I made Jev to talk to us like an LLM (almost) (github.com)

Turns out that Jev is pretty good at replaying true/false on which word would be best to append to the response. It can process up to 1000 words at the time.