Community · E2 · artifact verified
Review commits and copy with typed questions
nudgement reviews commit messages, code, comments, tests and UI copy before you commit: exact checks plus focused Jev questions catch what formatters cannot - a message that misrepresents the diff, or a test that passes even when behavior broke.
01 · Role in the system
What Jev does here
Each artifact gets exact checks first, then focused typed questions: clarity passes ask per-item Nouls such as whether an explanation is missing or a magic value is unexplained, while other passes judge whether a commit message honestly describes its diff and whether a test actually exercises the behavior it names. The design keeps questions narrow and mechanical - one concern per question, answers as calibrated Nouls - so findings read like lint with probabilities rather than an essay. An interactive introduction with real examples runs at nudgement.dev, and a recorded-replay script lets runs be replayed deterministically. It is the pre-commit complement to a formatter: honesty and consistency checks that no formatter can do.
02 · Control boundary
Where Jev sits
Jev as the judgment layer of a pre-commit reviewer: exact checks first, one narrow typed Noul per concern (missing explanation, magic value, message-diff honesty, test-behavior match), and a replay script for deterministic re-runs.
Code owns the loop, permissions, thresholds, validation, and side effects. Jev owns only the bounded judgments described above.
03 · Known limits
What this evidence does not prove
- New project (1 star at review time); question sets cover common cases, not exotic ones.
- Noul thresholds are fixed author choices; no benchmark against human reviewers.
- Focuses on single-repo artifacts; cross-repo or CI-wide sweeps are out of scope.
04 · Attribution
Public sources
This is a Community record: the project was published by a third-party community author.
- Nicholas-Westby ↗Community · github · public · checked 2026-09-30