On 48 real game questions, Jev plus a Gemini fallback matched Grok's accuracy with 84.0% lower mean latency and 88.5% lower mean cost.

3 points•pampas•9 days ago•3 comments•

3 comments

proc09 days ago
I'm curious about integrating Jev with testing. You throw it on a loop to check the test results and have the more general AIs fix things until tests pass. I'm sure something like this exists but it now can have more determinism and cheaper.
pampas9 days ago
It might be good at routing fixes to the most appropriate LLM.
pampas9 days ago
I used Jev to make my AI game 6x faster and 8.7x cheaper.

Read the full thread on Hacker News →

Related stories