strategy/ledger.json

7 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json

Mit deutschsprachigen Ersttestern Verständnis, strittige Antworten, genutzte Hilfen und eine freiwillige weitere Runde beobachten. Mit anderen Varianten derselben Richtung vergleichen; menschliche Rückmeldung steht aus.

Collect unprimed human feedback on comprehension, fairness, difficulty and replay interest; technical readiness and human feedback remain separate in conf/tasks/games-quality-20261013/game-quality.json.

Latest recorded source-informed local play/review; carried forward, not a fresh human or production playtest.

Maturity gap: Support and costs are designed constraints; finite themes and perceived plausibility need human feedback. Human beta feedback is still needed on onboarding, fairness, difficulty and replay interest.

The mapped /games/game13 player uses BridgeforgeAuthoredDeck, bridgeforge_beta.jte and bridgeforge-beta.js, with runtime identity game13 and rules forge-beta-1. Twelve authored English boards provide eight rotating UTC daily puzzles and four separate practices. Each supplies nine material cards: choose distinct left-bank, span and right-bank ideas that fit their scene roles, support the span load and stay within budget. Efficient and Stable are successful categories based on construction cost, not earned point scores or measured semantic strength. Three different failed constructions genuinely end unfinished; incomplete and repeated tests are free, Clear retains paid attempts, and Reveal is unscored. Retained provider/controller methods remain unmapped. Original native Farming play, fields-sky, completed an Efficient construction at cost 6, support 6 and load 6 after partial draft recovery, then a distinct Stable construction at cost 7, support 6 and load 3. Each review read all nine material scenes and values, three paid positional reasons, all four valid constructions and the cheapest example, with actual canonical Copy and terminal recovery. Desktop measured 1280 by 720 with page width 1265; phone measured 390 by 844 with page width 375. Native peer completion and Replay stranded focus on the document body. An exploratory incomplete Test also wrote a competing revision; a subsequently rejected selection was not counted as a completed construction. Original verification passed 33 of 41 clients with eight owning failures in 717.201208 milliseconds. Two of eight original desktop/phone browser journeys passed and six failed in 16.7 seconds; full constructions and all eight captures preceded the deferred boundary verdicts. Failures established stale instructions or success feedback after rejected actions, and invalid focus or Replay effects. Ordinary browser guidance already passed even though the separate DOM ordering fixture failed. Existing startup, scoring constraints, canonical Copy and accepted memory outcomes are not newly discovered defects. Three full-outcome regressions were appended after baseline closure and failed against the unchanged original in 424.311208 milliseconds. Cold unreadable storage with working writes overwrote an unknown genuine three-test loss through memory Efficient play and Replay, while hiding the storage warning; subsequent recovery handling and durable Stable cost 7 were completed before the aggregate verdict. Durable initial Reveal already copied correctly on its first request; that case failed only the known accepted-Replay focus expectation. The third case captured a focused construction summary being removed by peer Replay, then completed and recovered Stable cost 7 before its focus verdict. All original fifteen DOM cases and helpers remain an exact 18,309-byte prefix. The owning repair reconciles compatible saved state before requested actions and Copy. State identity controls feedback and follow-up effects separately from durability, so rejected requests do not announce success while failed writes can still support warned memory play. Existing readable-save protection preserves unknown bytes when reads are unavailable. Compatible observation aligns initialization and reconciliation; accepted Replay alone applies reset effects. Unchanged free Test results skip unnecessary save revisions. Focus ownership is captured before render and before the shared storage listener, preserving usable native controls and rescuing removed or disabled targets to the actual active select or result. The complete render, material/menu creation, tolerant five-field recovery, chronology, daily replay rule, costs, support/load constraints, content and shared persistence are unchanged. LocalStorage protection remains non-atomic. Repaired native Farming Efficient 6 and Stable 7 repeated the complete evidence, actual Copy and recovery. Peer completion now focused the visible result; Replay from a removed proof summary focused forge-left. All 44 final clients passed in 716.530417 milliseconds: eighteen owning DOM, thirteen owning pure, one mixed chronology and twelve Axis compatibility checks. Eight final first-round browser journeys passed in 26.2 seconds. Eleven fresh original Java checks passed at 15:17:30.807 through 15:17:32.265 UTC: two mixed mapped authored checks, three retained unmapped controller checks and six retained service checks. Their unchanged evidence is retained without claiming a final Java rerun or future Axis gameplay. The second native round used Language, words-code, on the measured phone. Three disjoint failed constructions exposed all nine paid positional reasons; the first exceeded budget 8 at cost 9. Exact repeat was free, Clear retained the attempt, and reload after two failures preceded the genuine third-test loss. All nine material scenes, four alternatives, loss Copy and terminal recovery were reviewed. Fresh Dictionary/Parser/Processor earned Efficient cost 7, support 7 and load 6 after complete-draft recovery. No further source change was justified. Fresh Dictionary/Machine translation/Processor then earned Stable cost 8 with the same support and load, full evidence, actual Copy and recovery; bounded phone content remained readable. Four final second-round browser journeys passed in 23.3 seconds. The final total is twelve distinct owning journeys and sixteen completed captures, none unavailable, on identical repaired source; earlier runs are additional executions, not additional distinct cases. Final runtime SHA-256 is 63d234476121473c5477899a1b12b7472d26b11d1a496f7e8b8b7100870dc354. Independent source review and the twenty-seven-input audit confirm only runtime and the three appended DOM cases changed. The immutable local classpath contains 596 entries and 17,868,319 copied bytes; served source and packed sixty-four-review base inventory agree. This is not a fat archive, deployment or live-provider check. Browser play covers two boards and eighteen distinct material scenes; all twelve boards and 108 material rows were source-reviewed, with pure checks covering affordable constructions. Maturity 4, its 10 October date, basis and historical gap remain unchanged. Finite repeating content, designed constraints, non-atomic device-local saves without account sync, and unverified human fairness, difficulty, enjoyment, retention and production remain limitations. Earlier known-issue record, retained for history: The 11 October review repaired rejected-action feedback, native focus, redundant free-Test saves and unreadable-save overwrite. Both local cycles passed; no further material defect was established. Maturity remains 4.

• modules/games/src/main/java/io/kgard/web/slice/games/game13/BridgeforgeController.java
• modules/games/src/main/resources/games/play/bridgeforge.json
• modules/games/src/test/java/io/kgard/web/slice/games/game13/BridgeforgeControllerTest.java
• modules/games/src/test/java/io/kgard/web/slice/games/game13/BridgeforgeRoundServiceTest.java
• modules/games/docs/games-maturity-20261010-next-prototype-beta.json
• conf/tasks/evidence/next-ten-games-20261010/root
• conf/tasks/evidence/next-ten-games-20261010/final-verification.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/35-bridgeforge/baseline.json
• conf/tasks/games-two-round-review-20261011/evidence/35-bridgeforge/actual-round1.json
• conf/tasks/games-two-round-review-20261011/evidence/35-bridgeforge/actual-round2.json
• conf/tasks/games-two-round-review-20261011/evidence/35-bridgeforge/original-client.log
• conf/tasks/games-two-round-review-20261011/evidence/35-bridgeforge/original-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/35-bridgeforge/original-browser-outcomes.json
• conf/tasks/games-two-round-review-20261011/evidence/35-bridgeforge/original-additional-regressions.log
• conf/tasks/games-two-round-review-20261011/evidence/35-bridgeforge/additional-regressions.json
• conf/tasks/games-two-round-review-20261011/evidence/35-bridgeforge/original-java-results.json
• conf/tasks/games-two-round-review-20261011/evidence/35-bridgeforge/round1-proposed-runtime.diff
• conf/tasks/games-two-round-review-20261011/evidence/35-bridgeforge/independent-path-review-round1.json
• conf/tasks/games-two-round-review-20261011/evidence/35-bridgeforge/round1-client.log
• conf/tasks/games-two-round-review-20261011/evidence/35-bridgeforge/round1-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/35-bridgeforge/round2-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/35-bridgeforge/final-browser-outcomes.json
• conf/tasks/games-two-round-review-20261011/evidence/35-bridgeforge/final-hashes.json
• conf/tasks/games-two-round-review-20261011/evidence/35-bridgeforge/classpath-round1-package.json
• conf/tasks/games-two-round-review-20261011/evidence/35-bridgeforge/verification.json
• modules/games/src/main/resources/static/js/games/bridgeforge-beta.js
• modules/games/src/main/resources/static/js/games/bridgeforge-rules.js
• modules/games/src/test/js/bridgeforge-ui.test.cjs
• modules/web-visual/playful-games-tests/bridgeforge.spec.ts
• modules/web-visual/playful-games-tests/bridgeforge-round2.spec.ts
• modules/web-visual/playful-games-tests/bridgeforge-support.ts
• conf/tasks/games-quality-20261013/results/verification-batch-014-forge-replay-followup.md

Verbinde benachbarte englische Wörter. Lange Ketten schaffen Platz für neue Ideen.

Deutsch bedienbar; englische Wörter werden zum Spielen benötigt.

Deutsche Oberfläche; englischer Wortschatz

Eigene und geteilte Runden werden unabhängig gespeichert.

Versioniertes, endliches redaktionelles Spielpaket mit erklärten Wortbeziehungen und eigenen Spielregeln.

Eigene Partie und teilbare Ausgangslage

Playable prototype

Endlicher Wortschatz, keine Live-Contextoid-Abfrage. Browseridentitäten sind keine verifizierten Personen; Tempovergleiche hängen auch von Gerät und Verbindung ab.

Agenten-Spieltests belegen keine menschliche Beliebtheit oder Verbreitung.

Versioniertes, endliches redaktionelles Spielpaket mit erklärten Wortbeziehungen und eigenen Spielregeln.

Teilbare Ausgangslagen und eigene Spielentscheidungen.

Spring prüft Züge; PostgreSQL speichert Partien, Beiträge, Ergebnisse und Wiederholungsbelege.

Menschliche Rückmeldungen zu Regeln, Wortschatz und Weitergabe einholen.

Mit unabhängigen Spielern Verständlichkeit, Spielspaß und Wiederholungswert prüfen.

Umsetzung eines eigenständigen Konzepts aus Word Game Ideas; lokale Spielzyklen werden separat belegt.

Maturity gap: Unabhängige menschliche Spieltests zu Verständlichkeit, Spielspaß und Zugänglichkeit stehen aus.

Ergänzung zur historischen 99-Spiele-Prüfung; deren Befunde bleiben erhalten. Am 25. Oktober 2026 wurden 6 neue lokale Browser-Spielzyklen geprüft, mit zwei abwechslungsreichen fehlerfreien Schlussrunden nach der letzten Korrektur. Das ist kein menschlicher Beliebtheitsnachweis; Bereitstellungen haben eigene datierte Nachweise.

• modules/games/docs/remaining-word-games.md
• conf/tasks/remaining-word-games-20261024/source-thread.md
• conf/tasks/remaining-word-games-20261024/chain-rush/play-review-summary.json
• conf/tasks/word-games-improvement-20261025/arcade/completion.json
• conf/tasks/word-games-improvement-20261025/README.md
• modules/games/docs/releases.md

Three trails lead to one word. Choose which clues to uncover.

Oberfläche, Regeln und vorbereitete Aufgaben benötigen derzeit Englisch. Die deutsche Beschreibung übersetzt das Spiel noch nicht.

English

Tagesrätsel plus sechs separate Übungsrätsel; dieselben Spuren erscheinen bei Wiederholung erneut.

18 authored boards (12 repeating daily + 6 separate practice)

Solo + friend challenge

Limited beta ready

The finite English bank contains twelve repeating daily boards and six separate practice boards; daily frames can be strongly revealing. Practice difficulty, clue usefulness, vocabulary inclusion, enjoyment and sustained return play still need unprimed human study. Progress and familiarity remain browser-local with no account or cross-device synchronization. Changed-save checks are a non-atomic preflight and do not provide a simultaneous-write transaction.

Resolved and locally verified 11 October 2026: changed saved attempts are adopted before stale actions can replace them; an optional revision distinguishes separate completed replays even when their scores and gameplay fields match. Only accepted Replay removes the reveal URL trigger, preserving other parameters and the hash. Opening a compatible older round no longer rewrites its scored save. Damaged JSON and incompatible saves are preserved, failed writes retain warned in-memory play, and delayed Copy feedback is discarded after the attempt changes. The game header now opens the complete catalog. The repaired local replays established no further defect. Independently accepted for invited human playtesting on 13 October 2026: Distinct paid final hillside clue preserves price and actual old90/content history; new90 reloads separately. Verification052 passes71 clients,12 Java and66 browser cases with native,history and recovery checks. Human difficulty,enjoyment and replay feedback remain pending.

Authored three-lane polysemy clues, candidate words, staged paid clues and explanations.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

Den tatsächlichen Weg vom geteilten Ergebnis zum spielenden Empfänger prüfen; keine unbestätigten Wachstumswerte ableiten.

Collect unprimed feedback on progressive clues, sense ambiguity and finite-bank replay.

Reassessed 11 October 2026 after two completed local play, review, improvement and replay cycles. Actual 75-point solve, three-miss zero and six-clue third-guess 30-point solve exercised clue costs, recovery and copying. Stale-result overwrite, reveal-URL Replay and identical-result attempt identity were repaired. Cycle one passed 33 engine/runtime checks, ten desktop/phone browser cases, focused Java checks and packaging; cycle two added two browser journeys against unchanged source. Retains Limited beta ready: the eighteen-board authored bank still needs unprimed human evidence for clue usefulness, vocabulary inclusion, difficulty, fairness and sustained return play. Independently accepted for invited human playtesting on 13 October 2026: Distinct paid final hillside clue preserves price and actual old90/content history; new90 reloads separately. Verification052 passes71 clients,12 Java and66 browser cases with native,history and recovery checks. Human difficulty,enjoyment and replay feedback remain pending.

Maturity gap: The 12-day bank plus six separate practice boards is enough for a limited beta, but difficulty, vocabulary inclusion and sustained return play still need unprimed human evidence.

The first source-informed local cycle on cv-13 explicitly replayed an existing result, checked Duty incorrectly, bought one Before-a-journey clue and restored a Charge draft on reload. Checking Charge earned 75 from two guesses and one paid clue; Copy and terminal reload retained that result. Actual ordinary two-tab play then reproduced a stale Duty selection replacing a completed 100-point solve, and a separate reveal=true URL caused a replayed draft to reload as a zero-point reveal. After repair, Replay removed only the reveal parameter; the paid-clue Charge draft, subsequent 75-point solve and copied/reloaded result survived, and another tab adopted the same result. A controlled regression also established that two completed 100-point replays need distinct save revisions. Cycle one passed 33 engine/runtime checks, ten complete desktop/phone browser cases, focused game80 and inventory Java checks, and packaging; the browser cases included 75/95-point recovery, changed-result protection and blocked storage. The second cycle on phone used cv-14: Crew, Company and Brace exhausted three distinct guesses for zero, with terminal recovery and no automatic clue charge. After Replay, all six clues were purchased; Crew and Company missed, then a Cast draft survived active reload with 30 points available. Checking Cast, Copy and terminal reload retained the third-guess 30-point solve. Two additional desktop/phone journeys passed for this loss and six-clue replay boundary. No further source change was justified in cycle two; the first-cycle checks apply to the unchanged final implementation. These are local source-informed agent results, not independent human fairness, difficulty, enjoyment or production-readiness evidence.

• modules/games/src/main/java/io/kgard/web/slice/games/game80/ConvergenceController.java
• modules/games/src/main/resources/games/convergence-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game80/ConvergenceSecurityWebMvcTest.java
• modules/games/docs/games-maturity-20261005.json
• modules/games/src/main/java/io/kgard/web/slice/games/game80/ConvergenceRoundService.java
• modules/games/src/main/java/io/kgard/web/slice/games/game80/convergence.jte
• modules/games/src/main/resources/static/js/games/convergence.js
• conf/tasks/evidence/games-maturity-20261005/termpath-b-game-80-phone.png
• conf/tasks/evidence/games-maturity-20261005/termpath-b-browser.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/91-convergence/round-1
• conf/tasks/games-two-round-review-20261011/evidence/91-convergence/round-2
• modules/games/src/test/js/convergence.test.cjs
• modules/games/src/test/js/convergence-runtime.test.cjs
• modules/web-visual/playful-games-tests/convergence.spec.ts
• conf/tasks/games-quality-20261013/results/batch-052-convergence-forensics-followup.md
• conf/tasks/games-quality-20261013/results/verification-batch-052-convergence-forensics.md

One missing word. Use the words around it to complete the connection.

Oberfläche, Regeln und vorbereitete Aufgaben benötigen derzeit Englisch. Die deutsche Beschreibung übersetzt das Spiel noch nicht.

English

Mehrere früher veröffentlichte Aufgaben sind am selben Tag als Übung verfügbar; sie sind kein zusätzlicher unabhängiger Vorrat.

12 authored boards; 12-board repeating daily; ordinary practice excludes today’s board (11 available during this review)

Solo + friend challenge

Limited beta ready

Twelve authored boards repeat daily; ordinary practice navigation excludes today’s board and exposed eleven choices during this review. The finite daily/practice pool does not provide unlimited new puzzles. Each board has six explained choices, one or two accepted central links, three distinct checked tries and a free hint; success and tries are reported without a points score. Authored validity is limited to the printed chain and explanations, not measured corpus proximity or unrestricted synonym acceptance. Practice Replay keeps the first result separately. Progress is browser-local without account or cross-device synchronization; changed-save protection is a non-atomic preflight. Unreadable or incompatible storage can leave a warned memory-only attempt that will not survive reload. The default route never calls the retained provider constructor. Unprimed-human fairness, return engagement, a larger difficulty-balanced schedule and public production readiness remain unverified.

The13 October independent review accepted Cluehinge unchanged after ordinary play, both optical answers Lens/Mirror, free hints, real three-miss loss, replay and exact first-result preservation. All12 boards and72 choices were reviewed; real read-only-save recovery and native normal-motion endings passed. Game61 is a separate implementation. Accepted for invited human playtesting; human feedback remains pending. This is local technical evidence, not deployment or public-release signoff.

Authored five-word chains with one central gap and six explained choices.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

Den tatsächlichen Weg vom geteilten Ergebnis zum spielenden Empfänger prüfen; keine unbestätigten Wachstumswerte ableiten.

Collect unprimed human feedback on comprehension, fairness, difficulty and replay interest; technical readiness and human feedback remain separate in conf/tasks/games-quality-20261013/game-quality.json.

Reassessed 11 October 2026 after two completed local play, review, improvement and replay cycles. Hinted Tendon in two tries, explicit fresh Lens and Mirror one-try attempts, preserved first-result benchmarks and genuine three-miss loss were checked with every accepted and rejected explanation, Copy and active/terminal recovery. The older Mirror result present on arrival was excluded from fresh-play credit. Scoped save, Replay, focus, guidance, Copy and catalog repairs passed 36 fresh client checks, eighteen fresh Java checks, packaging and twelve distinct final desktop/phone browser journeys; independent review found no blocker. Retains Limited beta ready, 4. The historical finite-board, unprimed-human fairness, return-engagement and difficulty-balanced scheduling gap remains open. Local correctness and reliability do not establish human enjoyment, live-provider behavior or public production readiness.

Maturity gap: Twelve repeating boards support a limited beta; unprimed-human fairness, return engagement and a larger difficulty-balanced content schedule are not evidenced.

The mapped default route is the game57 missing_link.jte template and missing-link.js custom client, with inventory identity game-57 and persisted runtime identity missing-link. It uses an authored five-word chain with one central gap and six explained choices; no live provider is called. Twelve authored boards repeat daily. Source and actual navigation confirm that ordinary practice excludes today’s board: the original label was Puzzle 8 of 11. The earlier look-ahead assumption that all twelve boards were offered in practice was corrected without changing release policy. The first complete actual local cycle played ml-09, Movement → Muscle → gap → Bone → Skeleton. A checked Ligament miss left two tries and explained its bone-to-bone role. The free hint left tries unchanged, but native focus fell to the page. Selected Tendon, the checked miss and the hint survived active reload; Check earned Link found in 2/3 tries with Hint used. Both Tendon connection readings and all five rejected-choice reasons were reviewed, and terminal recovery retained the result. Copy exposed the preview origin, and the owning All games link opened the daily list. The original root client run passed nineteen of 33 checks, with fourteen targeted failures. All eight original desktop/phone browser journeys reached complete earned Tendon original and Replay outcomes, full explanations, recovery and first-result review before their targeted failure verdicts. The reveal-loop failure was captured before a clean full replay earned the intended result. The scoped owning repair now reads the current save before every action and Copy; adopts compatible or deleted records; preserves changed invalid bytes; and distinguishes equal-looking attempts with optional flat attemptRevision metadata. Durable writes are gated while the current save is unreadable, with warned memory play still usable. This gate also fixed a reproduced candidate-stage read-outage regression that could otherwise overwrite an unknown or changed-invalid record. Replay changes the reveal URL and flag only after acceptance and retains the original result. Native choice and explanation nodes remain stable; accepted free Hint focuses its visible explanation; obsolete guidance clears and unavailable focus is rescued after adoption. Canonical Copy uses current request, state and save checks and provides a selected labelled readonly textarea when automatic copying is denied. The owning catalog destination is corrected. The pure engine/export prefix is byte-identical; the existing nine logical state fields and normalized recovery, stored identity, authored content, three-try limit and free-hint economy remain unchanged. Independent frozen source review found no material blocker, and the static design check returned no findings. Fresh verification passed 36 client checks: nine pure-engine, six legacy VM and twenty-one DOM checks. Eighteen fresh Java checks passed: three rendered-page, eight round-service and seven inventory checks. These client and Java runs had no failures or skips. Source, packaged and served runtime hashes agreed, and packaged findings matched the then-current 27-row refreshed canonical inventory. All eight repaired first-cycle browser journeys passed in 25.6 seconds. Fresh actual repaired ml-09 Replay kept the original hinted two-try benchmark. Ordinary peer Replay cleared old Copy guidance and moved hidden Copy focus to the visible Cartilage choice. Ligament miss, free Hint with visible explanation focus, selected Tendon and active reload retained two tries and the hint. Check completed Replay: link found in 2/3 tries with Hint used. Both readings, all five rejection reasons, the first-result benchmark, canonical Copy, terminal recovery and catalog destination agreed. Root reviewed desktop and phone complete-result captures as readable and usable without overflow. The second actual cycle opened ml-12, Sand → Glass → gap → Telescope → Stars, on the phone. Arrival contained an older completed Mirror one-try, no-hint benchmark; that arrival was not counted as fresh play. Explicit Replay started a new three-try attempt. All six choices and the chain were read, selected Lens survived active reload, and Check earned Replay: link found in 1/3 tries with No hint. All four accepted left/right readings and all four rejected-choice reasons were reviewed. The older first-result benchmark remained separate, and canonical Copy and terminal recovery agreed. Review established no new scoped defect, so the verified runtime and template were retained unchanged. A fresh explicit Mirror Replay then repeated the full choice, chain, active-recovery and explanation review and earned the alternate one-try, no-hint result. The original benchmark remained separate through Copy and terminal reload. All four second-cycle desktop/phone browser cases passed in 10.4 seconds. Their isolated coverage additionally preserved an original Lens snapshot through Mirror Replay and exercised a genuine three-miss memory-only loss followed by a Mirror solve, selected labelled manual Copy and honest unsaved reload. A bounded phone capture review confirmed readable explanations, fallback and storage warning. All 36 final client checks passed again. Across both cycles, twelve distinct final browser journeys passed. Runtime SHA-256 remained 3b6cbd84a1fd328dfe78f31d22ea0cfda037f9922b9dc3098cdd119d2ea9ef8e throughout the final replay checks. This is source-informed local agent and automated evidence, with no human, live-provider or deployment validation claimed. Earlier known-issue record, retained for history: Resolved and locally verified 11 October 2026: the custom client now adopts compatible or deleted peer saves and rejects stale actions without overwriting current progress or the first-result benchmark. Changed invalid saves are preserved, and durable writes wait until the current save can be read. Accepted Replay removes only the reveal parameter and clears its one-time flag; rejected Replay preserves saved progress and the URL. Native choice and explanation controls stay stable, obsolete guidance clears after adoption, and free Hint or hidden-control focus returns to visible content. Copy uses the public Contextoid URL, fences delayed feedback against a newer attempt, and offers a selected labelled readonly fallback. The owning All games link opens the catalog. Final replays established no further defect.

• modules/games/src/main/java/io/kgard/web/slice/games/game57/MissingLinkController.java
• modules/games/src/main/resources/games/missing-link-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game57/MissingLinkRenderedTest.java
• modules/games/docs/games-maturity-20261005.json
• modules/games/src/main/java/io/kgard/web/slice/games/game57/MissingLinkRoundService.java
• modules/games/src/main/java/io/kgard/web/slice/games/game57/missing_link.jte
• conf/tasks/evidence/games-maturity-20261005/termpath-a-browser-observations.json
• conf/tasks/evidence/games-maturity-20261005/termpath-a-play-observations.json
• modules/games/src/main/resources/static/js/games/missing-link.js
• modules/games/src/main/java/io/kgard/web/slice/games/release/GameDailySchedule.java
• conf/tasks/evidence/games-maturity-20261005/termpath-a-screens/game-57-390.png
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/72-the-missing-link/verification.json
• conf/tasks/games-two-round-review-20261011/evidence/72-the-missing-link/round-1
• conf/tasks/games-two-round-review-20261011/evidence/72-the-missing-link/round-2
• modules/games/src/test/js/missing-link.test.cjs
• modules/games/src/test/js/missing-link-runtime.test.cjs
• modules/games/src/test/js/missing-link-ui.test.cjs
• modules/games/src/test/java/io/kgard/web/slice/games/game57/MissingLinkRoundServiceTest.java
• modules/web-visual/playful-games-tests/the-missing-link.spec.ts
• modules/web-visual/playful-games-tests/the-missing-link-round2.spec.ts
• conf/tasks/games-quality-20261013/results/verification-batch-011.md

Type one missing word that fits two meanings, using an eight-word answer bank, accepted aliases and optional progressive clues. Visible, unlisted and repeated guesses cost nothing.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben und Rückmeldungen. Die englische Ausgabe bleibt separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle vier Aufgaben sind heute als Übung erreichbar. Tagesaufgaben und Übungen nutzen denselben Vorrat. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards (8 repeating daily + 4 practice)

Solo

Limited beta ready

Typed input remains restricted to an eight-word bank and finite accepted spelling or meaning aliases. Twelve authored English boards provide eight repeating UTC daily boards and four separate practice boards. Each distinct wrong concept costs 15 points, each of two progressive clues costs 12 points, and four distinct misses exhaust the attempt; free refusals and repeated guesses do not expand the supported vocabulary. Progress is browser-local without account or cross-device synchronization. Changed-save protection uses a non-atomic preflight rather than a simultaneous-write transaction. These checks cover the current authored route; retained legacy service tests do not establish live-provider behavior. Human early-clue discovery, rejected-synonym expectations, difficulty, enjoyment and public production readiness remain unverified.

The 13 October independent review accepted removal of six unsupported aliases and neutral accepted-form wording. Six changed editions preserve old saved bytes on arrival; deliberate new play uses the revised content. Earlier scored insect saves remain intact until new play. Valid alternate answers and normalised input passed. Accepted for invited human playtesting; human feedback remains pending. This is local technical evidence, not deployment or public-release signoff.

Authored dual-meaning word clues, eight-word banks and accepted aliases.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

Eine Verbesserung umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-20. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-sixth-ten-rounds-20261020/evidence/words/ledger.json

2 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json

Mit deutschsprachigen Ersttestern Verständnis, strittige Antworten, genutzte Hilfen und eine freiwillige weitere Runde beobachten. Mit anderen Varianten derselben Richtung vergleichen; menschliche Rückmeldung steht aus.

Collect unprimed human feedback on comprehension, fairness, difficulty and replay interest; technical readiness and human feedback remain separate in conf/tasks/games-quality-20261013/game-quality.json.

Reassessed 11 October 2026 after two completed local play, review, improvement and replay cycles. Complete 73, 46 and 100 point solves, four-miss exhaustion, free refusals and repeats, raw alias recovery, canonical and alternative meaning readings, receipts, Copy and recovery were checked. Reproduced focus loss, stale guidance, rejected-action effects and an undefined-receipt exception were repaired compatibly without changing the pure rules or score economy. Fresh verification passed 128 client checks, seventeen Java checks, packaging and twelve distinct final desktop/phone browser cases; independent review found no blocker. Retains Limited beta ready, 4. The historical finite-alias and early-clue discovery gap remains open. Local reliability does not establish human difficulty, enjoyment, live-provider behavior or public production readiness.

Maturity gap: Observe rejected synonyms and early-clue discovery; accepted spelling/meaning aliases remain finite. Eight repeating UTC daily boards and four separate practice boards, authored English content and device-local progress.

