Jev in production › Agent context and memory
Open-source coding agent that answers codebase questions using tree-sitter, SQLite memory and Jev judgments. Up to 1/18 of frontier model cost with equal or better accuracy (vendor claim).
On 24/10/2026 I have published as OSS a framework I call unified-brain: evidence gathering that is fast and cheap, and close to a frontier coding agent in accuracy. On the same nine-question, 74-point xz benchmark (evals/questions/xz.md) it has four measured profiles (env.example). The strongest agent I have measured is Claude Fable 5.2 at max reasoning effort: 72.75 / 74, in 5,372 s, for about $50 (evals/results/qa-xz-2026-11-13-fable-5-2-agents.md). Against it:
For the project's own README, linking back here: