What it is
jgrep is a semantic code search CLI for developers: it splits files or git diff hunks into 5–60 line chunks, asks Jev a yes/no question about each chunk, and prints file:line hits above a probability threshold, so diffs can be linted in CI with English rules.
How it uses Jev
Each file or diff hunk becomes a chunk state; jgrep packs up to 16 chunks with one Noul question per chunk asking whether the chunk matches the description, turns the returned probabilities into file:line hits above -t, and caches answers by (model, question, chunk).
Primitives:noul
Technique worth stealing
Packing many isolated chunks and one Noul question per chunk into a single request, then thresholding calibrated probabilities into grep-style hits.
Try it
npm install -g jevgrep; jgrep init; jgrep "catches an error and silently ignores it" src/
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-centricyes0.90
- Shows a System One patternyes0.88
- Handles uncertaintyyes0.67
- Measuredyes0.65
- Runnableyes0.98
- Worth recommendingunclear0.50
- Model replicano0.03
- Problem scopescore on a 0–2 scale1.94
- About Jevyes0.98
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.