The current default route uses the authored missing-link mode in fourth-deduction.js, with actual runtime identity game61; the inventory identity remains game-61. The retained older missing-link client and service are not the default player implementation. The first complete local cycle played missing-09, Bird to Computer. Endpoint and unlisted guesses were free. Wing cost 15 points, a blank guess cost nothing, and one progressive clue cost 12 points. The exact raw draft ' TWEETS ' survived reload and solved the puzzle for 73 points. Both the bird-call and short-posted-message readings, the Wing, hint and alias receipts, canonical Copy and terminal recovery were reviewed. Ordinary peer Replay from focused Copy left focus on the document body and retained obsolete fixed-result guidance. Another complete shared attempt reproduced repeated-guess guidance remaining after a peer bought a clue, then finished at 73 with both readings, all three receipts, Copy and recovery. The original focused client run passed 118 of 128 checks and failed ten targeted checks, with no skips. The original eight desktop/phone browser cases produced four passes and four targeted failures in 23.6 seconds. Both delayed Submit journeys raised the real undefined-last-receipt TypeError after adopting a newer attempt; meaningful full outcomes preceded the final boundary verdicts. No harness error was established. The scoped repair adds current-save preflight for free and refused actions, distinguishes rejected from invalid submissions, gates paid feedback and Replay effects on actual acceptance, retains native missing-word bank and clue nodes, clears obsolete guidance, rescues unavailable focus, and guards deferred result effects. Pure rules, strict snapshot fields, authored content, finite aliases, persistent identity and score economy remain unchanged. The shared runtime also serves other editions, but their compatibility checks are not evidence of gameplay in those editions. Independent review found no material blocker and the design check returned no findings. Fresh verification passed 128 client checks, comprising 112 pure-rule checks and sixteen DOM checks, with no failures or skips. Seventeen fresh Java checks passed: two current authored-deck checks, three retained legacy service checks, five shared authored-deck checks and seven inventory checks. Retained service checks are not current-route provider evidence. Packaging agreed with source and served runtime hashes and with the unchanged engine and then-current 23-row refreshed inventory. All eight repaired first-cycle desktop/phone browser cases passed in 19.8 seconds. Actual repaired play completed the raw-alias 73 point outcome again, with both readings, all receipts, Copy and active/terminal recovery. Peer Replay now focused the visible input, a native Wing bank button remained connected and focused through a peer draft change, and a peer clue cleared obsolete repeated-guess guidance before another complete 73 point result. The second complete cycle played missing-10, Bread to Ideas. Endpoint and unknown guesses were free; Yeast and Dough each cost 15 points, a repeated Yeast cost nothing, and two clues each cost 12 points. The final clue left usable input focus. The exact raw draft ' GROWTH ' survived reload and solved for 46 points. Both canonical Rise readings and the two independent Growth readings, all five chronological receipts, Copy and terminal recovery were reviewed. Review established no further scoped defect or source change. Fresh phone play made three distinct misses, Yeast, Thought and Dough, then recovered with one chance and 55 points available. Flour produced the fourth distinct miss and an honest exhausted outcome with no solved result. All four receipts, canonical and alternative meanings, Copy and terminal recovery were checked. A fresh accepted Replay with raw ' rising ' then solved for 100 points, with its single receipt, all readings, Copy and recovery reviewed. All four second-cycle desktop/phone browser cases passed in 7.4 seconds. The eight first-cycle and four second-cycle cases provide twelve distinct final passing browser cases; the final 128 client checks also all passed. Runtime SHA-256 remained 5e41a9a8c414475789f70267799b7c3b4099d4779058f7f2284625f519b144d7 through the second cycle, and the rules remained byte-identical. This is source-informed local agent and automated evidence, with no human, live-provider or deployment validation claimed. Earlier known-issue record, retained for history: Resolved and locally verified 11 October 2026: ordinary peer Replay rescues focus from hidden Copy to the visible answer input, peer changes clear obsolete result or repeated-guess guidance, and still-visible bank choices remain connected and focused. Main-save preflight protects blank, repeated, endpoint and unlisted guesses as well as paid actions. Accepted-state gates prevent false clue charges and submission feedback after adopting a newer attempt, including the reproduced undefined-last-receipt exception. Rejected Replay preserves the newer attempt and URL and focuses its actual active or terminal state; deferred result effects are fenced to the current state. No additional defect or runtime change was justified by the second completed cycle.

• modules/games/src/main/java/io/kgard/web/slice/games/game61/MissingLinkDailyController.java
• modules/games/src/main/resources/games/missing-beta-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game61/MissingLinkDailyAuthoredDeckTest.java
• modules/games/docs/games-maturity-20261010-fourth-prototype-beta.json
• conf/tasks/evidence/fourth-ten-games-20261010/deduction
• conf/tasks/evidence/fourth-ten-games-20261010/final-verification.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/76-missing-link-daily/verification.json
• conf/tasks/games-two-round-review-20261011/evidence/76-missing-link-daily/round-1
• conf/tasks/games-two-round-review-20261011/evidence/76-missing-link-daily/round-2
• modules/games/src/main/resources/static/js/games/fourth-deduction.js
• modules/games/src/main/resources/static/js/games/fourth-deduction-rules.js
• modules/games/src/test/js/fourth-deduction.test.cjs
• modules/games/src/test/js/fourth-deduction-ui.test.cjs
• modules/web-visual/playful-games-tests/missing-link-daily.spec.ts
• modules/web-visual/playful-games-tests/missing-link-daily-round2.spec.ts
• conf/tasks/games-quality-20261013/results/verification-missing-alias.md

Pass two-word clues from friend to friend. Lock a guess and the next clues before discovering the previous word, then compare the whole chain when it closes.

Bedienung, Wörter und Hinweisoptionen sind auf Englisch; freie eigene Wörter sind nicht vorgesehen.

English interface, words and clue choices.

Neue Ketten können dieselben Startwörter verwenden. Eine klar bezeichnete Soloübung umfasst eine Übergabe.

A friend chain closes after three handoffs or 24 hours. A separately labeled solo practice has one handoff.

Asynchronous friend chain, with one turn per participant browser identity; solo practice is labeled separately.

Playable prototype

Full chains require other participants. Browser identity is not a verified person. Full history stays hidden until the chain closes.

Six fresh local browser play/review cycles completed on 25 October 2026. Human playtesting and dated deployment verification are separate evidence.

Twenty curated starter words with four example clues each.

Participants choose an interpretation and two outgoing clues from the authored set.

Spring application; PostgreSQL stores chains and committed turns. A browser cookie identifies participants.

Rückmeldungen unabhängiger Spieler zur Verständlichkeit und zum gemeinsamen Aufdecken des Verlaufs einholen.

Gather independent human feedback on clue clarity, meaning changes and sharing.

Experimental Spring/PostgreSQL implementation. This entry records source facts and does not inherit earlier games' review results.

Maturity gap: Human enjoyment, retention, accessibility across real devices remain unverified.

Added after the original 99-edition review; historical findings and recommendation cohorts are unchanged. The 25 October local improvement pass retained two varied clean final rounds after response-recovery repairs. Deployment status belongs to dated release records.

• modules/games/src/main/java/io/kgard/web/slice/games/social/LostInWordsRules.java
• modules/games/src/main/java/io/kgard/web/slice/games/social/LostInWordsStore.java
• modules/games/src/main/resources/static/games/social/lost-in-words.js
• conf/tasks/social-word-games-20261024/continuation.md
• conf/tasks/word-games-improvement-20261025/social/completion.json
• conf/tasks/word-games-improvement-20261025/README.md
• modules/games/docs/releases.md

Choose one word that connects both ideas from six ranked choices. Optional hints, daily and practice puzzles, and explanations follow your locked answer.

Oberfläche, Regeln und vorbereitete Aufgaben benötigen derzeit Englisch. Die deutsche Beschreibung übersetzt das Spiel noch nicht.

English

Mehrere bereits veröffentlichte Rätsel sind als Übung zugänglich. Der freigegebene Übungsumfang hängt vom Datum ab; die heutige Aufgabe wird ausgelassen.

35 authored boards (35-board daily release; earlier 7-board schedule preserved)

Solo + friend challenge

Limited beta ready

Finite 35-board repeat cycle; quality ranking is editorial and fairness/retention are not human-validated. Browser-local progress, no cross-device account sync; observations are untrusted client events rather than unique-player statistics.

Resolved and locally verified 11 October 2026: explicit Replay removes the one-time reveal URL so a draft and paid hints survive reload without being revealed again; All games opens the full catalog. No additional defect was established in the second completed review cycle. Independent collection review, 15 October 2026: Unknown saved results survive read outages; returning durable daily progress replaces a memory-only result. Fresh60 client checks,24 browser cases,60 native transitions and actual saved-state/peer/Copy checks across both editions passed. Accepted for invited human playtesting; human feedback remains pending.

Authored editorial deck: six candidate bridge words, two-sided path/evidence narratives, hints and assigned quality ranks.

No live Termpath, Contextoid or LLM scoring call in player rounds.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

Den tatsächlichen Weg vom geteilten Ergebnis zum spielenden Empfänger prüfen; keine unbestätigten Wachstumswerte ableiten.

Run unprimed fairness and voluntary-next-puzzle sessions, then establish a reviewed publishing schedule.

Reassessed 11 October 2026 after two completed source-informed local play/review/improvement/replay cycles and final passing 30 client checks plus 4 desktop/phone browser cases. Retains Limited beta ready: actual ranked choices, hint costs and repaired replay recovery were exercised; finite editorial content, human fairness and return play still limit readiness. No public-release or production-readiness promotion.

Maturity gap: The answer remains editorial best-fit judgment;35 repeating boards and unmeasured human return play limit a release-readiness claim.

Two source-informed local cycles exercised tentative choices, locked editorial ranks, both paid hints, explanations, copied results, phone layout and reload. Actual results were 100 then 95 points on Lego/Veterinary, and 20 then 90 points on the phone River/Money practice; the first-attempt practice ledger stayed intact. The reveal-URL replay/reload defect was reproduced before repair and the same boundary passed afterward. Final local checks: 30 client tests and 4 browser cases across desktop and phone. Editorial fairness, human return play and production readiness remain unverified.

• modules/games/src/main/java/io/kgard/web/slice/games/game85/v3/BridgeHunterPublicBetaController.java
• modules/games/src/main/resources/games/bridge-hunter-public-beta-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game85/v3/BridgeHunterPublicBetaSecurityWebMvcTest.java
• modules/games/docs/games-maturity-20261005.json
• modules/games/src/main/java/io/kgard/web/slice/games/game85/v3
• conf/tasks/evidence/games-maturity-20261005/root-browser-baseline.json
• conf/tasks/evidence/games-maturity-20261005/root-play-observations.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/98-public/round-1
• modules/games/src/main/resources/static/js/games/bridge-hunter.js
• modules/games/src/test/js/bridge-hunter.test.cjs
• modules/games/src/test/js/bridge-hunter-runtime.test.cjs
• modules/web-visual/playful-games-tests/bridge-hunter-public.spec.ts
• conf/tasks/games-quality-20261013/results/batch-066-public-trio-repair.md
• conf/tasks/games-quality-20261013/results/verification-batch-066-public-trio.md

Select a two-sided bridge and the precise reading connecting it to each endpoint. All supported bridges receive equal credit.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben und Rückmeldungen. Die englische Ausgabe bleibt separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle vier Aufgaben sind heute als Übung erreichbar. Tagesaufgaben und Übungen nutzen denselben Vorrat. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards (8 repeating daily + 4 practice)

Solo

Limited beta ready

Twelve finite English authored boards comprise eight repeating UTC daily and four practice boards. The player selects a listed bridge and both directed relationship readings; every supported bridge earns equal credit. Unlisted or creatively reinterpreted bridges are outside the supplied scoring rules and need editorial review. A solve earns 100 minus 15 per earlier distinct failed complete inspection and 10 for the optional scene Hint. Three different failed reports exhaust the attempt honestly at 0; incomplete and repeated reports are free, and Reveal is unearned 0. Terminal audit rows include the four distinct supported meanings across two bridge alternatives; repeated endpoint readings are not additional unique meanings. Existing history compaction preserves paid inspections and Hint chronology while allowing free edits beyond the request-history boundary; literal valid 2,000-event recovery remains strict. Progress is browser-local without account or cross-device synchronization. Readable-save protection preserves unknown durable bytes during a read outage, but localStorage preflight cannot serialize simultaneous writers atomically. Failed saving permits explicitly warned memory continuation. The mapped route serves authored content; the retained unannotated custom method and provider-backed service were not live-verified. Human judgments about restrictive relation labels, creative bridges, difficulty and unprimed onboarding remain open. Linux, live-provider, account synchronization, production deployment and public-release readiness were not established by these local checks.

The 13 October independent review accepted full gameplay and the integrated hint repair: a previously explained first connection is free instead of charging again. Historical paid saves retain their original scores. Authored alternatives, keyboard/phone play, replay and recovery passed. Accepted for invited human playtesting; human feedback remains pending. This is local technical evidence, not deployment or public-release signoff.

Authored endpoint pairs, bridge words and typed relationship readings for both legs.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

2 Verbesserungen umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-20. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-fifth-ten-rounds-20261020/evidence/bridge/ledger.json

2 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json

Mit deutschsprachigen Ersttestern Verständnis, strittige Antworten, genutzte Hilfen und eine freiwillige weitere Runde beobachten. Mit den anderen Varianten derselben Richtung vergleichen; menschliche Rückmeldung steht aus.

Collect unprimed human feedback on comprehension, fairness, difficulty and replay interest; technical readiness and human feedback remain separate in conf/tasks/games-quality-20261013/game-quality.json.

Retains Limited beta ready, 4, and the historical question of whether explicit relation labels teach connections naturally or restrict creative bridges. Both completed local cycles reviewed original/repaired Ship and Ferry 100, Garden 75, genuine three-report 0 and fresh Herb 100. Each outcome included all six terminal audit rows, four distinct supported meanings, both bridge alternatives, exact paid receipts, active/terminal recovery and canonical Copy. The original 89-client/eight-browser baselines already passed; only a separately reproduced asymmetric read outage required the owning readable-save opt-in. Final evidence comprises ninety client checks, eleven fresh Java checks, twelve distinct owning desktop/phone browser journeys and eight earlier-game compatibility journeys, all with zero skips. Second-round review justified no further source change. Finite content, browser-local non-atomic saves and human semantic/onboarding gaps remain; no Linux, provider, account synchronization or deployment readiness is inferred.

Maturity gap: Does choosing explicit relation labels teach the connection naturally, or feel restrictive for users proposing creative bridges? Eight repeating UTC daily boards and four separate practice boards, authored English content and browser-local progress.

The mapped player is Clue Trestle at /games/game22, using bridge-readings.jte, bridge-readings-deck.json and the shared fifth-link.js/fifth-link-rules.js family. Inventory game-22 and persisted runtime game22 remain distinct. The authored-v1 envelope keeps rules identity fifth-link-2. Recovery is strict over version, boardId, events, phase and replay, replays legal events, limits stored histories to 2,000 and normalizes Replay to practice mode. Twelve authored boards comprise eight repeating daily and four practice boards. The separate language/source/target/bridge controller method is unannotated and is not a live-data mode on this route. No side store is used. Original actual desktop Ocean to Travel read the complete help, six candidate labels and both four-relationship menus. Empty Inspect was free. A full Ship draft using is-crossed-by and enables-a-journey survived reload before earning 100. Root reviewed all six audit rows, four distinct supported meanings, both valid bridges and the literal paid inspection receipt, then canonical Copy and terminal recovery. A separate explicit Replay used Ferry with both correct readings for equal 100 and repeated the full six-row audit, receipt, Copy and reload in a bounded 1280 by 720 view. Original runtime SHA-256 was a3acec834acf5a1bdef741299dade211e0bc718cac8d47ad01f20840fe186776. All 89 original client checks passed with zero skips in 821.6475 milliseconds, including thirteen owning DOM cases and earlier-family/pure compatibility. All eight original owning desktop/phone browser journeys passed with zero skips in 32.5 seconds. Inherited gameplay, native choices, Hint, startup alignment, history-cap completion, Replay and Copy guards already worked; these are compatibility evidence rather than newly repaired defects. The known asymmetric storage risk was then checked independently against the immutable original runtime. That appended cold-read-only case failed with zero skips in 355.522208 milliseconds: reads failed while writes worked, so a free Ship selection overwrote unknown peer 100 bytes and removed the warning. Full memory, recovered and fresh 100 outcomes were completed before the byte/warning assertion. Ordinary blocked-read-and-write memory coverage had not proved this boundary. The sole owning runtime change adds requireReadableSave only when data-game-id is game22. It uses the existing shared policy: accept explicitly warned memory actions while the current save is unreadable, but skip durable writes. The helper, default policy and Sense Junction/Pathbridge callers are unchanged. This is the sixth reproduced opt-in through this ordered review, after Hub Sheriff, Path Sorter, Mirror Paths, Trailsplice and Rank Ladder; other callers receive no new protection credit. The original thirteen DOM bodies are unchanged, with one appended actual-outcome outage regression. Nine protected pure/deck/template/CSS/earlier-test/browser inputs remain byte-identical. All 90 final client cases passed with zero skips in 784.236208 milliseconds: seventy owning-plus-pure cases and twenty earlier compatibility cases. Eleven fresh Java checks passed between 10:09:09.599 and 10:09:10.115 UTC; packaging passed in thirteen seconds and independent source review passed. Final runtime SHA-256 is 8f797e2de24769768b97e895bfa20d7a03a28bafcb63d1bc9a151302d0629a8d; owning DOM SHA-256 is 27445ac608776c60cce69525973a0de79aeba89610c7d8f4e7b6f5b4829033be. Unchanged template SHA-256 is 4153ec1add9bc972b340c5bcc7bfa650ad6b02f3f074a05e665dbda954fa27e8 and CSS SHA-256 is 861d74d1669afb6405022e65a23f794a075e6eb9d6d2c36db77f4e51e71b80d9. Immutable package and served source matched the source and 51-review canonical inventory. Actual repaired desktop Ship and Ferry each earned 100 after explicit fresh play, with full draft recovery, six audit rows, four meanings, both alternatives, exact paid receipt, canonical Copy and terminal reload. Sixteen final first-cycle browser journeys passed with zero skips in 51.6 seconds: eight owning Clue Trestle, four earlier Sense Junction and four earlier Pathbridge compatibility cases. The second actual baseline used Garden at a verified 390 by 844 phone viewport. Herb with the false same-spelling relation failed; an exact repeat was free. Root bought the Hint, cleared freely and recovered Vegetable with growing-place/adds-taste before Inspect earned 75. All six audit rows, four distinct meanings, both alternatives and both charged receipts were read, followed by canonical Copy, terminal recovery and a bounded phone result view. The explanation identified the supported relation, the arithmetic 100 minus 15 minus 10 was correct, and ordinary saves/recovery were warning-free. Result controls were already distinct, large and inside the viewport. No further source change was justified. Final actual phone Replay submitted three different complete wrong reports: unsupported Hammer with growing-place/adds-taste, Herb with same-spelling/adds-taste, and Herb with growing-place/opposite. Active reload retained the first two failures; the third ended honestly at 0. All six audit rows, four supported meanings, two bridge alternatives and all three literal charged receipts were reviewed, with canonical Copy, terminal reload and a bounded phone view. A clean Replay recovered Herb plus growing-place with the outgoing reading still empty, selected adds-taste and earned 100. The full six-row audit, both alternatives, four meanings and sole receipt were reviewed again with Copy and recovery. Page and viewport were both 390 pixels wide, with separated large result controls and no horizontal overflow. All four final second-cycle browser journeys passed with zero skips in 12.1 seconds. The owning total is twelve distinct browser journeys, with eight earlier-game compatibility journeys reported separately. Runtime, rules, content, template, styling and fixtures remained unchanged in the second cycle. Both actual local cycles are complete; no additional defect was established. These local engineering and bounded visual checks do not establish human play quality, live-provider behavior, Linux deployment or account synchronization. Earlier known-issue record, retained for history: Locally reviewed through two completed cycles on 11 October 2026. The original 89 client checks and eight owning browser journeys already passed, including inherited gameplay, capacity, startup, native focus, Replay and Copy protections. A separately reproduced cold-read-only outage could replace an unknown peer 100 with a free Ship selection when reads failed but writes succeeded, and hide the warning. The only owning runtime change enables the existing readable-save policy for persisted game22: warned memory actions may continue while the current save is unreadable, but the durable write is skipped. Other fifth-link callers retain their default policy and receive no new protection claim. The original thirteen owning DOM bodies, rules, deck, template and browser fixtures remain unchanged; one outcome-based outage regression was appended. Second-round Garden 75, genuine three-report 0 and fresh Herb 100 established no additional defect or justified source change. No original startup, capacity, scoring, focus or Copy defect is claimed for this edition. Local save preflight remains non-atomic.

• modules/games/src/main/java/io/kgard/web/slice/games/game22/BridgeHunterController.java
• modules/games/src/main/resources/games/bridge-readings-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game22/BridgeHunterAuthoredDeckTest.java
• modules/games/docs/games-maturity-20261010-fifth-prototype-beta.json
• conf/tasks/evidence/fifth-ten-games-20261010/link
• conf/tasks/evidence/fifth-ten-games-20261010/final-verification.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/48-bridge-hunter/verification.json
• conf/tasks/games-two-round-review-20261011/evidence/48-bridge-hunter/runtime.diff
• modules/games/src/main/java/io/kgard/web/slice/games/game22/bridge-readings.jte
• modules/games/src/main/resources/static/js/games/fifth-link.js
• modules/games/src/main/resources/static/js/games/fifth-link-rules.js
• modules/games/src/main/resources/static/js/games/playful-session.js
• modules/games/src/test/js/bridge-readings-ui.test.cjs
• modules/games/src/test/js/fifth-link.test.cjs
• modules/web-visual/playful-games-tests/bridge-hunter.spec.ts
• modules/web-visual/playful-games-tests/bridge-hunter-round2.spec.ts
• conf/tasks/games-quality-20261013/results/verification-batch-005.md

Compare two-leg word bridges within a printed effort budget. Prefer the strongest weaker leg, then lower effort; test and bank a route.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben und Rückmeldungen. Die englische Ausgabe bleibt separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle vier Aufgaben sind heute als Übung erreichbar. Tagesaufgaben und Übungen nutzen denselben Vorrat. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards (8 repeating daily + 4 practice)

Solo

Limited beta ready

Bridge strengths and effort are assigned rubric values, not live corpus measures. The finite English bank contains eight repeating daily boards and four practice boards. Progress remains browser-local with no account or cross-device synchronization; cross-tab checks do not make local writes an atomic transaction. Novice understanding of weaker-leg priority and paid extra tests, and enjoyment of comparing four routes, still need human observation.

Resolved and locally verified 11 October 2026: long sequences of free selection changes no longer strand an unfinished round at the 256-event limit. Legacy saves recover unchanged, and the next valid action removes redundant selections while retaining every tested candidate and paid report. Adopting another tab’s active or completed attempt now clears obsolete local playing instructions without rewriting that save. The repaired local replays established no further defect. Independently accepted for invited human playtesting on 13 October 2026: Results and replay reveal the useful context; a report after a tested maximum is refused without charge, while historical charged results remain exact. Fresh client, Java, browser, genuine historical-save and native checks pass. Human feedback remains pending.

Authored two-leg bridge candidates with declared leg strengths, effort and budget.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

Eine Verbesserung umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-22. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-eighth-ten-rounds-20261022/evidence/learning/ledger.json

4 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json

Mit deutschsprachigen Ersttestern Verständnis, strittige Antworten, genutzte Hilfen und eine freiwillige weitere Runde beobachten. Mit anderen Varianten derselben Richtung vergleichen; menschliche Rückmeldung steht aus.

Collect human clarity, difficulty and replay feedback.

Reassessed 11 October 2026 after two completed local play, review, improvement and replay cycles. Actual 100-, 95-, 80- and 90-point results exercised manual banking, three-test automatic banking, the report fee, over-budget routes and reload. Cycle one passed 20 focused client checks and four browser cases; cycle two passed 23 focused client checks and six browser cases, with focused Java and packaging verification recorded. Retains Limited beta ready: the finite authored rubric still needs unprimed human evidence for clarity, fairness, difficulty and engagement. Independently accepted for invited human playtesting on 13 October 2026: Results and replay reveal the useful context; a report after a tested maximum is refused without charge, while historical charged results remain exact. Fresh client, Java, browser, genuine historical-save and native checks pass. Human feedback remains pending.

Maturity gap: Observe whether novices understand the weaker-leg priority and paid extra tests, and whether comparing four routes feels engaging. Eight repeating UTC daily boards and four separate practice boards, authored English content and browser-local progress.

The first source-informed local cycle on bridge-daily-09 tested Network, reloaded its evidence and banked 100 points. Desktop and phone checks then reproduced enabled Test failing after 256 alternating free selection changes and reload. After repair, a desktop replay tested Traffic and Network, banked Network for 95 after the extra-test fee, and restored the same receipt on reload. The second cycle used the weaker-leg report and tested Network, Traffic and Hub; the third test automatically banked Network for 80, but a peer tab retained an obsolete instruction to choose an untested candidate. After the feedback repair, phone tested the over-budget Route for zero base, then Network and Traffic; automatic banking earned 90 without a report. Desktop adopted all three tests and the correct result and replaced the stale instruction with current evidence. Phone Copy and terminal reload retained the 90-point receipt. Recorded verification passed 20 focused client checks and four desktop/phone browser cases in cycle one, then 23 focused client checks and six browser cases in cycle two, plus focused Java checks and packaging. Blocked-storage checks retained warned in-memory testing and banking. These are local source-informed agent results, not independent human fairness, difficulty, enjoyment, semantic measurement or production-readiness evidence.

• modules/games/src/main/java/io/kgard/web/slice/games/game85/BridgeHunterDailyController.java
• modules/games/src/main/resources/games/bridge-daily-fifth.json
• modules/games/src/test/java/io/kgard/web/slice/games/game85/BridgeHunterDailyAuthoredDeckTest.java
• modules/games/docs/games-maturity-20261010-fifth-prototype-beta.json
• conf/tasks/evidence/fifth-ten-games-20261010/root
• conf/tasks/evidence/fifth-ten-games-20261010/final-verification.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/94-daily/round-1
• conf/tasks/games-two-round-review-20261011/evidence/94-daily/round-2
• modules/games/src/main/resources/static/js/games/bridge-daily-fifth.js
• modules/games/src/test/js/bridge-daily-fifth.test.cjs
• modules/games/src/test/js/bridge-daily-fifth-ui.test.cjs
• modules/web-visual/playful-games-tests/bridge-hunter-daily.spec.ts
• conf/tasks/games-quality-20261013/results/batch-055-bridge-hunter-path-poker-repair.md
• conf/tasks/games-quality-20261013/results/verification-batch-055-bridge-poker.md

Compare neutral clues, lock one bridge, and inspect its assigned rubric and two-sided evidence.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben und Rückmeldungen. Die englische Ausgabe bleibt separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle vier Aufgaben sind heute als Übung erreichbar. Tagesaufgaben und Übungen nutzen denselben Vorrat. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards (8 repeating daily + 4 practice)

Solo

Limited beta ready

Scores are editorial assignments; strong clues can turn selection into rubric recognition rather than discovery. The optional notebook keeps at most three candidate readings, with up to 160 characters per endpoint; the text is ungraded. Progress and writing remain browser-local. Cross-tab checks do not make local writes an atomic transaction.

Resolved and locally verified 11 October 2026: a rejected Replay preserves the peer notebook instead of clearing it; ordinary notebook updates are adopted, and stale reading actions are rejected before replacing newer saved text. Changed practice completion or Replay blocks pending notebook edits, damaged saves are preserved, temporary save failures retain warned in-memory text, and rejected Stage does not announce a selection. The repaired local replays established no further defect. Supplementary closure, 13 October 2026: The actual get-only outage previously overwrote an unreadable prior Toast reading during Keep and Replay. Seventeen failed reading reads now preserve exact prior bytes at all three checkpoints with warnings; read return restores the literal prior Toast interpretations. A separate genuine Breakfast peer arrival after delegated Replay clearing previously acquired an accepted fingerprint without adoption, so the next Toast edit erased it. Replay now adopts the actual arriving Breakfast immediately; both Toast edits retain it, and reload restores it. Whole Toast 90 and Breakfast 92 outcomes, all six authored meanings, manual Copy and fresh durable Breakfast 92 completed before both boundary verdicts. The bounded repair passed focused and final combined checks; simultaneous-write atomicity remains unverified. Independently accepted for invited human playtesting on 15 October 2026. Peer Replay restores useful focus and keeps the candidate heading and first choice visible, while valid notebook text/caret and outside focus remain stable. Delayed placement yields to newer input and raw peer changes. Final independent 63 client checks, six exact peer runs with 24 clear samples, ten intent/peer guards and two valid-focus peer controls pass; earlier full-game, content and storage evidence remains applicable. Human feedback remains pending.

Authored bridge candidates, neutral clues, two-sided evidence and assigned multidimensional scoring rubric.

language selects a legacy fixed LEGO→veterinary candidate lab; no live API call.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync. Kept-reading writes and Replay clearing require a readable side save. The intended empty post-clear fingerprint is reconciled against any actual arriving reading before later edits.

Eine Verbesserung umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-22. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-ninth-ten-rounds-20261022/evidence/words/ledger.json

4 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json

Mit deutschsprachigen Ersttestern Verständnis, strittige Antworten, genutzte Hilfen und eine freiwillige weitere Runde beobachten. Mit anderen Varianten derselben Richtung vergleichen; menschliche Rückmeldung steht aus.

Observe unprimed players comparing clues and interpreting the editorial rubric; use that evidence to tune contrasts and extend reviewed content.

