What it does
Every request is content-hashed and never re-sent. The project says plainly that it is not a benchmark; it is looking for where Jev's judgments are jagged.
Maker-reported (not independently measured by JevMade): ~99.5k questions answered by Jev on a 1,319-node tree · gateway p50 383 ms, p95 727 ms end to end; marketCost about $0.000017 per ~450-token call
- Primitives
- choice, score, noul
- Platform
- Python
- Added
- Project created
Source checked 2026-09-24 — opened the GitHub repository directly.