Survey counts 2,170 real-world uses of Jev
In one week 1,865 new repos and 305 existing projects adopted Jev, with 77% using it for attribute judgment, 52% for scoring and ranking and 31% for action selection, across browser automation, code review, model routing and a Mario emulator controller.
Chinese students found 2,170 real world uses for Jev, and it’s already replacing thousands of unnecessary LLM decisions 1 → 1,865 new repos appeared after launch and 305 existing projects added Jev. The new repos picked up 43,750 stars in one week 2 → 77% of projects use Jev for attribute judgment, 52% for scoring and ranking, 31% for action selection, and some for routing, filtering and model selection 3 → 36.8% use all three interfaces (Choice, Noul and Score), so one system can hand Jev several kinds of decisions 4 → The paper walks through real workflows: browser automation, code review triage, trimming tool history, model routing, risk gates on tool calls, and a controller for a Mario emulator 5 → Most of these are tiny calls: pick a tool, check an action, route a request, decide what context to keep. Jev handles them without another expensive LLM call Jev is starting to look like a cheap decision primitive for AI agents
This post quotes another post: x.com/leopardracer/status/2103433480885379248