Reassessed 11 October 2026 after two completed local play, review, improvement and replay cycles. Actual 92-, 90- and 83-point results, pinning, notebook recovery and desktop/phone collaboration were exercised; final verification passed 50 focused client checks, eight browser cases and focused Java checks. Retains Limited beta ready: the finite six-choice files and assigned rubric still need unprimed human evidence for fairness, learning, difficulty and engagement.

Maturity gap: Twelve small six-choice files, one irreversible decision and subjective score gaps provide limited depth and discovery. Human learning and engagement remain unverified.

The first source-informed local cycle on bh2-09 pinned Toast, kept two Breakfast endpoint readings, reloaded the selection, pin and text, then locked Breakfast for 92 points. Toast remained a 90-point unbanked alternative. After the rejected-Replay repair, a Toast replay earned 90 and retained its original kept text plus a separate unkept revision through reload. The second cycle reproduced ordinary notebook loss across actual tabs without delayed notifications; a subsequent Coffee lock earned 83 correctly despite the lost writing. After repair, phone and desktop visibly adopted the same kept Breakfast reading and current draft; edits to each endpoint survived reload together with the original kept text. Lock Breakfast, Copy and terminal reload retained 92 points and the ungraded read-only writing. Final recorded verification passed 50 focused client checks, eight desktop/phone browser cases and focused Java checks. These are local source-informed agent results, not independent human fairness, difficulty, enjoyment or production-readiness evidence. Supplementary closure findings, 13 October 2026. The actual get-only outage previously overwrote an unreadable prior Toast reading during Keep and Replay. Seventeen failed reading reads now preserve exact prior bytes at all three checkpoints with warnings; read return restores the literal prior Toast interpretations. A separate genuine Breakfast peer arrival after delegated Replay clearing previously acquired an accepted fingerprint without adoption, so the next Toast edit erased it. Replay now adopts the actual arriving Breakfast immediately; both Toast edits retain it, and reload restores it. Whole Toast 90 and Breakfast 92 outcomes, all six authored meanings, manual Copy and fresh durable Breakfast 92 completed before both boundary verdicts. Focused verification passed 90 existing checks across three suites and separately four unchanged closure bodies. The final combined local run passed the eight original suites, 251/251 checks with zero skips in 4141.056708 ms, and separately the same nine closure bodies, 9/9 with zero skips in 2017.653167 ms. These supplementary regression checks follow the existing two cycles and earn no extra cycle or maturity credit. Local save protection is optimistic and non-atomic. These DOM fault journeys do not establish native fault behavior, physical-device access, human enjoyment or production deployment. Independently accepted for invited human playtesting on 15 October 2026. Peer Replay restores useful focus and keeps the candidate heading and first choice visible, while valid notebook text/caret and outside focus remain stable. Delayed placement yields to newer input and raw peer changes. Final independent 63 client checks, six exact peer runs with 24 clear samples, ten intent/peer guards and two valid-focus peer controls pass; earlier full-game, content and storage evidence remains applicable. Human feedback remains pending.

• modules/games/src/main/java/io/kgard/web/slice/games/game85/v2/BridgeHunterDailyV2Controller.java
• modules/games/src/main/resources/games/bridge-hunter-v2-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game85/v2/BridgeHunterV2AuthoredDeckTest.java
• modules/games/src/test/java/io/kgard/web/slice/games/game85/v2/BridgeHunterV2PlayRouteTest.java
• modules/games/docs/games-maturity-followup-20261009.json
• conf/tasks/games-beta-improvements-20261009/queue.json
• conf/tasks/evidence/games-beta-improvements-20261009/game-85-v2/cycle1-independent-review.json
• conf/tasks/evidence/games-beta-improvements-20261009/game-85-v2/cycle2-independent-review.json
• conf/tasks/evidence/games-beta-improvements-20261009/game-85-v2/final-browser.json
• conf/tasks/evidence/games-beta-improvements-20261009/game-85-v2/final-phone-result.jpg
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/95-v2/round-1
• conf/tasks/games-two-round-review-20261011/evidence/95-v2/round-2
• modules/games/src/main/resources/static/js/games/bridge-hunter-v2.js
• modules/games/src/test/js/bridge-hunter-v2.test.cjs
• modules/games/src/test/js/bridge-hunter-v2-ui.test.cjs
• modules/web-visual/playful-games-tests/bridge-hunter-v2.spec.ts
• conf/tasks/games-two-round-review-20261011/evidence/common-closure-prepared/before-nine-extras-correct-jsdom.log
• conf/tasks/games-two-round-review-20261011/evidence/common-closure-repairs/high96-95-90/final-focused-review-v2.json
• conf/tasks/games-two-round-review-20261011/evidence/common-closure-repairs/high96-95-90/focused-owning-after-v1.log
• conf/tasks/games-two-round-review-20261011/evidence/common-closure-repairs/high96-95-90/four-extras-after-v1.log
• conf/tasks/games-two-round-review-20261011/evidence/common-closure-repairs/high96-95-90/source-proposal-and-inverse-v1.json
• conf/tasks/games-two-round-review-20261011/evidence/common-closure-repairs/high96-95-90/after-wrapper-manifest-v1.json
• conf/tasks/games-two-round-review-20261011/evidence/common-closure-prepared/root-after-seven-combined-review-v1.json
• conf/tasks/games-quality-20261013/results/verification-batch-069-expedition-bridge.md
• conf/tasks/games-quality-20261013/results/batch-069-final-three-private-repairs.md

Use a ten-word bank and two-sided feedback to find a top-quality bridge within six scored guesses.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben und Rückmeldungen. Die englische Ausgabe bleibt separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle vier Aufgaben sind heute als Übung erreichbar. Tagesaufgaben und Übungen nutzen denselben Vorrat. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards (8 repeating daily + 4 practice)

Solo

Limited beta ready

Free typing is bounded to the ten-word bank; six scored guesses. Quality grades are assigned, not corpus measures.

Resolved and locally verified 11 October 2026: rejected stale Replay preserves the newer attempt and its separate comparison save; a rejected guess no longer reports a point charge after adopting a peer attempt. Accepted in-memory moves and Replay still work when storage is unavailable. Final replays established no further defect. Supplementary closure, 13 October 2026: An actual per-key read outage, with set/remove still usable, previously let terminal comparison edits and Replay replace an unreadable prior comparison without warning. Five failed comparison reads now preserve the exact prior Wood/Story c2/c3 bytes at all three checkpoints with visible memory-only guidance; read return restores that actual comparison. Complete paid 80/100 play and Replay, three guesses with a miss and paid hint, all explanations, manual Copy, recovery and a fresh durable 80 result completed before the preservation verdict. The bounded repair passed focused and final combined checks; simultaneous-write atomicity remains unverified. Independent collection review, 15 October 2026: main-read and failed-Replay-write protection preserves the exact prior earned result and comparison; a guarded restored-result arrival keeps the full receipt below the sticky desk. Fresh 40 client checks, 12 browser cases, 24 native transitions and independent input/peer controls passed. Accepted for invited human playtesting; human feedback remains pending. Finite editorial rankings and browser-local, non-atomic saves remain limitations.

Authored ten-word bridge banks, two-sided feedback, aliases and assigned scoring rubrics.

language selects the retained legacy v4 experiment reading the public Clue Trestle deck (35 boards), not a live provider.

Spring/JTE + browser client; packaged authored content and browser-local progress, with a separate comparison save. Changed compatible saves are adopted before outdated actions can overwrite them. Comparison cleanup requires an accepted practice Replay. Unavailable storage permits warned in-memory play; no cross-device account sync. Comparison writes and Replay clearing independently require a readable comparison save; unavailable reads retain physical prior bytes while current comparisons remain usable in warned page memory.

Eine Verbesserung umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-22. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-eighth-ten-rounds-20261022/evidence/words/ledger.json

5 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json

Mit deutschsprachigen Ersttestern Verständnis, strittige Antworten, genutzte Hilfen und eine freiwillige weitere Runde beobachten. Mit anderen Varianten derselben Richtung vergleichen; menschliche Rückmeldung steht aus.

Observe unprimed players comparing endpoint readings and interpreting the assigned rubric and hint costs; use that evidence to tune contrasts and extend reviewed content.

Reassessed 11 October 2026 after two completed source-informed local play, review, improvement and replay cycles. Earned results 84, 92, 0 and 82 exercised paid misses and hints, accepted aliases, free unsupported/repeated guesses, six-attempt loss, comparison arithmetic, copying and recovery on desktop and phone. Two reproduced rejection defects were repaired. Verification recorded 1,507 passing module-client checks, including 64 focused checks, six focused browser cases and passing focused v4 Java checks. Retains Limited beta ready: independent human fairness, difficulty, learning and replay interest remain unverified; no public-release or production-readiness promotion.

Maturity gap: Twelve finite files, subjective score differences and strongly guiding relation hints limit discovery and strategic depth. Human learning, difficulty and engagement remain unverified.

Two completed local play, review, improvement and replay cycles used practice bh4-09. Butter plus a relation hint, then Breakfast, earned 84; reload retained the actions, costs and Butter/Breakfast comparison. An unhinted Breakfasts alias replay earned 92 and survived terminal reload. On phone, unsupported Umbrella and normalized repeated BUTTER__ used no attempt or points; six distinct misses ended at zero, with the strongest Toast quality 90 explicitly unbanked and the best-file benchmark 92. After Replay, both paid hints and a spaced mixed-case Breakfasts alias earned 82; copying and reload retained the one-guess, two-hint receipt. Controlled delayed-storage tests reproduced comparison loss on stale Replay and false charged feedback for a rejected answer before repair, then verified both repairs. Local verification passed 1,507 module-client checks, including the 64 focused v4 rules/UI/shared-session checks, plus six focused desktop/phone browser cases and the focused v4 Java checks. These source-informed local checks do not measure independent human fairness, difficulty, learning or engagement, establish simultaneous-write atomicity, or confirm production readiness. Supplementary closure findings, 13 October 2026. An actual per-key read outage, with set/remove still usable, previously let terminal comparison edits and Replay replace an unreadable prior comparison without warning. Five failed comparison reads now preserve the exact prior Wood/Story c2/c3 bytes at all three checkpoints with visible memory-only guidance; read return restores that actual comparison. Complete paid 80/100 play and Replay, three guesses with a miss and paid hint, all explanations, manual Copy, recovery and a fresh durable 80 result completed before the preservation verdict. Focused verification passed 90 existing checks across three suites and separately four unchanged closure bodies. The final combined local run passed the eight original suites, 251/251 checks with zero skips in 4141.056708 ms, and separately the same nine closure bodies, 9/9 with zero skips in 2017.653167 ms. These supplementary regression checks follow the existing two cycles and earn no extra cycle or maturity credit. Local save protection is optimistic and non-atomic. These DOM fault journeys do not establish native fault behavior, physical-device access, human enjoyment or production deployment.

• modules/games/src/main/java/io/kgard/web/slice/games/game85/v4/BridgeHunterV4Controller.java
• modules/games/src/main/resources/games/bridge-hunter-v4-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game85/v4/BridgeHunterV4AuthoredDeckTest.java
• modules/games/src/test/java/io/kgard/web/slice/games/game85/v4/BridgeHunterV4PlayRouteTest.java
• modules/games/docs/games-maturity-followup-20261009.json
• conf/tasks/games-beta-improvements-20261009/queue.json
• conf/tasks/evidence/games-beta-improvements-20261009/game-85-v4/cycle1-independent-review.json
• conf/tasks/evidence/games-beta-improvements-20261009/game-85-v4/cycle2-independent-review.json
• conf/tasks/evidence/games-beta-improvements-20261009/game-85-v4/final-browser.json
• conf/tasks/evidence/games-beta-improvements-20261009/game-85-v4/final-phone-result.jpg
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/96-v4/round-1
• conf/tasks/games-two-round-review-20261011/v4-rejection-fix-note.md
• modules/games/src/main/resources/static/js/games/bridge-hunter-v4.js
• modules/games/src/main/resources/static/js/games/playful-session.js
• modules/games/src/test/js/bridge-hunter-v4.test.cjs
• modules/games/src/test/js/bridge-hunter-v4-ui.test.cjs
• modules/games/src/test/js/playful-session.test.cjs
• modules/web-visual/playful-games-tests/bridge-hunter-v4.spec.ts
• conf/tasks/games-two-round-review-20261011/evidence/common-closure-prepared/before-nine-extras-correct-jsdom.log
• conf/tasks/games-two-round-review-20261011/evidence/common-closure-repairs/high96-95-90/final-focused-review-v2.json
• conf/tasks/games-two-round-review-20261011/evidence/common-closure-repairs/high96-95-90/focused-owning-after-v1.log
• conf/tasks/games-two-round-review-20261011/evidence/common-closure-repairs/high96-95-90/four-extras-after-v1.log
• conf/tasks/games-two-round-review-20261011/evidence/common-closure-repairs/high96-95-90/source-proposal-and-inverse-v1.json
• conf/tasks/games-two-round-review-20261011/evidence/common-closure-repairs/high96-95-90/after-wrapper-manifest-v1.json
• conf/tasks/games-two-round-review-20261011/evidence/common-closure-prepared/root-after-seven-combined-review-v1.json
• conf/tasks/games-quality-20261013/results/batch-067-bridge-v4-repair.md
• conf/tasks/games-quality-20261013/results/verification-batch-067-bridge-v4.md

Discover three distinct bridges by comparing supplied evidence on both sides of two endpoint roles. Find every candidate that satisfies both roles; one-sided associations do not count.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben und Rückmeldungen. Die englische Ausgabe bleibt separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle vier Aufgaben sind heute als Übung erreichbar. Tagesaufgaben und Übungen nutzen denselben Vorrat. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards (8 repeating daily + 4 practice)

Solo

Limited beta ready

Twelve finite authored English boards comprise eight repeating UTC daily boards and four separate practice boards. Seven candidates are judged against two stated endpoint roles using supplied readings; all three distinct valid bridges must be found. This is a designed evidence rubric, not unrestricted word association or a live graph measurement. One-sided plausibility does not earn a discovery. Each distinct wrong candidate costs 15 points, the optional exclusion report costs 12, and four distinct misses end the attempt without a solved result even if one bridge was found. Empty and repeated checks are free; Reveal is explicitly unscored. The fifteen-action ceiling exceeds legal seven-candidate attempts and is not a free-selection completion trap. Selected candidates and committed discoveries recover in this browser, with warned memory-only progress if storage fails and no account or cross-device synchronization. Save checks are non-atomic. The mapped player is authored only; retained unannotated custom-provider services were not exercised. Unprimed human role understanding, whether the three discoveries feel distinct, fairness, enjoyment, difficulty and replay interest remain unmeasured; live-provider and production readiness remain unverified.

Paid reports now exclude an unchecked candidate and retain the choice through later actions and reload. Stored action schema and score remain unchanged; historical hint text is reconstructed with the corrected prefix policy. Independent complete play and storage/peer checks passed. Technical readiness and human feedback are tracked separately in conf/tasks/games-quality-20261013/game-quality.json; human feedback remains pending.

Authored endpoint roles and candidates with separate evidence for each side.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

Eine Verbesserung umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-20. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-sixth-ten-rounds-20261020/evidence/words/ledger.json

2 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json

Mit deutschsprachigen Ersttestern Verständnis, strittige Antworten, genutzte Hilfen und eine freiwillige weitere Runde beobachten. Mit anderen Varianten derselben Richtung vergleichen; menschliche Rückmeldung steht aus.

Invite unprimed human playtesting and assess the recorded limitations; implement observed problems through bounded repairs with independent verification.

Reassessed 11 October 2026 after two completed local play, review, improvement and replay cycles. Photography 73 was completed before and after startup repair, then a clean 100; phone Astronomy 55 was followed by a genuine four-miss loss despite one discovery and a clean 100. All fourteen endpoint readings, seven candidate verdicts, paid chronology, canonical Copy and active/terminal recovery were inspected. Startup observation, temporary-read and fixed-daily Copy recovery were repaired with strict schemas, rules, deck and template unchanged. The second round corrected only singular status wording and four exact expectations after its original baseline. Final verification passed 143 client checks, eleven fresh Java checks before the display-only wording change, twelve distinct owning desktop/phone browser journeys and four additional inherited compatibility journeys. Independent source/wording reviews and bounded desktop/phone views passed. Retains Limited beta ready, 4, and the historical role-understanding and distinct-discovery gap. These local results do not establish atomic storage, human fairness, enjoyment or replay interest, live/custom-provider quality or production readiness.

Maturity gap: Test role misunderstanding and whether three bridge discoveries feel distinct enough. Eight repeating UTC daily boards and four separate practice boards, authored English content and device-local progress.

13 October coordinated quality acceptance: Paid reports now exclude an unchecked candidate and retain the choice through later actions and reload. Stored action schema and score remain unchanged; historical hint text is reconstructed with the corrected prefix policy. Independent complete play and storage/peer checks passed. Human feedback is pending; no production or live-provider validation is claimed. The historical maturity rating remains unchanged. Earlier finding: The original exact-key startup interleaving could leave the owning observer behind the shared recovered attempt and repeatedly refuse real actions. Observation now aligns after shared initialization and after reconciliation, using existing content/rules headers and recovered-current equality guards. Separate candidate regressions proved that one temporary alignment-read outage could strand practice progress even after reads recovered, including warned failed-write play, and that a daily result reconstructed as a new object could repeatedly refuse Copy. The final unconditional guarded observation addresses all three extensions without replacing malformed or incompatible saves. Inherited peer focus, accepted-action guards, protected Replay, paid rules and shared asynchronous Copy already passed the original browser baseline; they are compatibility evidence, not newly discovered faults. The second round corrected only the one-remaining-miss display and four corresponding wording expectations. Final owning and inherited replays passed; no further defect was established. Browser-local reconciliation remains non-atomic. The default annotated route is BridgeHunterProController.playable, BridgeHunterProAuthoredDeck and bridge_hunter_pro_beta.jte, using bridge-pro-beta-deck.json with the shared fourth-deduction.js and fourth-deduction-rules.js. Inventory identity game-31 remains distinct from persisted runtime identity game31. The authored-v1 envelope uses fourth-deduction-3 and exactly six state fields: actions, draft, text, selected, phase and replay. Strict committed-action recovery, daily Replay prohibition, authored content identity and scoring are unchanged. The retained unannotated language/source/target controller method and custom service are not the mapped authored player. The first actual desktop cycle used the Photography and Weather practice board. Root read all fourteen endpoint readings and the rules. Empty and repeated empty checks were free. Lens failed one role for a 15-point cost, and the one exclusion report cost 12 and explained Tripod. Light and Fog became distinct discoveries; selected Snow and two of three discoveries survived active reload. Snow completed the attempt at 73. Root read all seven full verdict/reason/two-sided explanations and all five chronological paid receipts, copied the canonical result, recovered the terminal result and inspected a bounded readable desktop view. The original 140-client run passed 139 checks and failed one, with zero skips in 827.937917 milliseconds. All 112 pure checks, sixteen inherited shared DOM checks and eleven other owning DOM checks passed. The exact-key startup fixture allowed arrival or first real action to adopt a newer selection, but the original observer stayed stale and the full attempt never completed. No completion is claimed for that controlled failing case. All eight original desktop/phone browser journeys passed with zero skips in 39.1 seconds after full 73, 100 and 88 outcomes, warned memory/manual Copy, all finite evidence and recovery. Existing ordinal-76 acceptance, focus, Replay and paid-rule protections therefore remain compatibility evidence; no new capacity or native-focus fault was established. The first narrow repair aligned the main observer after shared initialization. It passed all original 140 clients, eleven fresh Java checks and the initial preview build. Actual fresh desktop Photography again earned 73 with the paid Lens miss and exclusion report, all three discoveries, selected Snow recovery, all fourteen readings, seven verdicts, five receipts, canonical Copy and terminal reload. Independent source review then identified a temporary third-read failure sequence; it was reproduced in a separate new full-completion fixture against that candidate. After outer read A and shared recovery B, the alignment read failed once. Later readable storage could still cause repeated refusal of every action. A second separate case reproduced the same failure while writes remained unavailable. Both failed before the follow-up, so no full result is credited for those failures. An identity-gated observer update fixed both practice cases, but independent review identified the shared daily-terminal branch reconstructing an equal logical result as a different object. Its separate fixture completed a genuine daily 100, then reproduced repeated Copy refusal without saved-byte corruption. The final functional repair calls the existing guarded observeMain after shared reconciliation as well as after initialization. Its content/rules header and recovered-current equality checks prevent observing incompatible, malformed or genuinely different snapshots; failed-write local progress remains distinct from durable storage. The three new cases prove later-read practice completion, warned full memory 100 followed by durable fresh 100, and canonical fixed-daily Copy without rewriting the saved result. The original 140 test bodies, rules, deck, template and both owning browser files were unchanged during this functional repair. Root fresh verification passed 143 clients with zero skips in 722.990708 milliseconds, and eleven fresh Java checks at 07:15:28 UTC; the isolated build succeeded in eleven seconds. Served runtime and the packaged 42-reviewed inventory were checked. Independent final source review passed. The final reviewed actual desktop Replay earned a clean 100 after Light, Fog, selected Snow active recovery and final Snow, with all fourteen readings, seven full verdicts, three paid receipts, canonical Copy and terminal reload. Eight repaired first-round desktop/phone browser journeys passed with no skips in 40.6 seconds. The second actual phone cycle used Astronomy and Navigation. Route Motif, Telescope and Crater were three distinct misses costing 45 points; Coordinates and Horizon were found, and active reload retained two discoveries and five paid receipts. Star completed all three discoveries for 55. Root read all seven verdict/reason/two-sided explanations and six chronological receipts, copied the canonical result, recovered it after reload and reviewed a bounded readable phone view. The remaining-miss status incorrectly used the plural for one. All four original second-round browser journeys passed with zero skips in 23.5 seconds before that wording correction, including genuine exhaustion and a fresh 100. The authorized round-two repair changed only the visible count-dependent miss/misses wording. Exactly four matching expectations changed: one owning DOM regex, two owning second-round strings and one inherited completed-76 second-round string. Normalized byte comparisons confirmed every other fixture byte, boundary and case count unchanged; no new wording-only test was added. Rules, deck, template, strict schema and all three startup fixtures remained unchanged. All 143 client checks passed after the wording correction in 846.83675 milliseconds, and independent wording review passed. Under the final display runtime, root actually found Star, then missed Route Motif, Telescope and Crater on the phone. The corrected status showed one different miss left, 55 points available and one of three discoveries. Active reload retained all four chronological receipts. Rudder was a genuine fourth distinct miss, ending the attempt without a solved result despite the earlier Star discovery. Root read all seven full verdicts and fourteen endpoint readings plus five receipts, copied the canonical unsolved result and recovered it after terminal reload. A fresh explicit Replay found Star, Horizon and Coordinates for 100 with all seven verdicts, fourteen readings and three paid receipts, canonical Copy, terminal recovery and a bounded readable phone screenshot. The display-only package completed in eight seconds and its served runtime hash was verified. Four final owning second-round desktop/phone journeys passed with zero skips in 23.3 seconds. Four separately anchored inherited ordinal-76 second-round journeys passed with zero skips in 7.5 seconds, covering the shared singular wording without claiming new ordinal-76 review credit. The owning rounds provide twelve distinct passing browser journeys, eight after the functional repair and four after the second-round wording repair; the four pre-wording second-round passes are the same cases and are not double-counted. The eleven fresh Java checks preceded the display-only wording correction; Java behavior remained unchanged. Root read every final browser log, and independent functional and wording source reviews passed. The final runtime SHA-256 is 41254af5c8a21439f5c7b65609692e5a039f8727ab9db61c91167a1bd88feedb; rules and owning template hashes remain 0b2f7490175d8002c1c405f8c538c1d4f1105068b8499d7379b18ebaaad8a18d and 70ed900f74e8f76333fd5de919b851fa3ce844716445a36e41b56d8829007b6e. These local agent outcomes and controlled checks do not establish atomic storage, human semantic agreement or enjoyment, live/custom-provider quality, deployment or public readiness.

• modules/games/src/main/java/io/kgard/web/slice/games/game31/BridgeHunterProController.java
• modules/games/src/main/resources/games/bridge-pro-beta-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game31/BridgeHunterProAuthoredDeckTest.java
• modules/games/docs/games-maturity-20261010-fourth-prototype-beta.json
• conf/tasks/evidence/fourth-ten-games-20261010/deduction
• conf/tasks/evidence/fourth-ten-games-20261010/final-verification.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/57-bridge-hunter-pro/verification.json
• conf/tasks/games-two-round-review-20261011/evidence/57-bridge-hunter-pro/round-1-before
• conf/tasks/games-two-round-review-20261011/evidence/57-bridge-hunter-pro/round-1
• conf/tasks/games-two-round-review-20261011/evidence/57-bridge-hunter-pro/round-2-before-wording
• conf/tasks/games-two-round-review-20261011/evidence/57-bridge-hunter-pro/round-2-final
• conf/tasks/games-two-round-review-20261011/evidence/57-bridge-hunter-pro/inherited-76-r2
• modules/games/src/main/java/io/kgard/web/slice/games/game31/bridge_hunter_pro_beta.jte
• modules/games/src/main/resources/static/js/games/fourth-deduction.js
• modules/games/src/main/resources/static/js/games/fourth-deduction-rules.js
• modules/games/src/test/js/fourth-deduction.test.cjs
• modules/games/src/test/js/fourth-deduction-ui.test.cjs
• modules/games/src/test/js/bridge-hunter-pro-ui.test.cjs
• modules/games/src/test/js/bridge-hunter-pro-startup-ui.test.cjs
• modules/web-visual/playful-games-tests/bridge-hunter-pro.spec.ts
• modules/web-visual/playful-games-tests/bridge-hunter-pro-round2.spec.ts
• modules/web-visual/playful-games-tests/missing-link-daily-round2.spec.ts
• conf/tasks/games-quality-20261013/results/verification-bridge-hint.md

Infer a hidden concept from staged contextual clues. Use the visible answer bank and trade points for more evidence before four distinct misses.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben und Rückmeldungen. Die englische Ausgabe bleibt separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle vier Aufgaben sind heute als Übung erreichbar. Tagesaufgaben und Übungen nutzen denselben Vorrat. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards: eight rotating UTC daily + four separate practice. Eight-concept answer bank, five clues, four distinct misses.

Solo

Limited beta ready

Finite authored clues and aliases; no live Contextoid is redacted on the mapped route. Content repeats, and browser-local saving is best-effort and non-atomic. Human discovery, difficulty, enjoyment and retention remain unverified.

The 13 October independent review accepted the original Clueveil edition after its final constellation clue was narrowed to an official bounded sky region and help accurately described accepted names/forms. The changed edition preserves real old saved bytes on arrival; free repeats, paid clues, accepted normalized input and replay passed. Accepted for invited human playtesting; human feedback remains pending. This is local technical evidence, not deployment or public-release signoff.

Authored mystery concept, staged clues, aliases and clue interpretations.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

Eine Verbesserung umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-22. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-ninth-ten-rounds-20261022/evidence/words/ledger.json

3 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json

Mit deutschsprachigen Ersttestern Verständnis, strittige Antworten, genutzte Hilfen und eine freiwillige weitere Runde beobachten. Mit anderen Varianten derselben Richtung vergleichen; menschliche Rückmeldung steht aus.

Collect unprimed human feedback on comprehension, fairness, difficulty and replay interest; technical readiness and human feedback remain separate in conf/tasks/games-quality-20261013/game-quality.json.

Latest recorded source-informed local play/review; carried forward, not a fresh human or production playtest.

Maturity gap: Twelve authored English boards constrain replay and the eight-day daily cycle repeats. Source-informed functional play does not prove novice discovery, enjoyment or difficulty balance. Progress is device-local; human beta should observe clue use, misunderstanding and willingness to play again. Typed spelling aliases remain finite.

Two ordered local rounds completed on 11 October 2026. Original and repaired Fermenting practice plays earned 66 after one miss and two paid clues, then 100 on a fresh Replay. Each completed result reviewed all five clue interpretations, the global explanation, supported alias, title, note, last check and exact chronological receipts, followed by actual canonical Copy and terminal recovery. Active reload preserved padded raw drafts exactly. Ordinary native Baking focus, final-clue focus and peer Replay input focus already worked in the original. Help was read only in the original baseline. All 47 original clients, eight original browser journeys (22.4 seconds) and eleven fresh Java checks passed. Before that baseline, eleven prepared fixture literals across the owning DOM and two browser files were calibrated to the unchanged normalization contract: padded case works, internal ELECTRIC__BATTERY works, but trailing separators leave an unsupported space. A new cold-read regression then demonstrated unknown paid-66 save overwrite while writes still worked. Its calibrated version completed prior, memory, actual replacement and fresh durable outcomes before the captured byte verdict; the first version stopped early during fallback and remains calibration evidence only. A separate three-branch observer test completed fallback 88/100 reviews before failing its first deferred original-window completion assertion. Initial durable Reveal first manual Copy and Replay passed as compatibility. R1 enabled readable-save protection only for game5/redacted and appended the calibrated cold-read and durable-Reveal cases. All 61 selected clients passed, followed by eight R1 browser journeys in 22.1 seconds on runtime 23963e0d. Native 66 and 100 replays completed their full audits, Copy and recovery on bounded desktop and phone views. The observer test still failed on this revision and was explicitly carried into R2. No new ordinary focus, scoring or capacity defect is claimed. R2 first completed a genuine Battery loss on R1: four paid clues and four distinct misses, with repeated CIRCUIT free and all eight receipts reviewed. One guarded observeMain call after the existing reconciliation save then aligned the recovered snapshot; a separate frozen three-branch case covers writable, blocked-write and daily recovery. The existing header and recovered-state equality checks remain in force. This shared alignment call does not opt other editions into readable-save protection. The real prior loss recovered unchanged because content identity did not change. Final native replays completed genuine 0, assisted 42, one-guess unscored Reveal and clean 100. Every result included all five interpretations and exact receipts, actual canonical Replay Copy and terminal reload; active capacitor and padded ELECTRIC__BATTERY drafts recovered exactly. Root reviewed bounded desktop and phone views. Final runtime 7fa50f94ac12f70c145dbca43c73d6101f56da23d26e6a0c7fcbf662dedd0209 passed 62 clients in 958.99175 ms: fifteen family pure, seventeen old family DOM, eighteen owning DOM and twelve Hidden Destination compatibility checks. Four R2 browsers passed in 15.9 seconds. The eight R1 and four R2 cases are twelve distinct repaired journeys with twenty-two custom captures across two revisions, not twelve cases rerun on the final revision. Eleven distinct fresh original Java checks are retained without an R1/R2 rerun. All 37 backups verified; 33 current inputs stayed exact and four declared files changed. The 6,044-byte pure prefix, calibrated fifteen-case prefix and all seventeen R1 bodies remain exact. Three whole-outcome cases were appended. Deck, sixty authored interpretation rows, scoring, six normalized recovery fields, daily policy, JTE, shared helper and Java stayed unchanged. Both isolated compiled packages contained canonical73; no fat-archive or deployment claim follows. The mapped route uses twelve finite authored boards, eight daily and four practice. The retained custom/provider method is unannotated, and its fallback list does not establish live Contextoid provenance. Saving remains device-local, best-effort and non-atomic. Finite aliases, clue difficulty, novice discovery, enjoyment and retention remain unverified. Maturity 4, the 10 October review date, basis and historical gap are preserved. Earlier known-issue record, retained for history: Cold unreadable-save overwrite and transient observer-recovery failures were repaired. No additional unresolved defect was established within the finite authored checks.

• modules/games/src/main/java/io/kgard/web/slice/games/game5/RedactedContextoidController.java
• modules/games/src/main/resources/games/redacted-contextoid-beta-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game5/RedactedContextoidAuthoredDeckTest.java
• modules/games/src/test/java/io/kgard/web/slice/games/game5/RedactedContextoidRoundServiceTest.java
• modules/games/docs/games-maturity-20261010-third-prototype-beta.json
• conf/tasks/evidence/third-ten-games-20261010/deduction
• conf/tasks/evidence/third-ten-games-20261010/final-verification.json
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/actual-round1.json
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/actual-round2.json
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/baseline-closed.json
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/fixture-calibration.json
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/round1-improvement.json
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/round2-improvement.json
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/verification.json
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/final-source-preservation.json
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/original-native-cloud-report.json
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/round2-native-cloud-report.json
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/original-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/original-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/extra-original-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/extra-v2-original-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/round2-before-observer-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/round1-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/round1-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/round2-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/round2-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/original-java-results.json
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/round1-applied-source.json
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/round2-applied-source.json
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/classpath-round1-package.json
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/classpath-round2-package.json
• modules/games/src/test/js/redacted-contextoid-ui.test.cjs
• modules/web-visual/playful-games-tests/redacted-contextoid.spec.ts
• modules/web-visual/playful-games-tests/redacted-contextoid-round2.spec.ts
• modules/web-visual/playful-games-tests/redacted-contextoid-support.ts
• conf/tasks/games-two-round-review-20261011/evidence/26-redacted-contextoid/independent-lower-final-review.json
• conf/tasks/games-quality-20261013/results/verification-batch-006.md

Assign twelve words to four context shelves. Shared words can fit several shelves; every complete supported partition is accepted.

Deutsche Bedienung, Aufgaben, Rückmeldungen und Ergebnislinks im Starterpaket. Die ursprüngliche englische Ausgabe bleibt unter ihrer bisherigen Adresse erhalten; ihre Aufgaben und Spielstände sind getrennt.

English

Alle 6 Aufgaben sind heute als Übung erreichbar. Davon 2 zusätzliche Übungen außerhalb der Tagesrotation. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards: eight rotating UTC daily + four separate practice. 12 words, four shelves of three; several supported full partitions.

Solo

Limited beta ready

Finite authored English memberships and assigned costs, not live Contextoid ranks or measured audience agreement. Twelve repeating boards and browser-local progress; saving and peer preflight are non-atomic, with no account synchronization. Human difficulty, enjoyment, novice understanding and general device/accessibility behavior remain unverified. Retained unannotated provider code can expose synthetic fallback without a provenance marker; it was not played or verified live.

Verified unchanged after complete independent desktop/320 play, alternate paid outcomes, reload/replay and cold-read save recovery. Technical readiness and human feedback are tracked separately in conf/tasks/games-quality-20261013/game-quality.json; human feedback remains pending.

Authored context memberships and explanations; word-to-shelf assignment with overlapping allowed meanings.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

2 Verbesserungen umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-20. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-ten-rounds-20261020/evidence/words/ledger.json

Die Sortierung „Weitere Prüfrunden“ zählt alle nachfolgenden Prüfungen zusammen. Die ersten fünf Runden bleiben separat.

2 von höchstens 10 Runden am 2026-10-20: Kategorien, Teiltreffer, Hinweise und gespeicherte Erklärungen geprüft. Keine weitere konkrete Änderung aus den gespielten Fällen abgeleitet.

• conf/tasks/games-nonviral-further-rounds-20261020/evidence/words/ledger.json

3 frühere Prüfeinträge · 1 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json
• conf/tasks/games-nonviral-five-rounds-20261020/evidence/words/ledger.json

Neue und bisherige Aufgaben mit deutschsprachigen Ersttestern vergleichen: Verständnis, strittige Antworten, genutzte Hilfen und freiwillige weitere Runde beobachten. Die neuen Aufgaben sind noch nicht durch Teilnehmer validiert.

Invite unprimed human playtesting and assess the recorded limitations; implement observed problems through bounded repairs with independent verification.

Latest recorded source-informed local play/review; carried forward, not a fresh human or production playtest.

Maturity gap: Can novices recognize the supported alternate meanings and comfortably revise a complete partition? Eight repeating UTC daily boards and four separate practice boards, authored English content and browser-local progress.

13 October coordinated quality acceptance: Verified unchanged after complete independent desktop/320 play, alternate paid outcomes, reload/replay and cold-read save recovery. Human feedback is pending; no production or live-provider validation is claimed. The historical maturity rating remains unchanged. Earlier finding: The observed local save, peer-focus, stale-guidance and singular-wording issues were repaired and retested. No additional unresolved defect was established in these bounded journeys; broader human and live-provider gaps remain. Two local play–review–improve–replay–review cycles completed for ordinal 22, Context Clusters original at /games/game1. The annotated route uses twelve authored revision-two boards: eight rotating UTC daily boards and four practice boards. All twelve source boards contain 144 card incidences, 48 printed criteria and 182 literal membership readings. Native and browser journeys use Waves09 and Royal10, each with twelve audit rows and fourteen supported readings. The retained provider method is unannotated; its concrete fallback can return synthetic RelatedTerms without an explicit provenance marker. It receives no live-provider credit here. The immutable original was the already-reviewed shared a546 runtime, not the older historical manifest snapshot. Original 40 client cases passed 32 and failed 8; all 15 family pure and 9 older UI cases passed, while the 16 owning DOM bodies split 8/8. Original eight browser cases split 4/4 in 15.6 seconds after whole outcomes and finite proofs. Ordinary peer changes detached native selects, displaced focus and retained stale advice; delayed free repeated Check failed to adopt a completed peer and rejected actions left incorrect focus. Exact saved bytes and rejected Replay URL already passed. A separate original two-case run passed 1 and failed 1 in 602.9635 ms: cold unavailable reads with working writes replaced a genuine Waves loss. The first final byte comparison failed only after actual memory 77/100, read-return recovery and fresh durable 77/100; later captured warning/focus/share assertions were masked. FIRST durable initial-Reveal exact manual Copy and accepted Replay passed as compatibility. Original native play completed Waves 77 after one failed allocation, free repeat and paid Hint, then a primary 100 on a measured phone. Root reviewed all twelve rows/fourteen readings, criteria, plans, receipts and arithmetic, verified actual three-line canonical clipboard bytes and terminal recovery. The original Help paragraphs were read then. Native peer Replay and ordinary assignment lost focus; accepted Hint, own terminal H2, own Replay to the first physical Echo select and Copy-feedback cleanup already worked. R1 enabled the existing readable-save, preflight, observation and accepted-state machinery for game1 and retained native map-shaped shelf rows. It preserved earlier game18/game7 behavior, schema, scoring, content and share semantics. Two exact whole-outcome cases were appended after the unchanged original sixteen-body prefix. All 42 client cases passed in 1391.378334 ms. Repaired native Waves 77 and100 repeated complete evidence, exact clipboard and recovery; unrelated peer assignment retained Echo, and peer Replay reached co-feedback with current guidance. Eight R1 browsers passed in 15.2 seconds with eight custom captures. Help was not reopened. R2 first completed five Royal outcomes on R1: genuine three-check zero loss, alternate 100, two-miss 70, a two-assignment unearned Reveal and clean 100. Root observed singular remaining-check errors, alongside the earlier Waves one-failed-check formula. A new six-outcome test failed before repair only at its first final formula verdict (475.711 ms overall); later final assertions were masked. The repair only adds game1 to the existing singular-wording predicate, calibrates one exact browser-support expression and appends one whole-outcome DOM case. It changes no rules, content, identities or recovery contract. On final e377, root completed six native outcomes: Royal loss 0, alternate 100, primary 70, partial unearned Reveal and clean 100, then Waves 77. Every final outcome is Replay; the initial pre-change Royal loss was not. Each received full twelve-row/fourteen-reading review, four criteria/capacities, actual plan and ordered receipts, applicable formula or no-plan explanation, H2 focus, exact canonical clipboard bytes and full terminal reload. Desktop was 1280×720; both phone plays were independently measured 390×844 before actions. Root inspected the final native loss, alternate and Waves screenshots and all six final CLI captures. Correct 1 check left/remains and 1 earlier failed check were directly observed. No native cold simulation or new Help read is claimed. Final 43 clients passed in 1266.669583 ms: 15 family pure, 9 older family UI and 19 owning DOM, comprising 27 owning and 16 other-family checks. Four final R2 browser cases passed in 8.9 seconds. The twelve distinct repaired browser case/project checks are eight R1 plus four R2 across two revisions, with fourteen custom captures; they are not twelve runs on the final wording revision. Twelve distinct fresh original Java tests passed (four mapped authored, three retained controller, five retained service); none were rerun for these JavaScript-only repairs. Final 39-reference guards verify every backup, 36 unchanged inputs and only runtime/owning DOM/support edits. Pure/export 4892 bytes, original sixteen-body 18,083-byte prefix and R1 eighteen-body 26,008-byte prefix remain exact. Both browser journey bodies, all deck content, JTE, CSS, Java and shared helpers are unchanged. The final 596-entry classpath contains 18,060,310 copied bytes and contains canonical 77; source/compiled/HTTP identity is explicitly recorded. This is local evidence, not a fat-jar release, deployment, Linux/device-wide study or human beta. Browser-local synchronization is best effort and non-atomic; maturity 4, its 2026-10-10 date, basis and historical novice-understanding gap remain unchanged.

• modules/games/src/main/java/io/kgard/web/slice/games/game1/ContextClustersController.java
• modules/games/src/main/resources/games/context-clusters-original-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game1/ContextClustersAuthoredOriginalDeckTest.java
• modules/games/src/test/java/io/kgard/web/slice/games/game1/ContextClustersControllerTest.java
• modules/games/docs/games-maturity-20261010-fifth-prototype-beta.json
• conf/tasks/evidence/fifth-ten-games-20261010/context
• conf/tasks/evidence/fifth-ten-games-20261010/final-verification.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/baseline-closed.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/original-native-cloud-report.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/original-client-reached-review.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/independent-original-reached-review.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/original-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/extra-original-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/original-java-results.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/original-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/original-served-runtime.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/actual-round1.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/repaired-native-cloud-report.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/round1-improvement.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/round1-applied-source.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/round1-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/round1-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/round2-before-native.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/round2-before-grammar-client.log
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/independent-lower-round2-grammar-proposal.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/round2-applied-source.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/actual-round2.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/round2-native-cloud-report.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/round2-improvement.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/round2-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/round2-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/round2-served-runtime.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/classpath-round2-package.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/final-source-preservation.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/final-browser-outcomes.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/verification.json
• conf/tasks/games-two-round-review-20261011/evidence/22-context-clusters-original/independent-lower-final.json
• conf/tasks/games-quality-20261013/results/verification-batch-003.md

Find four explained word groups in eight rotating daily puzzles or four separate practice boards. Near-miss feedback, saved progress and shareable results.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben und Rückmeldungen. Die englische Ausgabe bleibt separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle vier Aufgaben sind heute als Übung erreichbar. Tagesaufgaben und Übungen nutzen denselben Vorrat. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards: eight rotating UTC daily + four separate practice. 16 words, four groups of four per board.

Solo

Limited beta ready

One authored partition per board rejects other plausible groupings. Experimental custom groups can overlap and use first-anchor allocation; no live-provider validation is claimed. Content repeats, and browser-local saving is best-effort and non-atomic without account sync. Human semantic agreement, difficulty, enjoyment and retention remain unverified.

Verified unchanged by the independent six-game pilot, including complete daily/practice/reload and compact layouts. Technical readiness and human feedback are tracked separately in conf/tasks/games-quality-20261013/game-quality.json; human feedback remains pending.

Authored four-way word groups and explanations.

Custom Contextoid board via anchors/language/scope and related settings; synthetic placeholders labelled. Separate experimental rules.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

Eine Verbesserung umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-20. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-fifth-ten-rounds-20261020/evidence/words/ledger.json

5 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json

Mit deutschsprachigen Ersttestern Verständnis, strittige Antworten, genutzte Hilfen und eine freiwillige weitere Runde beobachten. Mit den anderen Varianten derselben Richtung vergleichen; menschliche Rückmeldung steht aus.

Invite unprimed human playtesting and assess the recorded limitations; implement observed problems through bounded repairs with independent verification.

Latest recorded source-informed local play/review; carried forward, not a fresh human or production playtest.

Maturity gap: The finite English collection accepts one authored partition per board; broader plausible sets can still be rejected. Familiarity, semantic agreement and enjoyment need human testing. The retained custom provider mode remains experimental.

13 October coordinated quality acceptance: Verified unchanged by the independent six-game pilot, including complete daily/practice/reload and compact layouts. Human feedback is pending; no production or live-provider validation is claimed. The historical maturity rating remains unchanged. Earlier finding: Observed native-focus, stale-action adoption, rejected Replay URL and cold unreadable-save overwrite failures were repaired. Terminal guidance no longer promises persistence during memory-only play. No additional unresolved defect was established in the finite authored checks. Two ordered local rounds completed on 11 October 2026. Original and repaired Shape play completed four groups after two different misses, a free repeated hypothesis and journal correction, then a reverse-order clean win. Every result reviewed all four labels, sixteen words, four literal explanations and the actual miss receipts/grid, with active selection recovery, terminal reload, actual canonical clipboard bytes and bounded desktop/phone views. Phone gameplay used a separately measured 390 × 844 tab. Original peer selection displaced native word and still-eligible Revisit focus; peer Replay displaced Copy focus. R1 retained those eligible controls and moved stranded Replay focus to the play heading. Own terminal H2 focus, current Selected 0/4 guidance, cleared Copy feedback and correct adopted Quadrilaterals guidance already worked; they are compatibility, not newly fixed native faults. Help was read only originally. Original authored clients passed twenty-five and failed eight of thirty-three: nine pure and eight older UI checks passed, with eight passes/eight failures in sixteen owning DOM cases. Separate controlled custom rendering/Copy cases both passed. Fifteen fresh Java checks passed: three authored-deck, four mapped play-route, three retained controller and five retained service checks. Original browser cases passed four and failed four in 29.7 seconds after full outcomes and deferred boundary reviews. Delayed free repeated Check preserved saved peer bytes but failed to adopt its completed state; rejected Replay preserved latest saved results but removed show_context and focused the active heading. Earlier failures in loop-based client cases do not establish later unexecuted branches. Three additional full-outcome cases passed one and failed two: cold unreadable-but-writable storage overwrote a genuine Kitchen loss during memory play, and ordinary peer edits replaced an eligible Revisit control. Both final verdicts followed full outcomes, Copy and recovery. An earlier two-case selection repeats these same failures and adds no distinct tests. Initial durable Reveal FIRST manual Copy and accepted Replay already passed. R1 applied the reviewed adapter with readable-save protection, normalized observation, free/paid action preflight, accepted-effect and Replay URL gates, retained native word/Revisit controls and focus recovery. Three exact whole-outcome cases were appended after the original sixteen-case prefix. All thirty-six authored checks passed in 996.35075 ms and all eight R1 browser cases passed in 27.9 seconds on runtime 17caa960. Native replay repeated the complete Shape journal and clean-win flows, including actual Copy and recovery. Rules, authored content and the experimental custom behavior were preserved; no scoring, startup or capacity defect is inferred. R2 first completed four Kitchen outcomes on R1: a genuine loss with one earned group and four misses, a reverse clean win, a Reveal retaining only two earned groups, and a one-miss journal-corrected win. All group proofs, actual ordered receipts, grids, Copy and recovery were reviewed. Seeing fixed-on-device wording in durable native play did not itself prove a false claim. A separate calibrated memory-only case then failed only its final persistence-wording assertion, after two full four-group wins, canonical manual Copy, an actual empty reload and a visible storage warning had passed. The initial warning-element mismatch was corrected in source before any execution, so it receives no failed-fixture credit. R2 changed exactly one terminal sentence to This attempt is finished. Review the connections below. and appended that one whole-outcome regression. Final native replay repeated all four Kitchen outcomes on runtime b44b3b99ccc038fda265f47030c78cc6be2ce176b19a194908f3f0f544dcc22c. All final outcomes have Replay=true; the first before-change loss did not. The final loss preserves the actual second miss order fold, stir, whisk, poach after an unordered free repeat; the other misses use whisk, stir, fold with boil, simmer and steam. Loss, clean win, partial Reveal and corrected win each received all sixteen words/four reasons, earned-versus-revealed status, H2 focus, corrected feedback, actual canonical clipboard verification and full terminal recovery. Active reload retained the one-group/three-miss loss state and all four corrected selected words. Root reviewed bounded desktop/phone images and all six final CLI captures. No native cold-storage simulation or fresh Help reread is claimed. Final thirty-seven authored checks passed in 1039.296667 ms: nine pure, eight older UI and twenty owning DOM cases. The two original custom checks and fifteen distinct original Java passes are retained without repair reruns. Four R2 browser cases passed in 19.1 seconds; together with R1 this is twelve distinct repaired cases across two revisions and fourteen custom captures, not twelve cases rerun on the final revision. All thirty-eight original backups verified; thirty-six current inputs remained exact and only runtime/owning DOM changed. The one-literal inverse restores R1, the 4,804-byte pure/export prefix remains exact, and all original sixteen bodies plus the three R1 bodies remain unchanged before the final append. Deck, Java, templates, shared helpers and browser journeys are unchanged. Both isolated 596-entry compiled packages contained canonical76. A failed sandbox HTTP invocation provides no verification credit; root subsequently verified the final source/compiled/HTTP identity. These are local classpaths, not deployment evidence. The English authored collection has twelve finite sixteen-word boards, each accepting one four-group partition; plausible alternative groupings can be rejected. Custom provider groups use a separate experimental path and can overlap before first-anchor allocation; controlled tests are not live-provider validation. Browser-local progress is best-effort and non-atomic without account synchronization. Human semantic agreement, fairness, difficulty, enjoyment, retention and general device coverage remain unverified. Maturity 4, its 10 October review date, basis and historical gap are preserved.

• modules/games/src/main/java/io/kgard/web/slice/games/game1/v2/ContextClustersV2Controller.java
• modules/games/src/main/resources/games/context-clusters-v2-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game1/v2/ContextClustersV2AuthoredDeckTest.java
• modules/games/src/test/java/io/kgard/web/slice/games/game1/v2/ContextClustersV2RoundServiceTest.java
• modules/games/docs/games-maturity-followup-20261010.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-1-v2/round1-independent-review.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-1-v2/round1-root-transcript.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-1-v2/round1-partial-reload.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-1-v2/round1-terminal-reload.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-1-v2/final-transcript.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-1-v2/final-verification.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/actual-round1.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/actual-round2.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/baseline-closed.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/original-native-cloud-report.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/repaired-native-cloud-report.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/round2-before-native.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/round2-native-cloud-report.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/round1-improvement.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/round2-improvement.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/verification.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/final-source-preservation.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/final-browser-outcomes.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/original-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/original-custom-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/original-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/extra-original-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/extra-two-only-original-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/original-extra-review.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/round1-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/round1-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/round2-before-memory-client.log
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/round2-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/round2-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/original-java-results.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/round1-applied-source.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/round2-applied-source.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/classpath-round1-package.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/classpath-round2-package.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/round2-served-runtime.json
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/independent-lower-round2-honest-proposal.json
• modules/games/src/test/js/context-clusters-v2-runtime.test.cjs
• modules/web-visual/playful-games-tests/context-clusters-v2.spec.ts
• modules/web-visual/playful-games-tests/context-clusters-v2-round2.spec.ts
• modules/web-visual/playful-games-tests/context-clusters-v2-support.ts
• conf/tasks/games-two-round-review-20261011/evidence/23-context-clusters-v2/independent-lower-final.json
• conf/tasks/games-quality-20261013/results/verification-pilot.md

Recruit five words for your faction. Own, shared, rival and trap words score differently; finish with at least eight points and ahead of the rival.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben und Rückmeldungen. Die englische Ausgabe bleibt separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle vier Aufgaben sind heute als Übung erreichbar. Tagesaufgaben und Übungen nutzen denselben Vorrat. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards: eight rotating UTC daily + four separate practice. Eight word cards; recruit five.

Solo

Limited beta ready

Solo scoring against authored faction totals; no actual rival player or adaptive opponent. Finite repeating content and non-atomic browser-local saves. Corrected board fingerprints reset old affected progress with a notice. Human fairness, onboarding, difficulty, enjoyment and retention remain unverified.

The13 October independent review accepted Context Duel after stable display ordering removed the alternating faction shortcut, three rival-role wording conflicts were corrected, and private result focus was made visible. All96 cards and12display patterns were checked; old saved IDs/earned results and three changed editions remain protected. Actual loss/win, native play and true read-only/peer recovery passed. Accepted for invited human playtesting; human feedback remains pending. This is local technical evidence, not deployment or public-release signoff.

Authored own/shared/rival/trap word allegiances with explanations.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

Eine Verbesserung umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-22. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-ninth-ten-rounds-20261022/evidence/strategy/ledger.json

3 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json

Mit deutschsprachigen Ersttestern Verständnis, strittige Antworten, genutzte Hilfen und eine freiwillige weitere Runde beobachten. Mit anderen Varianten derselben Richtung vergleichen; menschliche Rückmeldung steht aus.

Collect unprimed human feedback on comprehension, fairness, difficulty and replay interest; technical readiness and human feedback remain separate in conf/tasks/games-quality-20261013/game-quality.json.

Latest recorded source-informed local play/review; carried forward, not a fresh human or production playtest.

Maturity gap: Human beta feedback is needed on fairness, unprimed onboarding, difficulty and replay interest. The authored starter bank is finite.

Two ordered local rounds completed on 11 October 2026. Original and repaired Laboratory plays earned 9–1 and 8–2. Every result reviewed all eight authored roles/reasons, recruited markers, title, note and totals, actual canonical Copy and terminal recovery. Active reload retained committed picks while deliberately discarding the uncommitted selected card. Repaired ordinary peer updates retained the focused Waiter button; peer Replay restored first-enabled focus and current advice. Help was read in the original baseline only. Original verification passed eight of nineteen clients and four of eight browser journeys (23.7 seconds). Browser failures followed complete earned outcomes and exposed replaced native controls, displaced focus and stale requested-action feedback. Controlled newer-save bytes, URL and active first-enabled focus already passed. A separate cold unreadable-but-writable regression completed memory, recovered and fresh durable outcomes before demonstrating unknown 3–8 save overwrite. The other extra case completed a fresh 9–1 before failing accepted-Replay focus; its first durable initial-Reveal manual Copy already passed. No original scoring, startup or capacity fault is claimed. The R1 adapter added current-save observation and preflight, readable-save protection, acceptance checks before requested effects, cached native cards, current feedback and scoped focus rescue. The pure score/recover prefix stayed exact: recovery returns only picks, phase and replay, tolerates discarded extra fields and legacy daily replay:true, and validates at most five distinct known picks. Selected cards remain intentionally unsaved. Shared helpers, templates, Java and scoring stayed unchanged. Two whole-outcome DOM cases were appended after the original seventeen bodies. All twenty-one clients passed, followed by eight R1 browser journeys in 22.2 seconds on runtime 2bf87e79. R2 Swimming Pool initially completed a genuine 3–8 loss with one- and four-pick recovery. It exposed the one-recruit resumed sentence and Goggles explanation grammar. One display expression, eleven related reason strings across six boards and one exact browser-support literal were corrected. Six content fingerprints therefore changed; the real old cd-10 loss correctly reset to 0/5 with the existing changed-edition notice, rather than retaining old progress. Fresh native 3–8 loss, 8–2 win, one-pick 0–3 unscored Reveal and reverse 9–1 win then completed all eight reasons/markers, actual Copy and recovery, with bounded desktop and phone views. Corrected one-recruit feedback and Goggles belong were directly observed; native play of the other ten boards is not claimed. Final runtime 2ba00750cf7a86908acde7f606130fb9c4f4a5ae1032d5009a371baefba25a78 passed all twenty-one clients in 853.785125 ms. Four final browser journeys passed in 16.4 seconds. The eight R1 and four R2 cases provide twelve distinct repaired journeys and twenty custom captures across two revisions, not twelve runs on the final wording revision. All thirty-five original backups verified; thirty-one inputs stayed exact and only runtime, deck, appended owning DOM and support changed. All nineteen DOM bodies stayed unchanged during R2, the 1,056-byte pure prefix remained exact and both browser journey files were preserved. Twelve distinct fresh original Java tests passed; the same three authored-deck tests freshly passed again after the deck change, without new controller/service rerun credit. Both isolated compiled packages contained canonical72; this inventory refresh is not a fat-archive or deployment claim. The mapped route uses twelve finite authored boards: eight daily and four practice. The retained custom/provider method is unannotated; its heuristic or synthetic fallback is not evidence of a live default provider. Solo faction totals are authored, not a human or adaptive rival. Local saving is best-effort and non-atomic. Human fairness, onboarding, difficulty, enjoyment and retention remain unverified. Maturity 4, the 9 October review date, basis and historical gap are preserved. Earlier known-issue record, retained for history: Local save, native-focus and stale-feedback faults were repaired; resumed wording and eleven explanation strings were corrected. No additional unresolved defect was established within the finite authored checks.

• modules/games/src/main/java/io/kgard/web/slice/games/game6/ContextDuelController.java
• modules/games/src/main/resources/games/context-duel-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game6/ContextDuelControllerTest.java
• modules/games/src/test/java/io/kgard/web/slice/games/game6/ContextDuelRoundServiceTest.java
• modules/games/docs/games-maturity-20261009-prototype-beta.json
• conf/tasks/evidence/prototype-beta-20261009/semantic-games
• conf/tasks/evidence/prototype-beta-20261009/final-verification.json
• conf/tasks/evidence/prototype-beta-20261009/final-surfaces/surface-check.json
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/actual-round1.json
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/actual-round2.json
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/baseline-closed.json
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/round1-improvement.json
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/round2-improvement.json
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/verification.json
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/final-source-preservation.json
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/original-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/original-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/extra-original-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/round1-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/round1-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/round2-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/round2-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/original-java-results.json
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/round2-java-results.json
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/round1-applied-source.json
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/round2-applied-source.json
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/round2-before-native.json
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/round2-native-cloud-report.json
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/classpath-round1-package.json
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/classpath-round2-package.json
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/independent-lower-proposal.json
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/independent-path-round2-wording-proposal.json
• modules/games/src/test/js/context-duel-ui.test.cjs
• modules/web-visual/playful-games-tests/context-duel.spec.ts
• modules/web-visual/playful-games-tests/context-duel-round2.spec.ts
• modules/web-visual/playful-games-tests/context-duel-support.ts
• conf/tasks/games-two-round-review-20261011/evidence/27-context-duel/independent-lower-final-review.json
• conf/tasks/games-quality-20261013/results/verification-batch-013.md

Aim word projectiles using the sense that fits the printed context. Break the kaiju shield within the printed charge budget while protecting the city; different successful volleys earn equal credit.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben und Rückmeldungen. Die englische Ausgabe bleibt separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle vier Aufgaben sind heute als Übung erreichbar. Tagesaufgaben und Übungen nutzen denselben Vorrat. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards: eight rotating UTC daily + four separate practice. Six projectile words with two readings each; five usable words and one decoy.

