What it does
The Codenames player measures associations for every board word, while ordinary game rules reject illegal clues.
Benchmarks & research
jev-lab
Experiment with a Jev-powered Codenames player and a separate CLI for checking determinism, sensitivity and calibration.
Bookmark: jev-lab
Keep this in your collection.
Leave a noteWhat would you try with this?
: jev-lab
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
How you can use it
You can adapt this idea to create word-association games like Codenames. Start by picking your word list and writing down your turn rules. A developer can build ordinary game rules to run the board, while asking the model to score how strongly a clue connects to each word on the table.
Running this project requires an access key from AI services like TypeSafe or OpenRouter so the software can reach the model. The model only estimates word links, while your own rules decide which scores count as safe, risky, or strong enough to play.