jevals: replacing LLM judges with typed Jev decisions for agent evals

jevals is an open-source tool that replaces LLM judges with typed Jev decision models for agent evals and guardrails. All evals for a trace are sent as one HTTP request costing a few thousandths of a cent and returning in a few hundred milliseconds, suitable for every trace or inside the agent loop. It supports Jev, Kev, Laya, and regular chat LLMs.

• Four AI Models Tested on Doom Control Benchmark2026-09-20
• Developer Says His Non-Autoregressive Decision Model Was Rebranded a Breakthrough2026-09-19
• Kev: Tiny decision-model family built on Qwen3.52026-09-21
• ProgramAsWeights Compiles English into Local Neural Programs2026-09-20
• Open-source 395M 'System One' model Von released2026-09-19