Solo

Limited beta ready

Twelve repeating authored English encounters with declared meanings and numeric effects. Original play targets a printed context, distinct from v2. Saves are device-local and non-atomic, without account sync. Human fairness, discovery, enjoyment, difficulty, retention, live-provider quality and production readiness remain unverified.

The13 October independent review accepted original Context Kaiju after deterministic private sense ordering removed the universal card-position shortcut. All144 meanings and12 displayed orders were reviewed. Native option identity, original reading IDs, all content editions and genuinely earned prior draft/100/assisted90 records remain intact; actual loss, alternate volleys and peer recovery passed. Accepted for invited human playtesting; human feedback remains pending. This is local technical evidence, not deployment or public-release signoff.

Authored word senses tied to a printed target context, with assigned damage and charge cost.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

Eine Verbesserung umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-20. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-seventh-ten-rounds-20261020/evidence/learning/ledger.json

4 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json

Mit deutschsprachigen Ersttestern Verständnis, strittige Antworten, genutzte Hilfen und eine freiwillige weitere Runde beobachten. Mit anderen Varianten derselben Richtung vergleichen; menschliche Rückmeldung steht aus.

Collect unprimed human feedback on comprehension, fairness, difficulty and replay interest; technical readiness and human feedback remain separate in conf/tasks/games-quality-20261013/game-quality.json.

Latest recorded source-informed local play/review; carried forward, not a fresh human or production playtest.

Maturity gap: Test how readily novice players understand the meaning choices and how quickly the finite encounters become familiar. Eight repeating UTC daily boards and four separate practice boards per edition; authored English meanings, connections and numerical effects; browser-local progress. Source-informed agent play does not establish unprimed human discovery, enjoyment, difficulty or retention. Production, real devices and assistive technology remain unverified.

The mapped /games/game14 route remains the original Context Kaiju player: ContextKaijuAuthoredDeck, context_kaiju_original_beta.jte and kaiju-original-beta.js, with runtime identity game14 and rules kaiju-original-3. Its twelve authored English boards comprise eight repeating UTC daily encounters and four practices. Each offers six words with two readings, five usable projectiles and one decoy. The printed target context and charge budget distinguish this edition from hidden-context v2. Retained provider/controller methods remain unmapped; no live custom mode is established. Matching readings damage the shield; every fired word consumes its assigned charges, while a miss removes 25 city safety. A win earns 100 minus 10 per miss and 10 for the single optional Aim assist. Reveal and unsuccessful defenses score zero. A surviving city can still lose when no unused projectile is affordable. Different affordable winning volleys receive equal base credit. Free selection and aiming snapshots are saved; they do not consume a paid event. The strict seven-field state, event chronology, daily replay restriction and scoring remain unchanged, with no reachable free-edit capacity trap established. The original native Workshop round, kaiju-original-09, recovered the selected Plane and literal reading before firing. Plane and Saw earned 100, clearing the shield with two charges and 75 city remaining. Review read all twelve meanings, both paid receipts, the score equation and the example plus seven-minimal-volley benchmark. Actual canonical Copy and terminal recovery passed at a measured 1280 by 720 desktop, page width 1265. A Grain misfire, Plane assist and Plane/Saw then completed 80 with no charges and 50 city, all twelve reasons and four paid rows reviewed, copied and recovered on a 390 by 844 phone. Native peer Replay left focus on hidden ko-share; peer assist also left obsolete Grain-miss feedback. These actual observations are separate from automated document-body focus failures. Original verification passed 38 of 46 client checks, with eight owning DOM failures and zero skips in 818.229209 milliseconds. Four of eight original desktop/phone browser journeys passed and four failed in 19.0 seconds; all completed their full outcomes before deferred boundary verdicts, and all eight outcome captures were available. The original shared helper already protected ordinary changed-save bytes; the failures concerned stale feedback, native control identity, free/refused-action reconciliation and focus or Replay effects. Existing startup and memory-play checks passed, so no original startup, scoring or generic storage-failure defect is claimed. After baseline closure, an appended cold-read case demonstrated unknown durable 80-point bytes being overwritten when only reads failed and writes worked. It completed memory 100, subsequent recovery handling and fresh durable assisted 90 with full reviews before failing the aggregate byte/warning verdict on the unchanged original in 400.5095 milliseconds. A separate durable initial-Reveal case passed the original in 368.62225 milliseconds; it is compatibility coverage for immediate Copy and accepted Replay, not an original Reveal defect. Both pass the repair. The original seventeen-case DOM file and fixture remain an exact 20,821-byte prefix, with only these two full cases appended. The owning repair checks current saved state before paid, free and refused actions, navigation and Copy. Accepted state identity controls effects independently of write durability. Existing requireReadableSave protection keeps unknown unreadable bytes intact while allowing warned memory play. Compatible recovered-state observation aligns initialization and reconciliation; Replay effects follow shared acceptance. Cached magazine and same-word aiming controls preserve native focus, while adoption clears stale guidance and rescues hidden or disabled focus to the actual active feedback or result. Existing asynchronous Copy protections remain. The embedded pure/export prefix is byte-identical at 3,865 bytes; strict recovery, rules, scores, deck, Java, JTE/CSS and shared persistence are unchanged. These localStorage checks are not atomic transactions. Repaired native Workshop Replay earned a fresh 100 with the complete twelve-reason, two-receipt, arithmetic and benchmark review, actual Copy and terminal recovery. Peer Replay now focused visible ko-feedback, and the same-word aiming select retained focus during peer changes. All 48 client checks passed in 786.800292 milliseconds: 23 pure, six existing DOM and nineteen owning DOM cases. All eight final first-round browser journeys passed in 17.1 seconds with eight completed captures. Sixteen fresh original Java checks passed at 15:05:35.652 through 15:05:37.069 UTC on 11 October: six mapped authored checks, three retained unmapped controller checks and seven retained service checks. Unchanged Java evidence is retained; subsequent compilation is not a fresh test rerun or provider validation. The second actual round used Space, kaiju-original-10, on a measured 390 by 844 phone. Stage assist, Stage and Burn earned 90 using all eight charges with city 75. Reload retained the assist and aimed draft; all twelve meanings, three paid rows, arithmetic and two-volley benchmark were read, copied and recovered. Probe's wrong reading followed by Burn produced a genuine affordability loss: score zero, shield 40, two charges and city 50. The surviving city and unaffordable remaining words explain that loss; it is not a city breach or v2 mutation outcome. Both receipts, all twelve reasons, audit and benchmark were reviewed with active and terminal recovery and actual Copy. No further source change was justified. Fresh Dock, Probe and Capsule earned 100; active reload after Probe retained shield 25, three charges and city 75. All twelve reasons, three receipts, arithmetic, benchmark, canonical Copy and terminal recovery passed. Bounded phone review showed readable content with page width 375 inside 390. Four final second-round browser journeys passed in 9.0 seconds with six completed captures. The final total is twelve distinct owning journeys on identical repaired source, with fourteen captures and none unavailable; original runs are additional executions, not more distinct final cases. Final runtime SHA-256 is 46eb2a693da5568f5310e686962297f5279adc15179114a3e809c4927ac39737. Independent reviews and the final audit confirm only runtime and the appended owning DOM suite changed among twenty-six protected inputs. The immutable local compiled classpath contains 596 entries and 17,852,383 copied bytes; served source matches the repair and packed canonical inventory matches the sixty-three-review base d1dae5bb8ab50a0f51564de9815f432ee5bbd354fdc321956cfdc1de52e4bb64. This is not a fat archive, deployment or Linux verification. Browser play covers two boards and twenty-four distinct meanings; DOM also reviews Harbor, while all twelve boards and 144 reasons were source-reviewed and pure checks enumerate their affordable wins. Maturity 4, the 10 October 2026 date, basis and historical gap remain exact. Finite authored judgments, non-atomic local saves, absent account/cross-device sync and unverified human fairness, discovery, enjoyment, difficulty, retention, live-provider quality, real-device assistive access and production remain limits. Earlier known-issue record, retained for history: The 11 October review repaired stale feedback, native focus, rejected-action effects and unreadable-save overwrite. Both local cycles and final checks passed; no further material defect was established. Maturity remains 4.

• modules/games/src/main/java/io/kgard/web/slice/games/game14/ContextKaijuController.java
• modules/games/src/main/resources/games/kaiju-original-beta.json
• modules/games/src/test/java/io/kgard/web/slice/games/game14/ContextKaijuAuthoredDeckTest.java
• modules/games/src/test/java/io/kgard/web/slice/games/game14/ContextKaijuRoundServiceTest.java
• modules/games/docs/games-maturity-20261010-final-prototype-beta.json
• conf/tasks/evidence/final-two-games-20261010/kaiju/cycle1/play.json
• conf/tasks/evidence/final-two-games-20261010/kaiju/cycle2/play.json
• conf/tasks/evidence/final-two-games-20261010/kaiju/cycle3/play.json
• conf/tasks/evidence/final-two-games-20261010/kaiju/final/play.json
• conf/tasks/evidence/final-two-games-20261010/kaiju/final-report.md
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/36-context-kaiju/baseline.json
• conf/tasks/games-two-round-review-20261011/evidence/36-context-kaiju/actual-round1.json
• conf/tasks/games-two-round-review-20261011/evidence/36-context-kaiju/actual-round2.json
• conf/tasks/games-two-round-review-20261011/evidence/36-context-kaiju/original-client.log
• conf/tasks/games-two-round-review-20261011/evidence/36-context-kaiju/original-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/36-context-kaiju/original-browser-outcomes.json
• conf/tasks/games-two-round-review-20261011/evidence/36-context-kaiju/original-cold-read.log
• conf/tasks/games-two-round-review-20261011/evidence/36-context-kaiju/original-durable-reveal.log
• conf/tasks/games-two-round-review-20261011/evidence/36-context-kaiju/additional-regressions.json
• conf/tasks/games-two-round-review-20261011/evidence/36-context-kaiju/java-original.json
• conf/tasks/games-two-round-review-20261011/evidence/36-context-kaiju/round1-proposed-runtime.diff
• conf/tasks/games-two-round-review-20261011/evidence/36-context-kaiju/round1-client.log
• conf/tasks/games-two-round-review-20261011/evidence/36-context-kaiju/round1-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/36-context-kaiju/round2-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/36-context-kaiju/final-browser-outcomes.json
• conf/tasks/games-two-round-review-20261011/evidence/36-context-kaiju/final-hashes.json
• conf/tasks/games-two-round-review-20261011/evidence/36-context-kaiju/classpath-round1-package.json
• conf/tasks/games-two-round-review-20261011/evidence/36-context-kaiju/verification.json
• modules/games/src/main/resources/static/js/games/kaiju-original-beta.js
• modules/games/src/test/js/kaiju-original-ui.test.cjs
• modules/web-visual/playful-games-tests/context-kaiju-original.spec.ts
• modules/web-visual/playful-games-tests/context-kaiju-original-round2.spec.ts
• modules/web-visual/playful-games-tests/context-kaiju-original-support.ts
• conf/tasks/games-quality-20261013/results/verification-batch-014.md

Fire words that identify a hidden context, avoid its named rival, and protect the city. Damage, mutation and wrong names have visible consequences.

Deutsche Bedienung, Aufgaben, Rückmeldungen und Ergebnislinks im Starterpaket. Die ursprüngliche englische Ausgabe bleibt unter ihrer bisherigen Adresse erhalten; ihre Aufgaben und Spielstände sind getrennt.

English

Alle 6 Aufgaben sind heute als Übung erreichbar. Davon 2 zusätzliche Übungen außerhalb der Tagesrotation. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards: eight rotating UTC daily + four separate practice. Eight projectile words per wave.

Solo

Limited beta ready

Twelve repeating authored English waves; no live Contextoid-generated waves. Health, city and mutation are game resources, not points or measured semantic truth. Saves are device-local and non-atomic. Human fairness, difficulty, enjoyment, retention and production readiness remain unverified.

The 11 October review repaired stale guidance, invalid native focus, rejected-action effects and unknown-save overwrite during read outages. Both local cycles passed; no further material defect was established. Maturity remains 4. Independent collection review, 15 October 2026: Varied signal display retains card identity; microphone-specific evidence and reliable complete context/result placement. Fresh three-game verification passed112 client checks,53 browser cases and a complete36-board review, including native context/receipt visibility, old earned records and separate storage/peer controls. Accepted for invited human playtesting; human feedback remains pending.

Authored hidden context, named rival, signal/bridge/rival cards and damage/mutation rules.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

2 Verbesserungen umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-20. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-third-ten-rounds-20261020/evidence/strategy/ledger.json

2 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json
• conf/tasks/games-nonviral-five-rounds-20261020/evidence/strategy-next/ledger.json

Neue und bisherige Aufgaben mit deutschsprachigen Ersttestern vergleichen: Verständnis, strittige Antworten, genutzte Hilfen und freiwillige weitere Runde beobachten. Menschliche Rückmeldung steht noch aus.

Invite a small cohort; record puzzle-specific confusion, unfair judgments and failures before expanding content or claiming public-release readiness.

Latest recorded source-informed local play/review; carried forward, not a fresh human or production playtest.

Maturity gap: Human beta feedback is needed on fairness, unprimed onboarding, difficulty and replay interest. The authored starter bank is finite.

The mapped /games/game14/v2 route remains Context Kaiju v2's authored player, using ContextKaijuV2Controller, context_kaiju_v2_beta.jte and the packaged kaiju-v2-beta.json deck. Runtime identity game14v2, authored-v1 storage and rules version kaiju-beta-1 are unchanged. Twelve English waves comprise eight repeating UTC daily boards and four separate practices, each with eight projectile words. Retained dynamic-provider methods are unmapped; their Java checks do not establish a live custom mode. The browser review played Camera and Kitchen, reading their sixteen distinct role explanations. All twelve boards and ninety-six authored explanations were source-reviewed; the pure checks cover safe outcomes for all twelve. This does not mean all twelve were played in a browser. Outcomes use Kaiju health, city integrity and mutation, not an earned points score. A signal removes 25 health; a bridge removes 10 health and adds 20 mutation; a rival removes 30 city and adds 30 mutation. A wrong context name removes 20 city and adds 20 mutation. Naming the correct context can win with positive health; reducing health to zero also wins if the city survives. Empty and normalized repeated names are free. All current waves contain four signals, two bridges and two rivals. The finite deck supports genuine city-breach losses; no mutation-only loss or event-capacity trap was established. These role assignments and consequences are authored game rules, not measured semantic truth or human fairness evidence. The original first native round used Camera, wave-09, at a measured 1280 by 720 desktop with page width 1265. All eight labels and the complete help were read. An empty name was free. Shutter committed health 75, city 60 and mutation 0; active reload retained that action and disabled projectile while intentionally clearing the unsubmitted text. Shutter, photograph, lens and viewfinder then completed a damage victory at health 0, city 60 and mutation 0. All eight terminal role explanations and four chronological action rows were inspected, actual Copy contained the canonical Contextoid URL, and terminal reload preserved the outcome. A separate original mixed attempt used the wrong name microphone, recording and raw ' CAMERA ' to win at health 90, city 40 and mutation 40, with all eight reasons and three actions reviewed, copied and recovered. Ordinary peer Replay exposed actual focus and guidance defects: on the 390 by 844 receiving phone, focused Copy became hidden while focus remained on beta-share rather than a usable ammunition button. After a peer Replay followed by microphone and recording, resources showed the active 90/40/40 attempt but the message still said Kaiju subdued. The original twenty-one client checks passed twelve and failed nine, with zero skips in 599.321459 milliseconds: nineteen owning checks and two Warm Cold compatibility checks. Original browser verification passed four and failed four of eight desktop/phone journeys in 23.5 seconds, after complete outcomes and finite reviews; eight completed-outcome captures were present and none unavailable. The failures concerned native control identity and focus, stale feedback and accepted-action effects after changed-save rejection. Existing shared changed-save protection and the passing original startup case are compatibility evidence, not newly discovered original scoring or startup failures. After those baselines closed, an appended cold-read regression used a real prior mixed victory at health 90, city 40 and mutation 40. With reads unavailable but writes working, the unchanged original overwrote unknown durable bytes and hid the warning. The case completed a full memory damage victory at 0/60/0 and a fresh durable naming victory at 100/60/0, including Copy and recovery, before its aggregate preservation verdict. It failed once with zero skips in 341.54475 milliseconds. A second appended durable initial-Reveal case passed the unchanged original in 229.519625 milliseconds and passed the repair; it protects immediate Copy and accepted Replay under the new observer without asserting an original Reveal defect. The entire original sixteen-case DOM file, including its fixture and test bodies, remains an exact prefix; only these two full behavioral cases were appended. The owning runtime now reconciles compatible saved state before paid actions, free or refused actions and Copy. Saved-state identity gates local effects separately from persistence success, preserving warned memory play without treating a rejected action as accepted. Fire, name submission and Replay apply their local focus, input and URL effects only after acceptance; a rejected name is not cleared. Initialization and post-reconciliation observation align only compatible recovered snapshots, and fingerprint comparison excludes savedAt. The existing requireReadableSave option is enabled for this player so unavailable reads cannot silently overwrite an unknown durable save. The shared persistence helper itself is unchanged. These are localStorage read/write guards, not atomic transactions or account synchronization. Ammunition controls are retained by identity while their text, details and availability update. Adoption clears obsolete terminal guidance, preserves still-usable native and outside focus, and rescues focus from hidden or disabled controls to the actual active or terminal target. Storage notification handling captures focus before shared rendering. Deferred finish focus is fenced against later state or user focus changes, and existing asynchronous Copy guards are retained. State recovery still derives health, city, mutation and phase from committed actions, tolerates extra legacy metadata and accepts only literal true for replay; the daily wrapper retains its existing replay restriction. Raw unsubmitted names remain intentionally unsaved. The rules, normalization, deck, Java, JTE, CSS and shared helper were not changed. All twenty-three repaired client checks passed with zero skips in 615.088042 milliseconds: twenty-one owning checks, comprising eighteen DOM and three pure checks, plus two Warm Cold compatibility checks. Actual repaired peer Replay focused enabled ammunition and replaced obsolete guidance. A fresh native desktop four-signal Replay ended at health 0, city 60 and mutation 0 with all eight reasons, four actions, actual canonical Copy and terminal recovery reviewed. All eight ammunition buttons were disabled at the terminal result. The eight final first-round browser journeys passed in 24.5 seconds with eight completed captures and none unavailable. Eleven fresh original Java checks passed on 11 October 2026, with XML timestamps from 14:49:10.491 through 14:49:11.820 UTC: one mapped authored-deck check, three retained unmapped controller checks and seven retained service checks. Java, JTE, deck and rules remained unchanged throughout this review. That fresh original evidence is retained; the subsequent successful compile and immutable classpath staging are not described as a final Java test rerun, a live-provider check or a deployment. The second native round used Kitchen, wave-10, at a measured 390 by 844 phone viewport. All eight labels and the complete rules were read. Window, the wrong name bedroom and duvet formed three committed actions; repeating raw ' BEDROOM ' was free, and active reload preserved the attempt. Raw ' KiTcHeN ' completed a surviving naming victory at health 90, city 10 and mutation 70. All four action rows and eight result reasons were reviewed, followed by actual canonical Copy and terminal recovery. A fresh Replay fired duvet and pillow, recovering after the first rival, and honestly lost at health 100, city 0 and mutation 60. Both paid actions and all eight role explanations were reviewed, copied and recovered. No points or mutation-only loss were inferred. The complete second-round review justified no further source change. A fresh Replay focused the name input and then fired the four signals, winning at health 0, city 60 and mutation 0. All four actions and eight reasons, actual canonical Copy, terminal recovery and eight disabled ammunition controls were checked. Bounded native phone screenshots were readable, with page width 375 inside the 390-pixel viewport. All four final second-round browser journeys passed with zero skips in 7.7 seconds, producing six completed-outcome captures and none unavailable. The final total is twelve distinct owning desktop/phone journeys, eight from round one and four from round two, on the identical repaired runtime. Fourteen captures were completed and none unavailable; original runs are additional executions, not extra distinct final cases. Independent recorded source, diff, log and prefix review passed, alongside root's complete evidence reviews and actual desktop/phone plays. Final runtime SHA-256 is 62356f4764010eb47c4e42c593e64162b7f2d9dd887b851d482669a6325cbd47; owning DOM SHA-256 is f8241b2fadfc6e0c3f584d38843368cdaa826244dea220d37b96670634335161. The twenty-six-input audit shows exactly those two changed files and twenty-four unchanged inputs, including all browser journeys and support. The immutable local compiled classpath has 596 entries and 17,832,782 copied bytes; its served runtime matches the source and its packed sixty-two-review canonical inventory matches cfe8356e802057c3d0cc7c0a702ba3299a540455ea8643c702569dea5f355dc0. No fat archive, production, Linux, live provider, account or cross-device verification is claimed. Finite repeating content, non-atomic local persistence and unverified human difficulty, enjoyment and retention remain limitations. Maturity 4, its 9 October 2026 date, basis and historical gap are preserved exactly, without adding a historical-gap field.

• modules/games/src/main/java/io/kgard/web/slice/games/game14/v2/ContextKaijuV2Controller.java
• modules/games/src/main/resources/games/kaiju-v2-beta.json
• modules/games/src/test/java/io/kgard/web/slice/games/game14/v2/ContextKaijuV2ControllerTest.java
• modules/games/src/test/java/io/kgard/web/slice/games/game14/v2/ContextKaijuV2AuthoredDeckTest.java
• modules/games/docs/games-maturity-20261009-prototype-beta.json
• conf/tasks/evidence/prototype-beta-20261009/temperature-kaiju
• conf/tasks/evidence/prototype-beta-20261009/final-verification.json
• conf/tasks/evidence/prototype-beta-20261009/final-surfaces/surface-check.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/baseline.json
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/actual-round1.json
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/actual-round2.json
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/original-client.log
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/original-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/original-browser-outcomes.json
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/additional-regressions.json
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/original-cold-read.log
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/original-durable-reveal.log
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/java-original.json
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/round1-proposed-runtime.diff
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/round1-source-audit.json
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/round1-client.log
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/round1-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/round2-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/final-browser-outcomes.json
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/final-hashes.json
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/classpath-round1-package.json
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/round1-served-parity.json
• conf/tasks/games-two-round-review-20261011/evidence/37-context-kaiju-v2/verification.json
• modules/games/src/main/resources/static/js/games/kaiju-v2-beta.js
• modules/games/src/main/resources/static/js/games/kaiju-v2-rules.js
• modules/games/src/test/js/kaiju-v2-ui.test.cjs
• modules/web-visual/playful-games-tests/context-kaiju-v2.spec.ts
• modules/web-visual/playful-games-tests/context-kaiju-v2-round2.spec.ts
• modules/web-visual/playful-games-tests/context-kaiju-v2-support.ts
• conf/tasks/games-quality-20261013/results/batch-064-resumed-three-variant-repair.md
• conf/tasks/games-quality-20261013/results/verification-batch-064-resumed-three-variant.md

Baut zu zweit eine gemeinsame Wortkette mit geheimen Zielen. Oder übe gegen einen Computer und erkenne seinen Plan.

Ohne Englischkenntnisse spielbar.

Deutsche Oberfläche und Wörter

Eigene und geteilte Runden werden unabhängig gespeichert.

Versioniertes, endliches redaktionelles Spielpaket mit erklärten Wortbeziehungen und eigenen Spielregeln.

Zwei Personen mit geheimen Zielen oder Computerübung

Playable prototype

Endlicher Wortschatz, keine Live-Contextoid-Abfrage. Browseridentitäten sind keine verifizierten Personen; Tempovergleiche hängen auch von Gerät und Verbindung ab.

Agenten-Spieltests belegen keine menschliche Beliebtheit oder Verbreitung.

Versioniertes, endliches redaktionelles Spielpaket mit erklärten Wortbeziehungen und eigenen Spielregeln.

Teilbare Ausgangslagen und eigene Spielentscheidungen.

Spring prüft Züge; PostgreSQL speichert Partien, Beiträge, Ergebnisse und Wiederholungsbelege.

Menschliche Rückmeldungen zu Regeln, Wortschatz und Weitergabe einholen.

Mit unabhängigen Spielern Verständlichkeit, Spielspaß und Wiederholungswert prüfen.

Umsetzung eines eigenständigen Konzepts aus Word Game Ideas; lokale Spielzyklen werden separat belegt.

Maturity gap: Unabhängige menschliche Spieltests zu Verständlichkeit, Spielspaß und Zugänglichkeit stehen aus.

Ergänzung zur historischen 99-Spiele-Prüfung; deren Befunde bleiben erhalten. Am 25. Oktober 2026 wurden 9 neue lokale Browser-Spielzyklen geprüft, mit zwei abwechslungsreichen fehlerfreien Schlussrunden nach der letzten Korrektur. Das ist kein menschlicher Beliebtheitsnachweis; Bereitstellungen haben eigene datierte Nachweise.

• modules/games/docs/remaining-word-games.md
• conf/tasks/remaining-word-games-20261024/source-thread.md
• conf/tasks/remaining-word-games-20261024/doppelkontext/play-review-summary.json
• conf/tasks/word-games-improvement-20261025/worlds/completion.json
• conf/tasks/word-games-improvement-20261025/README.md
• modules/games/docs/releases.md

Find one word that satisfies three authored semantic axes. Four guesses, optional clue, 30 daily/practice puzzles and local recovery.

Oberfläche, Regeln und vorbereitete Aufgaben sind auf Deutsch.

German

Neben dem Tagesrätsel ist Üben im vorhandenen Rätselbestand möglich; bekannte Rätsel bleiben Wiederholungen.

30 versioned German puzzles; three axes per puzzle; four guesses; daily and practice.

Solo daily and practice

Limited beta ready

Finite 30-day Berlin-time cycle; plausible unlisted answers may be rejected; no independent language-editor validation; client-side solutions visible; device-only progress. Local storage peer checks are best effort, not atomic or remote/account synchronization. Native final LEUCHTEN centre wrapped onto two lines at 686 px; full word remains in history/proofs and no CSS change was made. Controlled local desktop/emulated-phone and native laptop play does not establish unprimed human learning, physical-phone behavior, OS Share delivery or production durability.

No additional unresolved reproducible defect remains in the reviewed isolated local journeys after protecting unreadable prior progress, peer adoption, queued modal focus and obsolete Copy/Share feedback. See unchanged dated maturity gaps and limitations. Independently accepted for invited human playtesting on13 October2026. Canonical German/Unicode input, expanded uppercase saved attempts and native selected replacement are repaired; compatible peer progress preserves unfinished draft/caret while stale actions remain refused. Phone labels and full native endings pass the recorded checks. Independent032/037/041 evidence covers all30boards,177 readings, genuine old receipts, storage boundaries and native input/layout. Human feedback and production validation remain pending.

Authored three-axis word associations with explicit accepted answer dictionaries and explanations; intersection across all three axes decides a win.

None. No live KGARD Contextoid, Termpath, large lexicon, or model API connection.

Browser JavaScript; browser-local saved progress.

Den tatsächlichen Weg vom geteilten Ergebnis zum spielenden Empfänger prüfen; keine unbestätigten Wachstumswerte ableiten.

Collect uncoached German-language feedback on alternate meanings, difficulty and phone readability.

Latest recorded source-informed local play/review; carried forward, not a fresh human or production playtest. Independently accepted for invited human playtesting on13 October2026. Canonical German/Unicode input, expanded uppercase saved attempts and native selected replacement are repaired; compatible peer progress preserves unfinished draft/caret while stale actions remain refused. Phone labels and full native endings pass the recorded checks. Independent032/037/041 evidence covers all30boards,177 readings, genuine old receipts, storage boundaries and native input/layout. Human feedback and production validation remain pending.

Maturity gap: Thirty fixed puzzles are a useful beta bank, but unexpected valid connecting words still depend on manually listed alternatives.

Static browser JavaScript; per-puzzle progress remains in this device’s localStorage, with best-effort same-browser peer adoption and no account, remote or cross-device synchronization. Optional WebMCP is feature-detected. Original tolerant restoration, version/mode/date identities, four guesses, Hint marker, authored content and spoiler-free result/challenge sharing remain unchanged. Two ordered isolated play/review/improvement/replay cycles passed 10 first-cycle browser cases and 4 second-cycle browser cases; each of the two second-cycle finite cases completed 65 whole games and all 177 authored proofs. The six final owning supplemental DOM cases, syntax check and detector passed; simulated Share contracts supplement the actual stale Copy browser proof.

• modules/games/cloud-games/dreisinn/src/puzzles.js
• modules/games/cloud-games/dreisinn/src/engine.js
• modules/games/cloud-games/dreisinn/README.md
• modules/games/cloud-games/build-static.mjs
• modules/games/docs/games-maturity-20261005.json
• modules/games/src/main/resources/static/games/cloud/dreisinn/index.html
• conf/tasks/evidence/games-maturity-20261005/root-browser-baseline.json
• conf/tasks/evidence/games-maturity-20261005/root-play-observations.json
• conf/tasks/games-two-round-review-20261011/partition-8-1-20261012/evidence/1-dreisinn/final-review.json
• conf/tasks/games-two-round-review-20261011/partition-8-1-20261012/evidence/1-dreisinn/changed-files.json
• conf/tasks/games-two-round-review-20261011/partition-8-1-20261012/evidence/1-dreisinn/final-child-verification.json
• conf/tasks/games-two-round-review-20261011/partition-8-1-20261012/evidence/1-dreisinn/round1-final-root-review.json
• conf/tasks/games-two-round-review-20261011/partition-8-1-20261012/evidence/1-dreisinn/round2-before-root-review.json
• conf/tasks/games-two-round-review-20261011/partition-8-1-20261012/evidence/1-dreisinn/native-round2-final-review.json
• conf/tasks/games-two-round-review-20261011/partition-8-1-20261012/evidence/1-dreisinn/final-packet-validation.json
• conf/tasks/games-two-round-review-20261011/partition-8-1-20261012/evidence/1-dreisinn/round2-final-root-review.json
• conf/tasks/games-quality-20261013/results/verification-batch-037-dreisinn-recovery.md
• conf/tasks/games-quality-20261013/results/verification-batch-041-dreisinn-peer-draft.md

Infer two endpoints from the middle of a connection. Several supported pairs count, with explanations for both ends.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben und Rückmeldungen. Die englische Ausgabe bleibt separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle vier Aufgaben sind heute als Übung erreichbar. Tagesaufgaben und Übungen nutzen denselben Vorrat. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards (8 repeating daily + 4 practice)

Solo

Limited beta ready

Endpoint alternatives are restricted to the authored bank and supported-pair list, not open-ended semantic equivalence or measured live graph paths. The finite English collection contains eight repeating UTC daily boards and four separate practice boards. Up to three distinct wrong-pair checks cost 20 each, and two optional clues cost 10 each; an accepted pair can still win on the third check. Exhaustion or Reveal earns zero, while repeated pairs and incomplete checks are free. Explanations and reviewed alternatives establish only the authored puzzle model. Progress remains browser-local without account or cross-device synchronization, and changed-save protection is a non-atomic preflight. Unavailable storage permits warned memory-only play. Retained service tests do not establish custom or live-provider behavior. Human fairness, unprimed onboarding, difficulty, replay interest, enjoyment and public production readiness remain unverified.

The13 October independent review accepted Endglean after redundant known-endpoint hints stopped charging, named hints and supported Bottle/Paintbrush/Lake alternatives were corrected, and daily result focus was made immediately visible. All12 genuine old editions and a real prior80 result remain protected; historic paid70 and real read-only/peer recovery pass. A retained native menu may become partly clipped when a peer notice expands, but verified keyboard continuation remains usable. Accepted for invited human playtesting; human feedback remains pending. This is local technical evidence, not deployment or public-release signoff.

Authored middle traces, endpoint role clues and accepted alternative pairs.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

Eine Verbesserung umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-22. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-ninth-ten-rounds-20261022/evidence/learning/ledger.json

5 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json

Mit deutschsprachigen Ersttestern Verständnis, strittige Antworten, genutzte Hilfen und eine freiwillige weitere Runde beobachten. Mit anderen Varianten derselben Richtung vergleichen; menschliche Rückmeldung steht aus.

Collect unprimed human feedback on comprehension, fairness, difficulty and replay interest; technical readiness and human feedback remain separate in conf/tasks/games-quality-20261013/game-quality.json.

Reassessed 11 October 2026 after two completed local play, review, improvement and replay cycles. Full morning-cup 60, grape third-pair 40, three-miss zero and fresh 100 attempts were reviewed with all connection explanations, supported pairs, canonical Copy and active/terminal recovery. Current-save acceptance, focus, catalog and wording were corrected, followed by the observed terminal-header earned-score mismatch. The calibrated startup fixture passed on the original client and is compatibility evidence. Final verification passed 23 client checks, thirteen fresh Java checks and twelve distinct final desktop/phone browser journeys, plus packaging, independent source review and bounded readable visual reviews. Retains Limited beta ready, 4. The historical need for human feedback on fairness, unprimed onboarding, difficulty, replay interest and the finite starter bank remains open. Local correctness does not establish human enjoyment, live/custom provider behavior or public production readiness.

Maturity gap: Human beta feedback is needed on fairness, unprimed onboarding, difficulty and replay interest. The authored starter bank is finite.

The mapped default Endglean route uses origins.jte, origins-deck.json and origins.js. Runtime identity game53 differs from inventory game-53; shared authored-v1 storage uses rulesVersion origins-1. The default player uses a finite authored bank of middle traces, endpoint clues and supported pairs, with no live provider request. The first full local cycle used a fresh morning-cup board, reading all three middle words, six source choices, six destination choices and complete instructions. An empty Check was free. Quarry/Coffee was one wrong pair costing 20; repeating it was free. Both hints cost another 20. Farm/Espresso selections, the wrong-pair count and both clues survived active reload. The completed result earned 60/100, with all four connection explanations, all four reviewed endpoint alternatives, the score formula, canonical Copy and terminal recovery agreeing. The owning All games link opened the daily listing, and singular Copy wording was visibly wrong. The initial original client run reported 13 passes and 10 failures across 23 checks, but one early startup fixture had an overstrict immediate-arrival assertion. The corrected exact-main-key startup check passed 1/1 on the original single-read client: its first Check could adopt the newer attempt and the complete journey earned 80. Original startup stranding was not demonstrated. A later preserved-original comparison with the calibrated startup fixture and updated grammar expectations reported 12 passes and 11 failures across 23 checks. The original eight desktop/phone browser journeys all reached complete earned outcomes before verdicts; two passed and six targeted cases failed in 43.8 seconds. The first repair checks the current save before free, paid and Copy actions, and applies requested effects only when the attempted state was accepted. Save durability remains separate, so unavailable storage permits warned memory-only play. Native menus and still-valid focus are retained. Paid hints focus their visible text, while changed peer state and rejected actions rescue hidden or disabled focus to active work or the actual result. Replay protects the current save, URL and requested effects when rejected. The new owning observer is aligned after shared startup recovery as compatibility protection, not a correction of a demonstrated original startup defect. The owning catalog destination and singular wrong-pair, hint and remaining-attempt wording were corrected. Transition, recovery, scoring, exports, accepted pairs, tolerant seven-field saved-state normalization and all eight original pure test bodies remain unchanged; shareText changes concern wording. All 23 client checks then passed, comprising eight pure and fifteen DOM checks, along with thirteen fresh Java checks: three authored-deck, three retained service and seven inventory checks. The eight repaired first-cycle desktop/phone journeys passed in 31.5 seconds. A fresh actual repaired morning-cup Replay repeated the complete 60-point attempt, all middle words, all twelve endpoint choices, four explanations, four supported pairs, canonical Copy and active/terminal recovery. The first bounded desktop review found a readable result without overflow. Round-one runtime SHA-256 was b61edfb4b627b299f77222e7aa5e19b8a0493beec7b0132e5c5fab327dd5e0ff. The second full cycle used a fresh phone grape trace. Vineyard/Beer and Farm/Cider were two distinct wrong pairs; repeating the latter was free. Both clues left 40 potential points. The Farm/Wine third-pair draft, previous guesses and clues survived active reload before the third distinct Check earned 40. All four connection explanations and both supported Vineyard/Wine and Farm/Wine pairs were reviewed, with canonical Copy and terminal recovery agreeing. Review then found a separate display defect during an actual Mine/Wine, Factory/Wine and Forest/Wine loss: the result and Copy correctly earned zero, while the closed header still claimed 40 points available. The minimal second repair retains available points during active play and displays earned points with an earned-points label on every terminal result. The restored existing loss then showed zero earned. A fresh explicit Replay repeated all three distinct wrong sources and again ended honestly at zero, with all four learning explanations, two supported alternatives, canonical Copy and terminal recovery agreeing. A further fresh Vineyard/Wine draft survived active reload and earned 100, with the same complete explanation, alternative, Copy and recovery review. Final verification passed all 23 client checks and thirteen fresh Java checks with no failures or skips. All twelve distinct final desktop/phone browser journeys passed together in 41.6 seconds; this final run includes both cycles rather than adding the earlier eight journeys as new cases. Source, packaged and served runtime hashes agreed, and the package matched the then-current 31-row refreshed canonical inventory. Independent source review passed both repairs, and the final static design check returned no findings. Bounded phone views of the repaired zero result, all four learning explanations and earned-score header were readable without horizontal overflow. Final runtime SHA-256 is 90a953e5da841e646e883b5271534566595707e3295fd643f2649ccf166b87a5. This is local source-informed agent and automated evidence; no human, live-provider or deployment validation is claimed. Earlier known-issue record, retained for history: Resolved and locally verified 11 October 2026: free, paid, Copy and Replay actions reconcile the current save and apply effects only to accepted state; native choices and useful focus survive peer changes, paid hints focus visible clues, and rejected actions preserve the actual attempt. The owning All games destination and singular feedback were corrected. The second-cycle repair makes every terminal header display earned points, so a three-miss loss shows zero earned instead of 40 available. The calibrated original startup check passed; observer alignment preserves that compatibility and is not evidence of an original startup-stranding bug. Final replays established no further defect.

• modules/games/src/main/java/io/kgard/web/slice/games/game53/OriginsController.java
• modules/games/src/main/resources/games/origins-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game53/OriginsAuthoredDeckTest.java
• modules/games/docs/games-maturity-20261009-prototype-beta.json
• conf/tasks/evidence/prototype-beta-20261009/path-games
• conf/tasks/evidence/prototype-beta-20261009/final-verification.json
• conf/tasks/evidence/prototype-beta-20261009/final-surfaces/surface-check.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/68-origins/verification.json
• conf/tasks/games-two-round-review-20261011/evidence/68-origins/final
• modules/games/src/main/java/io/kgard/web/slice/games/game53/origins.jte
• modules/games/src/main/resources/static/js/games/origins.js
• modules/games/src/test/js/origins.test.cjs
• modules/games/src/test/js/origins-ui.test.cjs
• modules/web-visual/playful-games-tests/origins.spec.ts
• modules/web-visual/playful-games-tests/origins-round2.spec.ts
• conf/tasks/games-quality-20261013/results/verification-batch-011-lake-followup.md

Infer both ends of a word journey using independent role clues. Several supported pairs earn equal credit; paid exclusions and four distinct attempts help refine your choice.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben und Rückmeldungen. Die englische Ausgabe bleibt separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle vier Aufgaben sind heute als Übung erreichbar. Tagesaufgaben und Übungen nutzen denselben Vorrat. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards (8 repeating daily + 4 practice)

Solo

Limited beta ready

Supported endpoint roles and pairs are finite; plausible everyday alternatives require editorial review, especially multifunctional devices and repurposed objects. Twelve authored English boards provide eight repeating UTC daily boards and four separate practice boards. Each distinct wrong pair costs 15 points, each of two exclusions costs 12 points, and four distinct misses exhaust the attempt; incomplete and repeated pairs are free. Progress is browser-local without account or cross-device synchronization. Changed-save protection is a non-atomic preflight rather than a simultaneous-write transaction. The current default route uses the authored deck; retained legacy service checks do not establish live-provider behavior. Human clue discovery, interpretation of rejected alternatives, difficulty, enjoyment, repeat interest and public production readiness remain unverified.

The 13 October independent review accepted free already-disclosed endpoint exclusions and clarified ink-reservoir, chemical-cell/electric-light and road-span roles. Three changed editions preserve prior saves on arrival. Historical one-hint73 and two-hint61 results retain their original chronology and scores. Accepted for invited human playtesting; human feedback remains pending. This is local technical evidence, not deployment or public-release signoff.

Authored middle journeys, independent endpoint-role clues and multiple supported pairs.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

Eine Verbesserung umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-20. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-fifth-ten-rounds-20261020/evidence/words/ledger.json

3 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json
• conf/tasks/games-nonviral-five-rounds-20261020/evidence/words-next/ledger.json

Mit deutschsprachigen Ersttestern Verständnis, strittige Antworten, genutzte Hilfen und eine freiwillige weitere Runde beobachten. Mit den anderen Varianten derselben Richtung vergleichen; menschliche Rückmeldung steht aus.

Collect unprimed human feedback on comprehension, fairness, difficulty and replay interest; technical readiness and human feedback remain separate in conf/tasks/games-quality-20261013/game-quality.json.

Reassessed 11 October 2026 after two completed local play, review, improvement and replay cycles. Complete 61, 46 and 100 point solves, four-pair exhaustion with no earned score, free refusals and repeats, two paid exclusions, all middle meanings, supported pairs, candidate reasons and receipts, Copy and recovery were checked. Reproduced peer focus loss, stale guidance, rejected-action effects and an undefined-receipt exception were repaired without changing the pure API, recovery normalization, authored deck or score economy. Fresh verification passed 32 client checks, eighteen Java checks, packaging and twelve distinct final desktop/phone browser cases; independent review found no blocker. Retains Limited beta ready, 4. The historical human interpretation and finite-content gap remains open. Local reliability does not establish novice discovery, difficulty balance, enjoyment, live-provider behavior or public production readiness.

Maturity gap: Twelve authored English boards constrain replay and the eight-day daily cycle repeats. Source-informed functional play does not prove novice discovery, enjoyment or difficulty balance. Progress is device-local; human beta should observe clue use, misunderstanding and willingness to play again. Purpose-based exclusions need continued human interpretation review, especially multifunctional devices and repurposed objects.

The current default player route uses endpoints_beta.jte and the endpoints mode in deduction-beta.js, with runtime identity game60 and inventory identity game-60. The retained older client and service are not the mapped authored implementation. The first complete local cycle played ep-09, document to language to meaning. Both middle meanings and endpoint-role clues were read. An incomplete pair was free, museum with translator cost 15 points, and repeating that pair cost nothing. Two 12-point exclusions left 61 points, disabling museum and poet. The archive selection and existing translator selection survived active reload. Archive with literary translator then solved for 61 points. Both supported pairs, all ten candidate reasons, four chronological receipts, canonical Copy and terminal recovery agreed. Ordinary peer Replay from focused Copy left focus on the document body and retained obsolete fixed-result feedback. A focused native museum choice was detached by a peer right-choice update, and peer hint adoption kept repeated-pair guidance. That shared continuation also finished at 61 with all explanations, receipts, Copy and recovery reviewed. The original focused client run passed 22 of 32 checks and failed ten targeted checks, with no skips. The original eight desktop/phone browser cases produced four passes and four targeted failures in 54.2 seconds. Both delayed Check journeys raised the real undefined-last-receipt TypeError after adopting a newer attempt. Complete outcomes preceded the final boundary verdicts; no harness error was established. The repair adds current-save preflight for free and refused actions, accepted-save and commit gates with distinct rejection outcomes, accepted Replay, scoped focus rescue and cleared obsolete guidance. Native endpoint choices, middle evidence and immutable result disclosures stay stable while the latest receipt updates separately. Deferred effects are guarded against stale state. The pure engine and API, existing recovery normalization, authored decks, storage identities and score economy remain unchanged. Independent source review found no material blocker, and the design check returned no findings. Fresh verification passed 32 client checks, comprising fifteen pure-engine and seventeen DOM checks, with no failures or skips. Eighteen fresh Java checks passed: three authored Endpoints checks, three retained legacy service checks, five shared authored-deck checks and seven inventory checks. The legacy service checks are not current-route provider evidence. Source, packaged and served runtime hashes agreed; the pure engine/API prefix was byte-identical and the package matched the then-current 24-row refreshed inventory. All eight repaired first-cycle desktop/phone browser cases passed in 42.9 seconds. Two actual repaired 61 point outcomes were completed and reviewed. Peer Replay now focused a visible museum choice; that same native choice remained connected and focused through a peer right selection, and peer hints removed obsolete repeated-pair guidance. The second complete cycle played ep-10, flowing water to crossing to route. An incomplete river choice was free. River with ferry and stream with ferry each cost 15 points, a repeated river/ferry pair was free, and two 12-point exclusions left 46 points. The stream/road-bridge draft survived active reload and solved for 46. Both middle meanings, both supported pairs, ten candidate reasons, five receipts, canonical Copy and terminal recovery agreed. Review established no new scoped defect, so the runtime remained unchanged. Fresh actual phone Replay checked pond/dam, pond/ferry and lake/dam, then recovered with one chance and 55 points available. Sea/dam exhausted the fourth distinct miss with an explicit no-earned-score result. All ten reasons, both supported pairs, four receipts, unsolved Copy and terminal recovery were reviewed. A fresh accepted Replay with river/road bridge then solved for 100 points, with both middle meanings, both pairs, ten reasons, its single receipt, Copy and recovery checked again. All four second-cycle desktop/phone cases passed in 17.1 seconds, giving twelve distinct final passing browser cases. All 32 final client checks passed. Runtime SHA-256 remained 5f836dd40ef6dd736fa7307293480e9d73da1af8693a67dc0b2c4d442fe67dd7 through the second cycle, with the pure engine unchanged. This is source-informed local agent and automated evidence; no human, live-provider or deployment validation is claimed. Earlier known-issue record, retained for history: Resolved and locally verified 11 October 2026: ordinary peer Replay rescues focus from hidden Copy to a visible endpoint choice, still-visible native choices retain focus through peer selections, and peer exclusions clear obsolete repeated-pair guidance. Current-save preflight protects incomplete and repeated deductions and other free or refused actions. Accepted-save and commit gates prevent false paid feedback and the reproduced undefined-last-receipt exception when adopting another attempt. Rejected Replay preserves the actual newer attempt and URL, with focus on its active choices or terminal result. Stable middle evidence and result disclosures preserve ongoing reading, while guarded deferred effects avoid stale focus changes. No additional defect or runtime change was justified by the second completed cycle.

• modules/games/src/main/java/io/kgard/web/slice/games/game60/EndpointsController.java
• modules/games/src/main/resources/games/endpoints-beta-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game60/EndpointsAuthoredDeckTest.java
• modules/games/docs/games-maturity-20261010-third-prototype-beta.json
• conf/tasks/evidence/third-ten-games-20261010/deduction
• conf/tasks/evidence/third-ten-games-20261010/final-verification.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/75-endpoints/verification.json
• conf/tasks/games-two-round-review-20261011/evidence/75-endpoints/round-1
• conf/tasks/games-two-round-review-20261011/evidence/75-endpoints/round-2
• modules/games/src/main/java/io/kgard/web/slice/games/game60/endpoints_beta.jte
• modules/games/src/main/resources/static/js/games/deduction-beta.js
• modules/games/src/test/js/deduction-beta.test.cjs
• modules/games/src/test/js/deduction-beta-ui.test.cjs
• modules/web-visual/playful-games-tests/endpoints.spec.ts
• modules/web-visual/playful-games-tests/endpoints-round2.spec.ts
• conf/tasks/games-quality-20261013/results/verification-batch-006.md

Compare English and German words in real sentence contexts. Classify matching senses and false friends, then inspect the meanings.

Deutsche und englische Wörter gehören zur Aufgabe; Bedienung und Erklärungen sind Englisch. Ohne Englischkenntnisse noch nicht vollständig spielbar.

English / German

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards: eight rotating UTC daily + four separate practice. Four bilingual sense comparisons per case.

Solo

Limited beta ready

Finite authored English/German sentence senses, not arbitrary translation. Dictionary links and live provider quality remain unverified; retained provider fallback is unmarked. Device-local saves are non-atomic, with no account sync. Human fairness, difficulty, enjoyment, retention and production readiness remain unverified.

The 13 October independent review accepted False Friends unchanged. All 48 sentence judgments and 18 bilingual pairs were reviewed; ordinary and corrected play, loss and reveal, saved results and replay passed. The finite bank and editorial contextual readings remain disclosed. Accepted for invited human playtesting; human feedback remains pending. This is local technical evidence, not deployment or public-release signoff.

Authored English/German sentences and dictionary-linked word-sense judgments.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

Den tatsächlichen Weg vom geteilten Ergebnis zum spielenden Empfänger prüfen; keine unbestätigten Wachstumswerte ableiten.

Collect unprimed human feedback on comprehension, fairness, difficulty and replay interest; technical readiness and human feedback remain separate in conf/tasks/games-quality-20261013/game-quality.json.

Latest recorded source-informed local play/review; carried forward, not a fresh human or production playtest.

Maturity gap: Human beta feedback is needed on fairness, unprimed onboarding, difficulty and replay interest. The authored starter bank is finite.

The mapped /games/game9 route uses FalseFriendsAuthoredDeck and false_friends_beta.jte with runtime identity game9 and friends-2 rules. Twelve English/German files comprise eight rotating UTC daily boards and four separate practices. Their forty-eight case occurrences reuse eighteen distinct sentence pairs and dictionary-linked reasons. Each file asks four same/different judgments; one point per correct answer gives genuine clearance at three or four and a loss at zero through two. Free revisions replace the draft; Reveal is unscored. Recovery still returns only answers, phase and Boolean replay, tolerates extra fields, complete playing drafts and partial Reveal, and preserves the original daily replay:true compatibility. There is no paid Hint, chronological event ledger, capacity repair or new scoring rule. Original native Plans and stories completed four on desktop and three on phone, with partial and complete-draft reload, a free wrong-to-correct Arm revision, all eight sentences, four literal reasons and actual submitted judgments. Dictionary href, target and rel attributes, actual canonical Copy, terminal recovery and bounded views were inspected. Ordinary active peer edits detached Roman Same focus and kept obsolete guidance; terminal adoption left BODY focused. In a separate observed peer Replay, Copy remained focused inside the hidden result. Original clients passed eight of seventeen with nine owning failures in 737.004792 milliseconds. Original desktop/phone browser journeys passed two and failed six in 31.7 seconds; full earned outcomes preceded the deferred focus/guidance verdicts, with eight complete custom captures retained. Two cases appended after the closed baseline failed against the unchanged original in 480.25575 milliseconds. Unreadable storage with working writes overwrote an unknown genuine two-point loss during completed memory four and Replay. The regression reviewed the actual recovered state and fresh durable three/four before its captured preservation verdict. A separate durable initial Reveal first Copy already passed; that case failed only on accepted Replay focus after the subsequent complete four and recovery. Exact-key startup, scoring, valid help/outside focus and pending Copy protection remain compatibility, not newly discovered defects. The owning repair preflights every judgment, File, Reveal, Copy and Replay and gates requested effects on actual shared state acceptance, independently of durable-write success. Existing readable-save protection now prevents blind writes during unknown read outages while retaining warned memory play. Observation retains attempt identity and malformed bytes, excludes savedAt, and aligns after initialization, reconciliation and accepted writes only when unchanged recovery agrees with current memory. Native case/answer controls remain mounted with their bilingual text and pressed/disabled states; adoption clears transient counts and rescues unusable owned focus without stealing valid help or outside focus. Focus is captured before shared adoption can hide Copy. Accepted-only Replay preserves a rejected URL/save, and shared asynchronous Copy fences remain intact. Independent full source and evidence review passed. Repaired first-round desktop four and phone three repeated all sentences, four reasons, actual canonical Copy, recovery and bounded inspection. The unchanged help had been read during the original baseline and was not reopened in the repaired native round. The peer retained the same Roman Same control, terminal adoption focused ff-result-title, and peer Replay returned from Copy to the first Same judgment with current guidance. All eight first-round browser cases passed in 32.8 seconds. The second native round completed Plans zero, threshold three and fresh four on phone; an early incomplete zero draft had Roman unanswered and File disabled, was inspected and deliberately completed before the zero was counted. No cause is inferred for that incomplete draft. Desktop Devices and decisions completed three after two-choice recovery, an all-same genuine two-point loss, then fresh reverse-order four. Every completed outcome received all eight sentences, four complete reasons and submitted judgments, actual score-specific Copy and terminal recovery; bounded desktop/phone views were readable. Second-round desktop dictionary link attributes were explicitly checked only on the final four-correct report. No second-round source change was justified. Four second-round browser cases passed in 33.9 seconds. The unchanged-source rounds therefore supply twelve distinct final cases and sixteen custom captures, excluding repeated originals, failure screenshots and attachment duplicates. All nineteen final clients passed in 676.787666 milliseconds: two pure checks and seventeen owning DOM cases. Eleven Java checks passed freshly before repair at 16:35:32.096–16:35:33.489 UTC: two mapped authored, three retained controller and six retained service checks. Their unchanged-source original passes are retained without a final rerun claim. Final runtime bb6190a5e63e15358c45156ab7c5676b0e801af7f1a5e6d7ffcc7ca3e6d20c34 matched source, packaged resources and the locally served script. The immutable isolated classpath contains 596 entries and 17,930,064 copied bytes with the sixty-eight-review inventory; it is not a fat archive or deployment. Of thirty-four protected inputs, only the adapter and two appended DOM cases changed. The original fifteen-case 16,886-byte DOM prefix, 787-byte pure/export prefix, 1,399-byte terminal render, deck, Java, templates, CSS, shared code and browser fixtures remain exact. The retained provider/controller method is unannotated and is not selected by the default authored route. Injected empty/throwing providers differ from the concrete provider, which can return fabricated fallbackRelatedTerms when Contextoid retrieval fails or is unusable; the retained service does not mark that provenance. Rank overlap is not translation evidence. Dictionary URLs were not visited or independently verified, and no live provider, account synchronization, general translation quality, human fairness/enjoyment, Linux run or production validation is established. localStorage remains non-atomic. Both ordered local rounds are complete; maturity 4, its 9 October date, basis and historical human-beta gap remain unchanged. Earlier known-issue record, retained for history: The 11 October review repaired unreadable-save overwrite and stale guidance/focus during peer updates and Replay. Both local cycles passed; maturity remains 4.

• modules/games/src/main/java/io/kgard/web/slice/games/game9/FalseFriendsBureauController.java
• modules/games/src/main/resources/games/false-friends-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game9/FalseFriendsBureauRoundServiceTest.java
• modules/games/src/test/java/io/kgard/web/slice/games/game9/FalseFriendsBureauControllerTest.java
• modules/games/docs/games-maturity-20261009-prototype-beta.json
• conf/tasks/evidence/prototype-beta-20261009/semantic-games
• conf/tasks/evidence/prototype-beta-20261009/final-verification.json
• conf/tasks/evidence/prototype-beta-20261009/final-surfaces/surface-check.json
• conf/tasks/games-two-round-review-20261011/evidence/31-false-friends-bureau/baseline-closed.json
• conf/tasks/games-two-round-review-20261011/evidence/31-false-friends-bureau/actual-round1.json
• conf/tasks/games-two-round-review-20261011/evidence/31-false-friends-bureau/actual-round2.json
• conf/tasks/games-two-round-review-20261011/evidence/31-false-friends-bureau/original-native-cloud-report.json
• conf/tasks/games-two-round-review-20261011/evidence/31-false-friends-bureau/repaired-round1-native-cloud-report.json
• conf/tasks/games-two-round-review-20261011/evidence/31-false-friends-bureau/original-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/31-false-friends-bureau/original-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/31-false-friends-bureau/extra-original-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/31-false-friends-bureau/java-original.json
• conf/tasks/games-two-round-review-20261011/evidence/31-false-friends-bureau/round1-improvement.json
• conf/tasks/games-two-round-review-20261011/evidence/31-false-friends-bureau/independent-path-proposal.json
• conf/tasks/games-two-round-review-20261011/evidence/31-false-friends-bureau/independent-path-review-final.json
• conf/tasks/games-two-round-review-20261011/evidence/31-false-friends-bureau/round1-clients.log
• conf/tasks/games-two-round-review-20261011/evidence/31-false-friends-bureau/round1-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/31-false-friends-bureau/round2-browser.log
• conf/tasks/games-two-round-review-20261011/evidence/31-false-friends-bureau/final-source-preservation.json
• conf/tasks/games-two-round-review-20261011/evidence/31-false-friends-bureau/classpath-round1-package.json
• conf/tasks/games-two-round-review-20261011/evidence/31-false-friends-bureau/round1-served-runtime.json
• conf/tasks/games-two-round-review-20261011/evidence/source-baseline-inventory-31-false-friends-bureau.json
• conf/tasks/games-two-round-review-20261011/queue.json
• modules/games/src/main/resources/static/js/games/false-friends-beta.js
• modules/games/src/test/js/false-friends-ui.test.cjs
• modules/web-visual/playful-games-tests/false-friends.spec.ts
• modules/web-visual/playful-games-tests/false-friends-round2.spec.ts
• modules/web-visual/playful-games-tests/false-friends-support.ts
• conf/tasks/games-quality-20261013/results/verification-batch-009.md

Compare explained routes across three English example pairs, or inspect up to four cached routes between your own words. Seven authored routes are available on matching cache misses. An unsaved explorer with no score or background search.

Vollständig deutsche, vorbereitete Beispiele ohne Live-Datenzugriff. Die ursprüngliche englische Ausgabe bleibt separat erreichbar. Entwürfe bleiben nur auf der geöffneten Seite, nicht nach dem Neuladen.

English / German

Alle 5 Paare und beide Wege sind sofort vergleichbar. Die Beispiele sind fest; es gibt keine automatische Suche nach weiteren Wegen.

Sieben erklärte Beispielwege über drei englische Endpunktpaare

Keine Tagesrotation.

Alle Beispiele sind direkt wählbar. Eigene Anfragen zeigen höchstens vier bereits zwischengespeicherte Wege; ohne Treffer entsteht kein neuer Inhalt.

Up to 4 cached routes per pair; 3 English authored example pairs with 7 routes and 23 connection readings. Original Sushi → Japan → JAXA → Spacecraft → Spaceship route remains first.

Solo

Concept demo

Finite unsaved route comparison without a score or timer; reload returns to the first connection. Authored routes are interpretations, not measured rankings or exhaustive answers. Cached coverage and metadata limit provider content; a snapshot is neither a full search nor proof of semantic quality. Other misses stay empty. Live-provider behavior, human usefulness and engagement remain unverified.

