What it is
A 12-notebook Python course for developers who want to combine Jev, a fast decision model, with a slow LLM. It teaches agent safety, routing, triage, and evaluation through 13 real use cases.
How it uses Jev
Before every tool call, Jev answers a Choice (allow/ask/block) with Noul checks for irreversibility and exfiltration; ask verdicts go to a human and errors fail closed. A Choice model router picks fast or capable LLM with confidence fallback, and a Noul gate checks if the agent is done.
Primitives:choicenoul
Technique worth stealing
Use Jev for fast, calibrated decisions inside an agent loop, and let the LLM handle language tasks.
Try it
Clone repo, run ./setup.sh, add OpenRouter key to .env, then uv run jupyter lab and open 01_hello_jev.ipynb.
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-centricyes0.94
- Shows a System One patternyes0.92
- Handles uncertaintyyes0.91
- Measuredunclear0.51
- Runnableno0.15
- Worth recommendingyes0.79
- Model replicano0.06
- Problem scopescore on a 0–2 scale1.47
- About Jevyes0.99
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 23 Sept 2026.