One-Room Semantic Party

Host a solo forecasting puzzle with three authored guests. Pick Safe, Bold or Weird: each clue invites words from the printed families. Classify the stated scene as Written text, Objects or Places; a word’s other senses do not apply. A guest first seeks their highest preference that appears in those families. All words at that preference are equally acceptable; your correct prediction becomes their response. A wrong prediction uses the first eligible top-choice word in the printed word list. Each guest responds once. Earn up to 50 for accurate predictions, 25 for each distinct goal reached, and lose 20 for each trap response; total is clamped to 0–100. Commit one complete plan, or reveal unscored. Guests are written rules, not measured human responses.