All seven authored routes and23 readings across three fallback pairs passed independent desktop390/320, no-JavaScript, copy and empty-custom checks. Cached paths still take priority. This remains an unsaved explorer. Technical readiness and human feedback are tracked separately in conf/tasks/games-quality-20261013/game-quality.json; human feedback remains pending.

Up to four matching, deduplicated cached Termpath routes from one Games Data API lookup. When no usable cached result exists, three English example pairs offer seven authored routes with an interpretation for each connection and no invented measurements.

Custom language/source/target supported. Unlisted endpoint pairs and non-English requests remain empty on cache miss. Cached results take priority over authored examples.

Spring/JTE + browser client; unsaved route selection/inspection; Games Data API cache dependency, one explicit default authored example and empty states for other cache misses.

2 Verbesserungen umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-20. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-fourth-ten-rounds-20261020/evidence/learning/ledger.json

Die Sortierung „Weitere Prüfrunden“ zählt alle nachfolgenden Prüfungen zusammen. Die ersten fünf Runden bleiben separat.

2 von höchstens 10 Runden am 2026-10-20: Eingabefehler und Vergleich sind erreichbar; eigene Deutungen bleiben unbewertet. Weitere Änderungen brauchen menschliche Rückmeldung.

• conf/tasks/games-nonviral-further-rounds-20261020/evidence/learning/ledger.json

4 frühere Prüfeinträge · 1 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json
• conf/tasks/games-nonviral-five-rounds-20261020/evidence/learning-next/ledger.json

Die deutsche Ausgabe mit deutschsprachigen Ersttestern vergleichen: Verständnis, strittige Antworten, genutzte Hilfen und freiwillige weitere Runde beobachten. Menschliche Rückmeldung steht noch aus.

Invite unprimed human playtesting and assess the recorded limitations; implement observed problems through bounded repairs with independent verification.

Reassessed 11 October 2026 after two completed local inspection, review, improvement and replay cycles. All four authored readings, forward, non-linear and reverse navigation, custom empty-result retry and return, public Copy and the deliberately unsaved reload were exercised. Endpoint focus, canonical normalized/encoded sharing and authored JAXA display were repaired without changing stored route identity or cached metadata. Nine client checks, sixteen focused Java checks and packaging passed in cycle one; six first-cycle and two second-cycle desktop/phone cases passed. Cycle two needed no additional source change. Retains Concept demo: this remains an unsaved explorer without a scored game objective, and live cached-provider behavior, unprimed human usefulness and engagement, and production readiness are unverified.

Maturity gap: Historical 5 October gap: simulated discovery exposed preloaded routes and generic fallback without a decision loop. Those misleading states are removed; the current edition is an explicit unsaved explorer and still has no scored game objective.

Gap at the original review: The player has no inferential, construction or scored decision loop after choosing endpoints. The polling and discovery times are presentation of paths already supplied, and arbitrary-pair fallback routes are generated practice.

13 October coordinated quality acceptance: All seven authored routes and23 readings across three fallback pairs passed independent desktop390/320, no-JavaScript, copy and empty-custom checks. Cached paths still take priority. This remains an unsaved explorer. Human feedback is pending; no production or live-provider validation is claimed. The historical maturity rating remains unchanged. Earlier finding: The 13 October pilot expands the single authored fallback into three selectable pairs and seven explained routes. Worker checks pass; independent content acceptance is recorded separately in the central quality tracker. Earlier endpoint-focus, canonical Copy and JAXA-label repairs remain intact. Current 13 October quality pilot: Up to four matching, deduplicated cached Termpath routes from one Games Data API lookup. When no usable cached result exists, three English example pairs offer seven authored routes with an interpretation for each connection and no invented measurements. Up to 4 cached routes per pair; 3 English authored example pairs with 7 routes and 23 connection readings. Original Sushi → Japan → JAXA → Spacecraft → Spaceship route remains first. Finite unsaved route comparison without a score or timer; reload returns to the first connection. Authored routes are interpretations, not measured rankings or exhaustive answers. Cached coverage and metadata limit provider content; a snapshot is neither a full search nor proof of semantic quality. Other misses stay empty. Live-provider behavior, human usefulness and engagement remain unverified. See conf/tasks/games-quality-20261013/game-quality.json for separate technical readiness and human feedback. The earlier review history below is retained and describes the previous default where applicable. Route identity and endpoint validation prevent unrelated cached paths from appearing; the date label does not imply newly generated content. The first source-informed local cycle inspected all four authored connections, Sushi → Japan → JAXA → Spacecraft → Spaceship, and read each explanation. The screen correctly identified one authored example with no measured ranks or weights, timer, score or saved attempt. Actual review found that final Next disabled the focused control and left focus on the page body, Copy included the local preview address, and the route displayed Jaxa while its explanation used JAXA. The repair moves focus to the selected connection only at a disabled navigation endpoint, builds the public URL from normalized lookup parameters with UTF-8 encoding, and changes only the authored acronym display. Java coverage verifies defaults, Unicode normalization and encoding, omission of reveal state, and preservation of stored nodes and cached display/measurement semantics. The repaired actual replay inspected all four steps in order: enabled Next retained focus, final Next focused connection four, and Copy retained labelled authored provenance with https://contextoid.com/games/game73?language=en&source=sushi&target=spaceship. Reload returned to connection one as stated. Nine client checks, sixteen focused Java checks (nine Free Roam and seven inventory checks), six desktop/phone browser cases and packaging passed in cycle one. The second cycle loaded Html → Mayonnaise and observed an honest empty result retaining both inputs, offering retry and fabricating neither routes nor inspection controls. Retry kept the same words. Try Sushi to Spaceship returned to the example, whose four explanations were inspected in order 4, 2, 3, 1 before Copy and another first-step reload. No additional source change was warranted. The actual phone replay inspected all four connections in reverse; enabled Previous retained focus, and the first-step endpoint focused Sushi → Japan. Copy succeeded with authored provenance, the route and explanation wrapped without horizontal overflow at 390px, and the Copy controls were visually inspected. Reload again returned to step one. Two additional desktop/phone cases passed, giving eight distinct browser cases across both cycles. The second cycle retained the first cycle's tested implementation. These local source-informed checks establish the repaired inspection behavior, not a scored gameplay loop, live provider availability, human engagement or production readiness.

• modules/games/src/main/java/io/kgard/web/slice/games/game73/FreeRoamController.java
• modules/games/src/test/java/io/kgard/web/slice/games/game73/FreeRoamRoundServiceTest.java
• modules/games/docs/games-maturity-20261005.json
• modules/games/src/main/java/io/kgard/web/slice/games/game73/FreeRoamRoundService.java
• modules/games/src/main/java/io/kgard/web/slice/games/game73/free_roam.jte
• modules/games/src/main/java/io/kgard/web/slice/games/TermpathGameSupport.java
• conf/tasks/evidence/games-maturity-20261005/termpath-b-game-73-phone.png
• conf/tasks/evidence/games-maturity-20261005/termpath-b-browser.json
• conf/tasks/games-autonomous-hardening-20261010.md
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/86-free-roam/round-1
• conf/tasks/games-two-round-review-20261011/evidence/86-free-roam/round-2
• modules/games/src/main/java/io/kgard/web/slice/games/game73/FreeRoamRound.java
• modules/games/src/main/java/io/kgard/web/slice/games/game73/FreeRoamPath.java
• modules/games/src/main/resources/static/js/games/free-roam.js
• modules/games/src/test/js/free-roam.test.cjs
• modules/web-visual/playful-games-tests/free-roam-integrity.spec.ts
• conf/tasks/games-quality-20261013/results/pilot-b-content.md
• conf/tasks/games-quality-20261013/results/verification-content-game73.md

Follow an authored word journey one connection at a time, then choose its concealed destination. Extra evidence costs points; three distinct wrong destinations end the attempt.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben und Rückmeldungen. Die englische Ausgabe bleibt separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle vier Aufgaben sind heute als Übung erreichbar. Tagesaufgaben und Übungen nutzen denselben Vorrat. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards (8 repeating daily + 4 practice)

Solo

Limited beta ready

A finite authored English candidate bank has one designated destination per board; alternative natural readings depend on editorial acceptance. Twelve boards provide eight repeating daily scenes and four separate practice scenes. The first connection is free, each additional Explore costs 12 points and each distinct wrong destination costs 15; three distinct wrong destinations end the attempt without an earned score. Empty or repeated checks are free, while Reveal is unearned. Buying all connections can make the final role definition direct, so difficulty depends on reasoning from earlier clues. Progress and selected drafts are browser-local without account or cross-device synchronization; changed-save checks are not atomic between simultaneous tabs, and unavailable storage permits warned memory-only play. The default mapped route has no live-data or custom-board mode. Retained cached-Termpath and fallback code is unmapped and was not validated by these authored cycles. Human beta feedback is still needed on clue use, misunderstanding, semantic acceptance, enjoyment, willingness to replay and difficulty balance. Public production readiness remains unverified.

The 13 October independent review accepted Hidden Destination unchanged. All12 clue/path sequences and exclusions were reviewed, ordinary and deliberate-mistake play completed, and score/replay/reload behavior passed. Real read-only outages preserve prior results while allowing warned memory play and deliberate durable recovery. Accepted for invited human playtesting; human feedback remains pending. This is local technical evidence, not deployment or public-release signoff.

Authored word chains, progressively disclosed connection readings and destination clues.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

Eine Verbesserung umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-22. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-ninth-ten-rounds-20261022/evidence/words/ledger.json

3 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json

Mit deutschsprachigen Ersttestern Verständnis, strittige Antworten, genutzte Hilfen und eine freiwillige weitere Runde beobachten. Mit anderen Varianten derselben Richtung vergleichen; menschliche Rückmeldung steht aus.

Collect unprimed human feedback on comprehension, fairness, difficulty and replay interest; technical readiness and human feedback remain separate in conf/tasks/games-quality-20261013/game-quality.json.

Reassessed 11 October 2026 after two completed local play, review, improvement and replay cycles. Actual outcomes included 61 before and after repair, 58, a genuine three-miss loss and fresh 100-point replays, with all path and exclusion readings, committed receipts, canonical Copy and active/terminal recovery. Controlled startup-observer and queued-focus regressions were repaired, and singular/plural display wording was corrected. Existing inherited browser acceptance already passed on the original. Final verification passed 44 client checks and twelve distinct desktop/phone browser journeys, with independent source review and bounded desktop/phone visual inspection. Nineteen fresh Java checks preceded the final one-line display correction; packaging and all twelve owning browser journeys were repeated afterward. Retains Limited beta ready, 4. The historical finite-content, novice-discovery, clue-use, enjoyment and difficulty questions remain open; local correctness does not establish human semantic quality, live/custom-provider behavior or public production readiness.

Maturity gap: Twelve authored English boards constrain replay and the eight-day daily cycle repeats. Source-informed functional play does not prove novice discovery, enjoyment or difficulty balance. Progress is device-local; human beta should observe clue use, misunderstanding and willingness to play again. Some final definitions are direct once all connections are bought; difficulty depends on early-clue reasoning.

The mapped player uses HiddenDestinationController.playable, HiddenDestinationAuthoredDeck, hidden_destination_beta.jte, hidden-destination-beta-deck.json and the shared deduction-beta.js runtime. Inventory ID game-37 and persisted runtime ID game37 remain distinct. The authored-v1 envelope and deduction-3 rules identity are unchanged. Recovery normalizes six logical fields: actions, draft, left, right, phase and replay; tolerated unrelated top-level metadata is not a new schema. The paid event chronology, scoring and pure/exported engine prefix remain byte-identical. The retained hiddenDestination controller method, HiddenDestinationRoundService.generateRound and hidden_destination.jte use cached Termpath or fallback material but are unmapped; language, source and target query parameters do not select that legacy method. No external provider or custom-board quality was exercised. The first actual desktop cycle used hd-09, The patient trail. A blank Check spent nothing. Wax was a distinct miss, and repeating wax was free. Two paid Explores revealed the seed to plant to flower to nectar journey and sharpened the destination role. The selected honey draft survived active reload; explicit Check earned 61/100 from one miss and two extra connections. Root inspected all three path readings, the final explanation, all seven excluded-destination reasons and four chronological events, then copied the canonical spoiler-free result and recovered it after reload. A bounded original desktop result was visually inspected. The pristine original client run passed 41 of 43 checks with two targeted failures and zero skips. One exact-key startup interleaving could leave the owning observer behind the newer state already recovered by shared initialization, repeatedly refusing valid actions. The other failure combined queued final Explore focus, native self-disable blur and terminal peer adoption, leaving focus on the page body. All eight original anchored desktop/phone browser journeys passed in 32.0 seconds. Their full 61 and 100 outcomes, native choices, peer recovery, delayed accepted-state boundaries and memory/manual Copy behavior were inherited compatibility, not additional original defects. The minimal repair observes the matching saved state immediately after shared startup and tracks pending in-game focus intent through adoption. Pending intent is consumed or invalidated so a native-blurred page body can hand off to the actual active or terminal state, while deliberate outside focus is preserved. An appended complete 88-to-fresh-100 outside-focus guard passed on the original and repaired runtimes; the updated original comparison was 42 passes and two failures out of 44, separate from the pristine 43-check baseline. Existing status and result counts were also given singular/plural wording. No template, authored deck, paid chronology, rules, storage identity or normalized field contract changed. The repaired build retained the original saved 61. Root explicitly started a fresh desktop Replay, repeated wax and its free duplicate, bought two connections, recovered the selected honey draft and checked it for 61. All readings, exclusions and four committed events were reviewed, copied and recovered again, with a bounded final desktop inspection. The first repaired client run passed all 44 checks. Nineteen fresh Java checks passed: thirteen owning and inventory checks plus six existing Endpoints compatibility checks. Twenty anchored browser journeys passed in about 1.5 minutes: eight owning first-round cases and twelve completed Endpoints compatibility cases. These compatibility cases do not add another completed edition. Browser expected wording was updated after the original run without removing gameplay, score, persistence or focus assertions. Independent source review passed on the frozen candidate. The second actual cycle used hd-10 on a phone. Theory was a distinct miss and its repeat was free. The final paid Explore exposed observation to evidence, sharpened the testable-proposal role and handed focus to the visible role clue. Experiment was the second distinct miss. The selected hypothesis draft survived active reload with one miss remaining and 58 points available; explicit Check earned 58 on the third deduction. Both path readings, the final explanation, seven excluded-destination reasons and four committed events were reviewed, copied and recovered. Root also completed and reviewed a genuine three-distinct-miss loss with no earned score, then a clean 100-point Replay. Final phone inspection found the remaining generic connection(s) text in committed destination receipts. The second-cycle refinement changed one destination description expression to the existing singular/plural count helper and updated the exact owning browser expectations. It made no rule, state, deck or template change. Independent review reconstructed the prior runtime hash by reverting only that expression, confirming the bounded display-only delta. After packaging, the retained saved 100 displayed one connection explored. A fresh phone Replay selected and checked hypothesis for 100, then reviewed both path readings, the final explanation, all seven exclusions and its committed receipt, copied the canonical result and recovered it after reload. The bounded phone visual review confirmed the corrected receipt. Final verification passed all 44 client checks with zero skips in 703.5 milliseconds and all twelve distinct owning desktop/phone browser journeys across both rounds with zero skips in 48.7 seconds. The nineteen fresh Java checks and twelve completed Endpoints browser compatibility cases preceded the final destination-only receipt correction; final packaging and the twelve owning journeys were repeated afterward. Final runtime SHA-256 is 6429924b88a066a007a7fbc3b9563ca7efd4ceff4da55a80c40d8550a3a76f36; the dedicated DOM suite remained dc7c6d5d81b346047617505c09305b768e5d2118c04fa4e0193cb011feb24056. This evidence is local agent gameplay and automated/source review, not human enjoyment, difficulty fairness, external-provider validation, deployment or public production readiness. Earlier known-issue record, retained for history: Resolved and locally verified 11 October 2026: startup observation now aligns with the state actually recovered by the shared session, preventing repeated refusal after the controlled initial-read race. Pending in-game focus intent survives peer adoption when a self-disabling final Explore has already blurred focus to the page body; deliberate outside focus remains undisturbed. Status, result and committed-receipt counts use singular/plural wording. The second cycle changed only the remaining connection-count receipt text. All eight original browser journeys already passed, so inherited native controls, accepted-action protection, Replay and Copy behavior are compatibility evidence rather than newly discovered defects. Final replays established no further defect. Local storage reconciliation remains non-atomic.

• modules/games/src/main/java/io/kgard/web/slice/games/game37/HiddenDestinationController.java
• modules/games/src/main/resources/games/hidden-destination-beta-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game37/HiddenDestinationAuthoredDeckTest.java
• modules/games/docs/games-maturity-20261010-third-prototype-beta.json
• conf/tasks/evidence/third-ten-games-20261010/deduction
• conf/tasks/evidence/third-ten-games-20261010/final-verification.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/63-hidden-destination/verification.json
• conf/tasks/games-two-round-review-20261011/evidence/63-hidden-destination/round-1-before
• conf/tasks/games-two-round-review-20261011/evidence/63-hidden-destination/round-1-after
• conf/tasks/games-two-round-review-20261011/evidence/63-hidden-destination/round-2-final
• modules/games/src/main/java/io/kgard/web/slice/games/game37/hidden_destination_beta.jte
• modules/games/src/main/resources/static/js/games/deduction-beta.js
• modules/games/src/test/js/deduction-beta.test.cjs
• modules/games/src/test/js/deduction-beta-ui.test.cjs
• modules/games/src/test/js/hidden-destination-ui.test.cjs
• modules/web-visual/playful-games-tests/hidden-destination.spec.ts
• modules/web-visual/playful-games-tests/hidden-destination-round2.spec.ts
• conf/tasks/games-quality-20261013/results/verification-batch-006.md

Count the fewest open arrows on a declared word map, including closed links, zero-arrow journeys and unreachable destinations.

Deutsche Bedienung, Aufgaben, Rückmeldungen und Ergebnislinks im Starterpaket. Die ursprüngliche englische Ausgabe bleibt unter ihrer bisherigen Adresse erhalten; ihre Aufgaben und Spielstände sind getrennt.

English

Alle 6 Aufgaben sind heute als Übung erreichbar. Davon 2 zusätzliche Übungen außerhalb der Tagesrotation. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards (8 repeating daily + 4 practice)

Solo

Limited beta ready

Distances count only the finite authored open, one-way arrows; they are not measured corpus semantic distances, Contextoid ranks or live graph results. Twelve authored English boards provide eight repeating UTC daily boards and four separate practice boards. The complete four-answer plan submits once, with 25 points per exact decision; a lower score is a completed result, while Reveal earns no score. Map-reading workload, zero-arrow and unreachable-route understanding, and enjoyment of repeated arithmetic remain untested with novices. Progress is browser-local without account or cross-device synchronization. Changed-save protection is a non-atomic preflight rather than a simultaneous-write transaction. Retained legacy service checks do not establish current-route provider behavior. Human difficulty, enjoyment, repeat interest, live-provider behavior and public production readiness remain unverified.

The 13 October normal-motion follow-up independently accepted authoritative Submit, Reveal, Replay and peer recovery visibility after the focused heading had remained offscreen. Instant centered reveal preserves intended focus and exact saved results. All 74 client checks, 54 browser cases and 18 independent follow-up journeys passed; prior full content/gameplay evidence remains applicable. Accepted for invited human playtesting; human feedback remains pending. This is local technical evidence, not deployment or public-release signoff.

Authored directed maps with open/closed arrows and shortest-hop queries.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

3 Verbesserungen umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-20. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-fourth-ten-rounds-20261020/evidence/learning/ledger.json

4 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json
• conf/tasks/games-nonviral-five-rounds-20261020/evidence/strategy-next/ledger.json

Die deutsche Ausgabe mit deutschsprachigen Ersttestern vergleichen: Verständnis, strittige Antworten, genutzte Hilfen und freiwillige weitere Runde beobachten. Menschliche Rückmeldung steht noch aus.

Collect unprimed human feedback on comprehension, fairness, difficulty and replay interest; technical readiness and human feedback remain separate in conf/tasks/games-quality-20261013/game-quality.json.

Reassessed 11 October 2026 after two completed local play, review, improvement and replay cycles. Full Market 100, Studio 50 and corrected Studio 100 outcomes covered closed and directed arrows, zero-arrow identity, unreachable destinations, every map node and arrow, all four shortest-route proofs, Copy and recovery. The incomplete 250-edit trap, stale-action effects, native focus, obsolete guidance and catalog destination were repaired without changing the pure engine/API, strict recovered state or score economy. Fresh verification passed 71 client checks, twenty-one Java checks, packaging and twelve distinct final desktop/phone browser cases; independent review found no blocker. Retains Limited beta ready, 4. The historical novice understanding, map-reading workload and repeated-arithmetic enjoyment gap remains open. Local reliability does not establish human enjoyment, live-provider behavior or public production readiness.

Maturity gap: Test zero-arrow and unreachable understanding, map-reading workload and repeated arithmetic enjoyment with novices. Eight repeating UTC daily boards and four separate practice boards, authored English content and browser-local progress.

The mapped default route uses hops_beta.jte, the authored Hopreckon deck and the Hopreckon mode in prediction-fifth-beta.js, with runtime identity game59 and inventory identity game-59. The retained older Hopreckon client and cached-path service are not the current default player implementation. The first complete local cycle played hops-09, Market map. All nine nodes and eleven directed arrows were read, including the closed Apple-to-Nest arrow. An incomplete Submit was free. Answers 6 and 3 survived active reload; No route and 0 completed the plan for 100 points. All four committed answers, best routes and breadth-first proofs, canonical Copy and terminal recovery agreed. The owning All games link incorrectly led to daily games. Source-informed regression work then reproduced an incomplete plan trapped after 250 free edits, ordinary native-focus and guidance failures, and delayed free-map or rejected-action focus effects. The original focused client run passed 59 of 71 checks and failed twelve targeted checks, with no skips. All eight original desktop/phone browser cases ended in targeted failures after completing their full outcomes. Each trapped edit-limit plan was honestly revealed for no earned score and followed by a fresh complete 100 point attempt before its final verdict. No runtime exception or harness error was established. The scoped runtime repair normalizes only valid free Choose or Clear actions at the 250-edit bound, preserving the current answer vector as legal events. The pure engine/API, strict six-field state, existing saved histories and 251-event terminal recovery, authored maps and scoring remain unchanged. Main-save preflight and accepted-save or Replay gates protect free views, refusals and delayed actions. Stable native case, navigation, draft, map and audit nodes retain valid focus and disclosures, with obsolete guidance cleared and unavailable focus rescued to actual active or terminal content. The owning catalog destination and singular one-arrow audit wording were corrected. An initial template build failed because GAMES_CATALOG did not exist; the owning template was corrected to the established literal /games/catalog before the final fresh build. Independent source review found no material blocker and the design check returned no findings. Fresh verification passed 71 client checks, comprising forty pure-engine and thirty-one DOM checks, and twenty-one Java checks: six authored Hopreckon checks, three retained legacy service checks, five shared authored-deck checks and seven inventory checks. The retained service checks do not establish live-provider behavior. Source, packaged and served runtime hashes agreed, the pure engine/API prefix was byte-identical, and the package matched the then-current 25-row refreshed inventory. All eight repaired first-cycle desktop/phone browser cases passed, with the runner reporting 1.2 minutes. Actual repaired Market play again reviewed all nine nodes and eleven arrows, restored the 6/3 partial plan, then completed No route/0 for 100 points. All four best routes and breadth-first proofs, canonical Replay Copy, terminal recovery and the catalog destination agreed. Fresh desktop result and phone draft/map captures were visually reviewed for readable controls, correct hidden panels and no overflow. The second complete cycle played hops-10, Studio map, on the phone. All nine nodes and eleven open arrows were read. A complete 5/No route/1/4 draft survived active reload and submitted for 50 points. The wrong visited-word count and longer route were explicitly marked for revision; the unreachable and one-arrow answers matched. Every committed answer, best route and breadth-first proof, the singular one-arrow wording, canonical Copy and terminal recovery were reviewed. No new scoped defect was established. Fresh actual Studio Replay re-read the full map, restored the corrected 4/No route/1/3 draft and completed 100 points, with all four audits, proofs, Copy and recovery checked again. All four second-cycle desktop/phone cases passed in 33.1 seconds, including a complete warned memory-only result and selected manual Copy. The eight first-cycle and four second-cycle cases provide twelve distinct final passing browser cases. All 71 final client checks passed. Runtime SHA-256 remained cf5e8279119b65401b5c05ec33eec4021ad111d9d279299830d2249861794933 through the second cycle, with the template and pure engine/API frozen. These are source-informed local agent and automated checks; no human, live-provider or deployment validation is claimed. Earlier known-issue record, retained for history: Resolved and locally verified 11 October 2026: valid free Choose or Clear actions at the 250-edit boundary normalize only the current answer vector into legal choice events, allowing unanswered cases to be completed. Existing saved histories, including complete 251-event terminal attempts, retain their recovery contract. Current-save preflight and accepted-action gates protect free navigation, map reading, refusals and Replay from effects based on stale attempts. Stable native cases, navigation, map rows and proof disclosures preserve usable focus and reading; peer changes clear obsolete guidance and rescue unavailable focus. Rejected active and terminal actions focus the actual current state. The owning All games link now opens the catalog and the one-arrow audit uses singular wording. No additional defect or runtime change was justified by the second completed cycle. Earlier known-issue record, retained for history: The 13 October independent review accepted Hopreckon after map Return was repaired for both normal play and newer peer progress. All 48 cases were independently audited for shortest open paths, closed connections, zero hops and unreachable destinations. Real saved 75-point progress survived a read-only storage outage, memory play and deliberate durable recovery. Accepted for invited human playtesting; human feedback remains pending. This is local technical evidence, not deployment or public-release signoff. Earlier known-issue record, retained for history: The earlier map Return repair remains integrated, but readiness was reopened after a new actual-viewport check found Reveal leaves its focused result hundreds of pixels above the viewport when the planning panel collapses. A bounded shared focus follow-up is assigned; previous shortest-path and recovery evidence is retained. Technical readiness is pending the follow-up; human feedback remains pending.

• modules/games/src/main/java/io/kgard/web/slice/games/game59/HopsController.java
• modules/games/src/main/resources/games/hops-beta.json
• modules/games/src/test/java/io/kgard/web/slice/games/game59/HopsAuthoredDeckTest.java
• modules/games/docs/games-maturity-20261010-fifth-prototype-beta.json
• conf/tasks/evidence/fifth-ten-games-20261010/prediction
• conf/tasks/evidence/fifth-ten-games-20261010/final-verification.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/74-hops/verification.json
• conf/tasks/games-two-round-review-20261011/evidence/74-hops/round-1
• conf/tasks/games-two-round-review-20261011/evidence/74-hops/round-2
• modules/games/src/main/java/io/kgard/web/slice/games/game59/hops_beta.jte
• modules/games/src/main/resources/static/js/games/prediction-fifth-beta.js
• modules/games/src/test/js/prediction-fifth-beta.test.cjs
• modules/games/src/test/js/prediction-fifth-beta-ui.test.cjs
• modules/web-visual/playful-games-tests/hops.spec.ts
• modules/web-visual/playful-games-tests/hops-round2.spec.ts
• conf/tasks/games-quality-20261013/results/verification-batch-008.md
• conf/tasks/games-quality-20261013/results/prediction-ending-assessment.json
• conf/tasks/games-quality-20261013/results/verification-batch-008-visibility-followup.md

Build a directed route that visits the fewest board-local hubs, then minimizes its declared fare. Retrace to compare branches before banking a fixed result.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben, Rückmeldungen und Ergebnislinks. Die englische Ausgabe und ihre Aufgabenkennungen bleiben separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle 4 Aufgaben sind heute als Übung erreichbar. Tagesaufgaben und Übungen nutzen denselben Vorrat. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards (8 repeating daily + 4 practice)

Solo

Limited beta ready

Hub status and fares are board-local authored rules, not live global ranks or measured semantic distances. Twelve finite English boards provide eight repeating daily scenes and four separate practice scenes. A non-endpoint with at least three incident arrows is a hub; incoming and outgoing arrows both count, and endpoints are exempt. Scoring minimizes hub visits before fare: the best objective earns 100, minimum hubs with a higher fare earns 70, and extra hubs earn 40; the optional Hint subtracts 10. Reaching the destination still requires explicit Bank. Free Retrace refunds removed legs and fare; there is no supplies limit or genuine exhaustion/loss state, and Reveal is an unearned zero. The result shows the final route and every complete alternative, not a chronological list of free erased detours. Progress is browser-local without account or cross-device synchronization; changed-save protection is not an atomic transaction between simultaneous tabs, and unavailable storage permits warned memory-only play. The mapped route has no live-data or custom-board mode. Retained cached-Termpath/provider-oriented service and template code are unmapped, and these authored cycles do not validate that legacy path. Whether unprimed players understand that fewer hubs outranks cheaper fare, enjoy freely comparing routes, learn from the supplied readings or find the difficulty fair remains unverified by human beta feedback. Public production readiness also remains unverified.

