What it does
Jev scores shuffled word buckets, then a separate Noul question decides when the reply is complete.
Humor & parody
jevbot
Make a non-generative decision model talk by holding a 20,000-word tournament for every broken sentence it posts to Discord.
Bookmark: jevbot
Keep this in your collection.
Leave a noteWhat would you try with this?
: jevbot
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
How you can use it
For a playful writing experiment, try changing the allowed word list to shape what your bot can say. This project chooses each next word through several rounds of voting, then checks whether the reply is finished. A developer can adapt that approach. It makes many paid AI requests per word, and the sample replies are often broken sentences, not polished conversation.