New episode: Jev and the new decision layer for coding agents.
@typesafeai’s Jev gives us a new place to put intelligence: between deterministic code and expensive LLM reasoning.
We look at:
- fast browser-use loops, including
@thdxr’s OpenCode demo.
- live feedback while you type, including
@rileybrown’s real-time post analyzer.
- our own experiments with test selection and local decision models, plus how fine-tuning could work
- model, reasoning-effort, skill, and tool routing.
If cheap, fast judgment can sit inside these loops continuously, how should we redesign agent systems around it?