Resolved and locally verified 11 October 2026: changed-save preflight and accepted-action gates prevent stale guidance, requested-state focus and rejected Replay effects; valid native choice focus is retained and hidden Copy or disabled Hint receives a visible destination. A reachable 2,000-event completion trap now compacts only erased free detours while retaining surviving moves and the paid Hint in their original order. Legacy 2,000-event saves still recover unchanged. The owning catalog link and singular hub wording are corrected. The first repair candidate briefly reused an old directed-edge explanation when a peer changed the source station; a complete 70-point regression exposed it and current choices now refresh their actual edge text. This candidate regression was not an original-game defect. The original calibrated startup check passed; startup alignment is compatibility coverage. No further source change was justified by the second cycle, and final replays established no further defect. Browser storage preflight remains non-atomic. Independently accepted for invited human playtesting on 13 October 2026. Terminal restoration and real keyboard Tab-to-Replay remain visible;24 native probes/96 samples and ten continuation/peer observations pass across the two hub games. Content and save contracts remain unchanged. Human feedback and deployment remain pending.

Authored directed word graph; board-local hub labels and declared fares.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

Eine Verbesserung umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-20. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-fourth-ten-rounds-20261020/evidence/strategy/ledger.json

6 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json
• conf/tasks/games-nonviral-five-rounds-20261020/evidence/routes/ledger.json

Die deutsche Ausgabe mit deutschsprachigen Ersttestern vergleichen: Verständnis, strittige Antworten, genutzte Hilfen und freiwillige weitere Runde beobachten. Menschliche Rückmeldung steht noch aus.

Collect unprimed human feedback on graph planning, fairness, difficulty and replay interest.

Reassessed 11 October 2026 after two completed local play, review, improvement and replay cycles. Actual Language routes earned 100 before and after repair; Astronomy routes earned a hinted 30 and a fresh hinted best 90. All directed readings, board-local degree classifications, complete alternatives, final routes, canonical Copy and active/terminal recovery were reviewed. Current-save acceptance, scoped focus, the reachable history-capacity trap, the catalog link and singular wording were repaired. An edge-reading cache regression introduced by the first candidate was separately reproduced and fixed; the original startup check already passed. Final verification passed 52 client checks, thirteen fresh Java checks and twelve distinct desktop/phone browser journeys, with independent source review and bounded desktop/phone visual inspection. Retains Limited beta ready, 4. The historical questions about unprimed understanding of hubs before fare and the enjoyment of freely comparing routes remain open. Local correctness does not establish human semantic quality, fairness, enjoyment, live/custom-provider behavior or public production readiness. Independently accepted for invited human playtesting on 13 October 2026. Terminal restoration and real keyboard Tab-to-Replay remain visible;24 native probes/96 samples and ten continuation/peer observations pass across the two hub games. Content and save contracts remain unchanged. Human feedback and deployment remain pending.

Maturity gap: Do new players notice that fewer hubs outranks a cheaper fare? Does freely comparing routes feel engaging rather than mechanical once the scoring order is understood? Twelve authored boards and local progress bound replay; interpretation and difficulty need human beta feedback.

The mapped Hub Avoider player uses HubAvoiderController, HubAvoiderAuthoredDeck, hub_avoider_beta.jte, hub-avoider-beta-deck.json and hub-network-beta.js. Inventory ID game-39 and persisted runtime ID game39 are deliberately distinct and unchanged. Shared authored-v1 storage uses rulesVersion hub-1 and strict five-field state: boardId, events, phase, replay and version. Events reconstruct a simple directed route and the paid Hint. The mapped player uses twelve finite authored English boards. The retained hubAvoider method, HubAvoiderRoundService and hub_avoider.jte are unmapped legacy cached-Termpath/fallback code; language, source or target parameters do not select that method. No external provider was exercised. The first actual cycle used the Language to Computing practice scene. Root read all twelve directed explanations, eight non-endpoint degree classifications and the scoring rule. Language to Code to Poetry reached a genuine dead end. Two free Retrace actions returned to Language and refunded the surviving fare to zero. Translation to Compiler recovered active at zero hubs and fare 6. Moving to Computing reached fare 8 but did not bank automatically; explicit Bank earned 100/100. All three alternatives, the final route, every reading and degree classification were inspected. Canonical Copy and terminal reload retained the result, and the bounded original desktop result was visually inspected. The pristine original client run passed 40 of 51 checks with eleven targeted failures and no skips. It included twelve owning engine cases, fifteen owning DOM cases and twenty-four existing generated compatibility cases for other network editions; those compatibility checks carry no future-gameplay completion credit. All eight anchored original desktop/phone browser journeys failed at their targeted final verdicts after complete outcomes. The capacity journey first earned a normal 100, then reached 2,000 real events and demonstrated that a valid Bank was refused; that trapped attempt was not credited as a completed hinted result. Other journeys completed 100, 90, 70 and 40 outcomes before the catalog, focus or stale-action verdicts. The exact-key calibrated original startup case passed, so no original startup stranding is claimed. The original catalog destination and singular hub wording were visibly incorrect. The repair adds saved-at-independent observation and current-save preflight, gates local feedback and focus on actual accepted state, preserves useful native choices and register disclosures, rescues focus after hidden Copy and directs an accepted Hint to its shown text. Rejected Replay retains newer saved state, its URL and current result. Memory-only accepted actions remain distinct from failed persistence. At the history bound, only erased free detours are compacted; surviving move indices and the paid Hint retain their relative order. Old valid 2,000-event active and terminal saves recover exactly as before, oversized records remain invalid, and normal histories are not rewritten on load. The public engine API, five-field state schema, scoring economy, board bytes and content versions remain unchanged. The owning All games link now targets the catalog; result, share, hint and solution wording use the singular hub form where appropriate. During independent review of the first candidate, root found a cache keyed only by destination that could retain the Orbit to Shadow explanation after a peer adopted Sundial to Shadow. An appended full 70-point case passed on the original runtime and failed on the first candidate, which passed 51 of 52 checks. Refreshing both label/fare text and the actual directed-edge reading on every render fixed the regression while keeping native controls. The original-runtime comparison with the updated template passed 42 of 52 checks; it is not a replacement for the pristine 40-of-51 baseline. Browser wording expectations were changed only after the frozen original run to reflect the corrected singular form; no gameplay, score or acceptance requirement was removed. The repaired actual replay retained the original saved 100 through the new build, then explicitly started a fresh desktop attempt. Code to Poetry again reached a dead end, free Retrace refunded the route, and Translation to Compiler survived active reload before Computing and explicit Bank earned 100. All twelve readings, eight degrees, three solutions and the final route were inspected, copied and recovered. A bounded final desktop review found the result readable. All 52 final client checks passed with zero skips: twelve owning engine checks, sixteen owning DOM checks and twenty-four unchanged generated compatibility checks. Thirteen fresh Java checks passed. All eight first-round desktop/phone browser journeys passed in 21.9 seconds, including the reached history boundary, honest score tiers, hinted result, peer adoption and rejected effects. Independent final source reviews passed. The second actual cycle used Astronomy to Timekeeping on a phone. Root read all thirteen arrows and eight non-endpoint degrees, then followed Observation to Telescope to Lens into a dead end. Three free Retrace actions returned to Astronomy and refunded fare to zero. A paid Hint transferred focus to its visible clue. Orbit to Shadow recovered active at two hubs and fare 6 with the Hint retained. Timekeeping and explicit Bank earned 30/100 at fare 9: the extra-hub base score of 40 less the ten-point Hint. All thirteen readings, eight degrees, four alternatives and the final route were inspected; canonical Copy and terminal reload retained the outcome. Review found no additional defect and justified no second-cycle source change. A fresh practice Replay bought the same Hint and chose Orbit to Period, recovered active, then banked Timekeeping for 90/100 at one hub and fare 4. All readings, degree classifications, alternatives and final route were inspected again; singular-hub canonical Copy and terminal recovery agreed. The bounded phone result was visually inspected. Four second-cycle desktop/phone journeys passed in 14.1 seconds, including the lower 30 and best hinted 90, honest unearned Reveal and warned memory-only play with manual Copy. This gives twelve distinct final browser journeys across both rounds. Final runtime SHA-256 is 97b6d929b8ceb94b7dfcb5fc80e6746fc53827cdbc51b72d4068a4e722f2c3f1; final template SHA-256 is 4d8d012a884d5225b133b3b9613f61f0fa5a23e5faeb26e1525025c44a8fb396. The authored deck remained byte-identical at 5f32dbf3e74eec334fe5dd8f8344cdf690b91c3d971b639be14edbb8cc5e5cda. Final source remained frozen during the second cycle. This is source-informed local agent and automated evidence, not human enjoyment, fairness, provider/custom-content quality, deployment or public production validation.

• modules/games/src/main/java/io/kgard/web/slice/games/game39/HubAvoiderController.java
• modules/games/src/main/resources/games/hub-avoider-beta-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game39/HubAvoiderAuthoredDeckTest.java
• modules/games/docs/games-maturity-20261010-third-prototype-beta.json
• conf/tasks/evidence/third-ten-games-20261010/network
• conf/tasks/evidence/third-ten-games-20261010/final-verification.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/64-hub-avoider/verification.json
• conf/tasks/games-two-round-review-20261011/evidence/64-hub-avoider/round-1-before
• conf/tasks/games-two-round-review-20261011/evidence/64-hub-avoider/round-1-after
• conf/tasks/games-two-round-review-20261011/evidence/64-hub-avoider/round-2
• modules/games/src/main/java/io/kgard/web/slice/games/game39/hub_avoider_beta.jte
• modules/games/src/main/resources/static/js/games/hub-network-beta.js
• modules/games/src/test/js/network-third-beta.test.cjs
• modules/games/src/test/js/hub-avoider-ui.test.cjs
• modules/web-visual/playful-games-tests/hub-avoider.spec.ts
• modules/web-visual/playful-games-tests/hub-avoider-round2.spec.ts
• conf/tasks/games-quality-20261013/results/verification-batch-020-replay-followup.md
• conf/tasks/games-quality-20261013/results/verification-batch-020-terminal-focus.md

Preview and undo shared-word bans, then lock a surviving route that meets an authored game-rarity target.

Deutsche Bedienung, Regeln, vorbereitete Aufgaben und Rückmeldungen. Die englische Ausgabe bleibt separat erhalten. Menschliche Sprach- und Spieltests stehen noch aus.

English

Alle vier Aufgaben sind heute als Übung erreichbar. Tagesaufgaben und Übungen nutzen denselben Vorrat. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards (8 repeating daily + 4 practice)

Solo

Limited beta ready

Twelve finite authored English challenges cover six endpoint maps with two- and three-ban variants. Only supplied paths survive bans, in fixed file order. Rarity is the rounded average of assigned intermediate game values, not measured word rarity, semantic distance or an earned point score. Exact collateral previews, free full inspection and unlimited Undo or Reset constrain difficulty and depth. Progress is browser-local without account or cross-device synchronization; changed-save protection is a non-atomic preflight rather than a simultaneous-write transaction. The retained optional cached-Termpath mode was not validated by these authored-route checks. Human discoverability, difficulty, enjoyment and repeat interest, live-provider behavior and public production readiness remain unverified.

Resolved and locally verified 11 October 2026: ordinary peer Replay moves focus from hidden Copy to the visible route heading, while peer selections preserve the same still-visible native route button. Main-save preflight and acceptance gates protect free or refused choices, view switches, nested comparison Preview, Replay and Copy. Rejected Preview adopts the actual route without forcing Ban plan; a free phone view switch reconciles a newer terminal result. Rejected Replay retains the newer completed attempt and URL and focuses its result heading. Deferred effects check the current state and view, and invalid or disabled controls receive a visible focus handoff. No additional defect or runtime change was justified by the second completed cycle. Independently accepted unchanged for invited human playtesting on 13 October 2026: unfamiliar daily/practice wins, all 12 boards, 37 fresh client checks, 36 native geometry frames and genuine read-only-save recovery. A 2.83px immediate reload edge settles by 700ms. Human feedback and live-provider/deployment validation remain pending.

Authored alternate paths, shared words and assigned game-rarity credits.

Supplying language selects the retained Termpath experiment: Games Data API cached paths with rank/weight/node dimensions, falling back to practice data.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

Eine Verbesserung umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-20. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-sixth-ten-rounds-20261020/evidence/strategy/ledger.json

5 frühere Prüfeinträge · 0 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json

Mit deutschsprachigen Ersttestern Verständnis, strittige Antworten, genutzte Hilfen und eine freiwillige weitere Runde beobachten. Mit anderen Varianten derselben Richtung vergleichen; menschliche Rückmeldung steht aus.

Collect unprimed human feedback on route planning, finite-bank depth, difficulty and repeat interest; optional live cached-provider mode remains separately unverified.

Reassessed 11 October 2026 after two completed local play, review, improvement and replay cycles. First-survivor wins at 90/80 game rarity with one, two and three bans, all-removed Undo recovery, budget-exhausted hypothetical previews and a 20/80 below-target loss were checked with complete route readings, formulas, Copy and active/terminal recovery. Reproduced focus loss and unaccepted view effects were repaired compatibly. Fresh verification passed 37 client checks, sixteen Java checks, packaging and twelve distinct final desktop/phone browser cases; independent review found no blocker. Retains Limited beta ready, 4. The historical six-map difficulty and depth gap remains open: exact previews and unlimited Undo are still present, and assigned values are not measured rarity. Local reliability does not establish human enjoyment, learning, live-provider behavior or public production readiness. Independently accepted unchanged for invited human playtesting on 13 October 2026: unfamiliar daily/practice wins, all 12 boards, 37 fresh client checks, 36 native geometry frames and genuine read-only-save recovery. A 2.83px immediate reload edge settles by 700ms. Human feedback and live-provider/deployment validation remain pending.

Maturity gap: Six endpoint maps with two variants each; exact previews and unlimited undo limit difficulty and depth. Values are assigned game weights rather than measured rarity.

The first complete local cycle played hb-09, Book to Forest. Original Paper, Story and Page bans progressed through p1, p3, p4 and p6, with game rarity 20, 55, 42 and 90. Two-ban active recovery retained the intended plan. Lock finished p6 at 90/80 game rarity with three bans; all six route statuses and twenty link explanations, rounded-average formulas and collateral previews, canonical Copy and terminal recovery were reviewed. Actual ordinary peer Replay from focused Copy left focus on the document body. A separate actual phone continuation banned Tree, Author and Story until every route was removed, then used free Undo to restore p5 and lock a 90/80 win with two bans. All six routes, twenty explanations, Copy and terminal recovery were reviewed again. The original desktop/phone browser run produced four passes and four targeted failures in 20.4 seconds, with complete p6/p5 wins and all twenty readings before the soft boundary verdicts. It confirmed detached native choices after ordinary peer selection, rejected comparison Preview forcing the wrong view and focus, a phone free-view action missing a newer terminal result, and rejected Replay leaving focus on Replay instead of the adopted terminal heading. No harness error was established. Original focused client checks passed 30 and failed seven; all sixteen existing DOM checks passed. The scoped runtime repair adds main-save fingerprints and accepted-action gates for free/refused choices, view switches, nested Preview, Replay and Copy, retains keyed native route nodes, and guards delayed effects with the current state and view. Pure engine behavior and API, strict snapshot keys, game-rarity values, first-survivor priority, persistent explanations and storage identity remain unchanged. Independent source review found no blocker. Fresh verification passed 37 client checks, comprising nine engine and 28 DOM checks, and sixteen Java checks, comprising nine Hubbreaker checks plus seven inventory checks. Packaging and source/served runtime hashes agreed. All eight repaired first-cycle desktop/phone browser cases passed in 17.2 seconds. Actual repaired play again completed the three-ban p6 90/80 win with active and terminal recovery, formulas, all twenty readings and Copy. Peer Replay now focused visible route content, and a focused native Paper button remained connected and focused while another tab selected Tree. The shared Tree ban then completed a one-ban p6 90/80 win with all six routes, twenty readings, Copy and terminal recovery. The second complete cycle played hb-10, Train to City. All six routes and nineteen explanations were inspected. Station and Ticket bans produced p6; Undo Ticket restored p4, re-banning Ticket restored p6, and the two-ban state survived reload. Banning Passenger left p5 at 90/80 with three bans. The free hypothetical Rail preview honestly showed that no route would survive, while Ban stayed disabled at the exhausted budget. Reload retained that selection, the three actual bans and the current 90 rarity. Lock completed p5 at 90/80; all nineteen readings, six route statuses, the rounded-average formula, Copy and terminal recovery agreed. Review found no further scoped defect. Fresh phone play locked the initial p1 at 20/80 with zero bans and an explicit no-win outcome, then reviewed all six routes, nineteen readings, Copy and terminal recovery. An accepted fresh Replay followed by Station, Ticket and Passenger bans completed p5 at 90/80, with all readings, statuses, formula, Copy and recovery checked again. All four second-cycle desktop/phone browser cases passed in 8.3 seconds. Eight first-cycle and four second-cycle cases provide twelve distinct final passing browser cases. Runtime SHA-256 remained bafe414f3bafc522225895bf4d9eb0b729d04044de72322939116fd2c6f5ead2 through the second cycle. These outcomes are game-rarity target wins and losses, not earned point scores. This is source-informed local agent and automated evidence, with no human, live-provider or deployment validation claimed.

• modules/games/src/main/java/io/kgard/web/slice/games/game64/HubbreakerController.java
• modules/games/src/main/resources/games/hubbreaker-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game64/HubbreakerPlayRouteTest.java
• modules/games/src/test/java/io/kgard/web/slice/games/game64/HubbreakerAuthoredDeckTest.java
• modules/games/docs/games-maturity-followup-20261010.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-64/round1-reviewer-source-notes.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-64/round1-independent-review.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-64/baseline-transcript.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-64/final-transcript.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-64/final-play-checks.json
• conf/tasks/evidence/games-beta-improvements-20261009/additional-20261010/game-64/final-verification.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/77-hubbreaker/verification.json
• conf/tasks/games-two-round-review-20261011/evidence/77-hubbreaker/round-1
• conf/tasks/games-two-round-review-20261011/evidence/77-hubbreaker/round-2
• modules/games/src/main/resources/static/js/games/hubbreaker.js
• modules/games/src/test/js/hubbreaker.test.cjs
• modules/games/src/test/js/hubbreaker-ui.test.cjs
• modules/web-visual/playful-games-tests/hubbreaker.spec.ts
• modules/web-visual/playful-games-tests/hubbreaker-round2.spec.ts
• conf/tasks/games-quality-20261013/results/verification-batch-019-hubbreaker.md

Choose the broad connector that includes both other ideas. Seven selections build a streak, and every choice receives an explanation.

Deutsche Bedienung, Aufgaben, Rückmeldungen und Ergebnislinks im Starterpaket. Die ursprüngliche englische Ausgabe bleibt unter ihrer bisherigen Adresse erhalten; ihre Aufgaben und Spielstände sind getrennt.

English

Alle 8 Aufgaben sind heute als Übung erreichbar. Davon 4 zusätzliche Übungen außerhalb der Tagesrotation. Die Gesamtzahl zählt jede Aufgabe einmal.

12 Aufgaben: 8 in der Tagesrotation, 4 separate Übungen

Tagesauswahl aus 8 Aufgaben; danach werden bekannte Aufgaben erneut verwendet.

Neben der Tagesaufgabe sind 4 separate Übungsaufgaben am selben Tag spielbar. Wiederholungen bleiben bekannte Inhalte.

12 authored boards (8 repeating daily + 4 practice)

Solo

Limited beta ready

Twelve finite authored English packs provide eight repeating daily boards and four practice boards, each with seven three-word rounds. A hub is the broad category that includes the other two words; it is not measured graph centrality or the older global-rank concept. Each round has one committed Check: a correct choice earns 100 points and a wrong choice earns zero. Empty Check and selection changes before commitment are free. Explicit Next follows the committed explanation. Revealing the current or remaining rounds earns zero for those rounds while preserving earlier earned points; the displayed streak adds no bonus. The default route makes no live Contextoid or Termpath request. Progress and drafts are browser-local without account or cross-device synchronization, and changed-save protection remains a non-atomic preflight. Unavailable storage permits warned memory-only play. The retained provider-backed service was not exercised as live play in these authored cycles. Human fairness, unprimed onboarding, difficulty, enjoyment, replay interest and public production readiness remain unverified.

The13 October independent review accepted Hub Hunter unchanged after full seven-round play, partial/revealed500 and replay700, exact reload and real read-only-save recovery. All12 boards and84 category decisions were reviewed; native input and normal-motion endings passed. The finite introductory bank remains a replay limitation. Accepted for invited human playtesting; human feedback remains pending. This is local technical evidence, not deployment or public-release signoff.

Authored category/member triples; broadest category is the connector.

No live data mode on the default player route.

Spring/JTE + browser client; packaged authored content and browser-local progress; no cross-device account sync.

5 Verbesserungen umgesetzt. Die Höchstzahl ist kein Soll; eine begründete frühere Beendigung bleibt sichtbar.

Stand: 2026-10-20. Browserprüfungen belegen keine menschliche Spielerfahrung.

• conf/tasks/games-nonviral-ten-rounds-20261020/evidence/learning/ledger.json

Die Sortierung „Weitere Prüfrunden“ zählt alle nachfolgenden Prüfungen zusammen. Die ersten fünf Runden bleiben separat.

2 von höchstens 10 Runden am 2026-10-20: Eigene Antworten und aufgelöste Lösungen bleiben unterscheidbar. Fehlerkorrektur und vollständige Runde geprüft; keine weitere Änderung begründet.

• conf/tasks/games-nonviral-further-rounds-20261020/evidence/learning/ledger.json

3 frühere Prüfeinträge · 1 in diesen Runden behobene Befunde.

Prüfeinträge können unterschiedliche Tests und Nachprüfungen enthalten. Ihre Anzahl ist keine Qualitätsnote. Eine Runde umfasst Review, Änderung oder begründeten Verzicht, Spielen und erneuten Review. Automatisierte Browserpartien ersetzen keine menschlichen Spieltests.

Stand: 2026-10-20

• conf/tasks/games-quality-20261013/game-quality.json
• conf/tasks/games-nonviral-five-rounds-20261020/selection.json
• conf/tasks/games-nonviral-five-rounds-20261020/evidence/learning-next/ledger.json

Neue und bisherige Aufgaben mit deutschsprachigen Ersttestern vergleichen: Verständnis, strittige Antworten, genutzte Hilfen und freiwillige weitere Runde beobachten. Die neuen Aufgaben sind noch nicht durch Teilnehmer validiert.

Collect unprimed human feedback on comprehension, fairness, difficulty and replay interest; technical readiness and human feedback remain separate in conf/tasks/games-quality-20261013/game-quality.json.

Reassessed 11 October 2026 after two completed local play, review, improvement and replay cycles. Fresh People at work 700 before and after repair, mixed Travel 500, honest partial 200 with five unearned reveals and fresh all-seven 700 were reviewed with every category explanation, chronological outcome row, canonical Copy and active/terminal recovery. Reproduced native-focus, changed-save acceptance and owning-navigation defects were repaired while preserving the pure rules, tolerant recovery, identities and point economy. Verification passed 24 fresh client checks, fourteen fresh Java checks, packaging and twelve distinct final desktop/phone browser journeys; independent source and bounded visual reviews found no blocker. Retains Limited beta ready, 4. The historical finite-bank, human fairness, unprimed onboarding, difficulty and replay-interest gap remains open. Local correctness and reliability do not establish human enjoyment, live-provider behavior or public production readiness.

Maturity gap: Human beta feedback is needed on fairness, unprimed onboarding, difficulty and replay interest. The authored starter bank is finite.

The mapped default Hub Hunter route uses hub_hunter.jte, the finite authored hub-hunter-deck.json and hub-hunter.js. Runtime identity game55 differs from inventory game-55; shared authored-v1 storage uses rulesVersion hub-hunter-1. Retained provider-backed service methods and older rank-based hub assumptions do not define this player route. The first full cycle began with a fresh People at work attempt on hh-09. Empty Check used no round; revising Poet to Writer was free, and active reload preserved Writer. Writer earned 100 and its Poet/Novelist category explanation was read. Performer then earned 100; committed reload preserved 200 points, streak two, frozen choices and the explanation before explicit Next. Scientist, Athlete, Musician, Craftsperson and Artist each earned 100 after their respective explanations and explicit progression. The complete result was 700/700, seven of seven correct and streak seven. All seven chronological outcome/explanation rows, canonical Copy and terminal recovery agreed. The original root client run passed fourteen of 24 checks, comprising eight pure-engine and sixteen DOM checks, with ten targeted failures and no skips. All eight original desktop/phone browser journeys reached complete earned outcomes before their verdicts; two passed and six failed targeted catalog, native/help focus, delayed free-refusal adoption or rejected-action focus assertions in 51.4 seconds. The scoped repair keeps each current round’s native radio nodes stable, checks current saves before actions, empty Check, premature Next and Copy, and applies requested effects only after accepted state changes. Peer adoption and rejected Replay restore focus to the actual active round, committed explanation or complete result, preserving still-valid native and disclosure focus. The startup observer aligns after shared recovery without rewriting arrival bytes. The owning All games destination was corrected to the catalog. The pure engine/export prefix, tolerant six-field recovered state, shared-session identity, authored deck, chronology, scoring and reveal semantics remain unchanged. Fresh final verification passed all 24 client checks: eight pure-engine and sixteen DOM checks. Fourteen fresh Java checks passed, and neither client nor Java verification skipped checks. All eight repaired first-cycle desktop/phone browser journeys passed in 32.4 seconds. Source, packaged and served runtime hashes agreed, and packaged findings matched the then-current 29-row refreshed inventory. Independent source review passed and the static design check returned no findings. Fresh actual repaired People at work Replay repeated the free empty Check, Poet-to-Writer revision and active draft recovery, then earned all seven categories for 700. Every explanation, committed recovery boundary, terminal row, canonical Copy and terminal reload agreed. The first bounded visual review inspected complete desktop 700 and phone memory 700 results, including all seven explanations and selected manual Copy; both were readable without overflow. The second full cycle used a fresh phone Travel attempt on hh-10. Hostel was a committed miss for zero, Luggage earned 100, and the current Public transport round was revealed for zero. Committed reload preserved 100 earned points and reset streak. Rail vehicle, Aircraft, Vessel and Road each earned 100 after explicit Next and explanation review. The complete result was 500/700, five correct and final streak four, with all seven outcome rows, canonical Copy and terminal recovery checked. Review found no additional scoped defect, so the runtime and template remained frozen. A fresh explicit Travel Replay earned Accommodation and Luggage for 200. Public transport was selected but remained uncommitted through active reload. Reveal remaining settled the five unplayed rounds for zero while retaining the two earned rounds, producing an honest 200/700 result. All seven explanation rows, canonical Copy and terminal reload agreed. A further fresh explicit Replay earned Accommodation, Luggage, Public transport, Rail vehicle, Aircraft, Vessel and Road first try for 700/700. Each explanation was read before Next, and all seven terminal rows, canonical Copy and recovery were reviewed. All four second-cycle desktop/phone cases passed in 14.9 seconds, giving twelve distinct final browser passes across both cycles; all 24 final client checks passed again. The second bounded phone visual review confirmed that the partial 200 result clearly distinguished five revealed zero-point rounds. Runtime SHA-256 remained 73ead2862eafe4bc18c2b4fbf317eda00097d98257b84fa0f7c60b8a30bfcf74 throughout final replay verification. This is source-informed local agent and automated evidence; no human, live-provider or deployment validation is claimed. Earlier known-issue record, retained for history: Resolved and locally verified 11 October 2026: native choice controls remain stable within the current round, and current-save checks cover actions, empty Check, premature Next and Copy. Rejected actions adopt current progress without performing the requested effects. Peer adoption and rejected Replay rescue hidden, detached or disabled focus to the actual active round, committed explanation or complete result while retaining valid native focus. Startup observation aligns after shared recovery without rewriting the arrival save. The owning All games link now opens the catalog. Final replays established no further defect.

• modules/games/src/main/java/io/kgard/web/slice/games/game55/HubHunterController.java
• modules/games/src/main/resources/games/hub-hunter-deck.json
• modules/games/src/test/java/io/kgard/web/slice/games/game55/HubHunterAuthoredDeckTest.java
• modules/games/docs/games-maturity-20261009-prototype-beta.json
• conf/tasks/evidence/prototype-beta-20261009/path-games
• conf/tasks/evidence/prototype-beta-20261009/final-verification.json
• conf/tasks/evidence/prototype-beta-20261009/final-surfaces/surface-check.json
• conf/tasks/games-two-round-review-20261011/queue.json
• conf/tasks/games-two-round-review-20261011/evidence/70-hub-hunter/verification.json
• conf/tasks/games-two-round-review-20261011/evidence/70-hub-hunter/round-1
• conf/tasks/games-two-round-review-20261011/evidence/70-hub-hunter/round-2
• modules/games/src/main/java/io/kgard/web/slice/games/game55/hub_hunter.jte
• modules/games/src/main/resources/static/js/games/hub-hunter.js
• modules/games/src/test/js/hub-hunter.test.cjs
• modules/games/src/test/js/hub-hunter-ui.test.cjs
• modules/web-visual/playful-games-tests/hub-hunter.spec.ts
• modules/web-visual/playful-games-tests/hub-hunter-round2.spec.ts
• conf/tasks/games-quality-20261013/results/verification-batch-011.md

Investigate an undirected district graph, then flag every node meeting the printed degree and neighboring-district warrant rules. Two optional reports support the investigation.