You and an LLM each play a connecting word per turn and try to meet in the middle within eight rounds; the model drops riddle hints in the final two turns.
Quick answer:Converge is one of only 9 live word games where the AI produces language during play. It scores 8/10. It is a daily puzzle, on the web. Verified live August 13, 2026.
Disclosure: WordGameAI is operated by MidnightDev, who make Yapword. This page scores Converge by exactly the same published rubric we apply to our own game.
Score breakdown
The AI Involvement Score measures how much the AI does while you play — not how good the game is. Full rubric here.
✓Runs at play time A model executes while you are playing.3/3
✓Generates language back to you It produces original text you read during play.2/2
✓Evaluates your input semantically Your guess is judged by meaning, not spelling.2/2
✓Responsive hints Hints are derived from the model and the current board.1/1
×Adapts to you Every player gets the same behaviour; the AI holds no state about you.0/1
×Model disclosed by the operator Third-party reporting names a model, but the operator does not — so it does not count.0/1
AI Involvement Score8/10
Without the model-disclosure point: 8/9.
The record
Developer
Chase Wackerfuss
Launched
2025
Model disclosed
unnamed LLM (two structured-output calls per turn)
Daily puzzle
yes
Unlimited play
not verified
Difficulty levels
not verified
Multiplayer / challenges
not verified
Themed or custom games
not verified
Account required
not verified
Price
not verified
Playable archive
not verified
Platforms
Web
AI type
AI collaborator
Last verified
2026-08-13
We could not verify: Unlimited play, Difficulty levels, Multiplayer / challenges, Themed or custom games, Account required, Price, Playable archive. These are recorded as unknown rather than guessed, and unknown counts against the score.
Yapword is one of 9 live word games where the AI produces language during play, and one of 3 that respond to how you personally played. Free daily, no account.
Converge scores 8 out of 10 on the WordGameAI rubric. A model runs while you play, so yes — the AI is doing work every turn. You and an LLM each play a connecting word per turn and try to meet in the middle within eight rounds; the model drops riddle hints in the final two turns.
What AI model does Converge use?
Not officially. Third-party reporting points to unnamed LLM (two structured-output calls per turn), but the operator does not state it, so we do not award the disclosure point. We only count what the operator says about their own product.
What could you not verify about Converge?
We could not verify unlimited play, difficulty levels, multiplayer / challenges, themed or custom games, account required and more. Those fields are recorded as unknown rather than guessed, and unknown counts against the score. If you operate Converge and can fill them in, we will update the record and re-date it.