jevbooks

← All projects

beat-the-reviewer

Ash20pk/beat-the-reviewer · Homepage

Ten approvals to win, judged by a probability, not a sentence.

Verification & Guardrails76%Runner-up: Evaluation & Benchmarking

What it is

A game where you write pull request bodies to convince a reviewer to approve ten risky production requests. For developers who want to practice precise, scoped arguments under a probabilistic gate.

How it uses Jev

Each level is a typed judgement rule: a deterministic gate, a question, and a threshold. The question goes to jev-1.13.0, which returns a probability. The threshold turns that number into APPROVED or REFUSED. The verdict and number appear in the PR comments.

Primitives:score

Technique worth stealing

Pinning the model version and hashing the question so thresholds as fine as 0.25 cannot silently retune.

Try it

Use the GitHub template, clone your copy, create a branch named level-01, and open a pull request with your request in the body.

View on GitHub

judged by Jevjev-1.13.0

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-centricyes0.91
  • Shows a System One patternyes0.73
  • Handles uncertaintyunclear0.48
  • Measuredunclear0.52
  • Runnableunclear0.53
  • Worth recommendingno0.39
  • Model replicano0.17
  • Problem scopescore on a 0–2 scale1.68
  • About Jevyes0.89

Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.