What it is
A standalone SO-101 playground for developers: a Bun/Effect coordinator, React workbench, Rerun visual workspace, explicit robot tools, and a LeRobot hardware boundary, defaulting to mock hardware on a fresh checkout.
How it uses Jev
The decision runner (bun run jev) calls Jev via the AI SDK experimental_evaluate API on Vercel AI Gateway. Jev picks bounded joint steps from typed candidate actions under a spend budget; runs write decision.* events and JSONL logs.
Primitives:choice
Technique worth stealing
Treat the evaluation model as a decision runner over typed candidate actions, capped by a cumulative USD budget.
Try it
bun install && uv sync --extra dev && bun run build && bun run dev, then bun run jev --smoke
Evidence
Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.
- Jev-centricyes0.71
- Shows a System One patternyes0.69
- Handles uncertaintyno0.05
- Measuredno0.11
- Runnableyes0.91
- Worth recommendingno0.31
- Model replicano0.24
- Problem scopescore on a 0–2 scale0.80
- About Jevyes0.94
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.