Sphinx-style pre-ship understanding gate

Quiz developers on AI-written code before it ships

Problem

More and more code is written by AI agents; it runs, but the author often can't explain why. A Show HN project that pops a questionnaire before pushing to prod shows this is a real anxiety, and team leads worry about juniors merging AI output straight into main.

Solution

A lightweight GitHub App/CLI that generates 3-5 targeted questions about each PR diff (why this change, edge cases, how to debug failures), scores the developer's answers, and flags low-understanding changes for reviewers.

Users

Small teams heavy on AI coding agents, contract developers, tech leads reviewing lots of AI code

MVP

GitHub PR webhook to read diffs; LLM-generated targeted Q&A; post answers and comprehension score as PR comment; gate only high-risk diffs; ignore rules to reduce noise

Revenue

Per-repo or per-seat subscription; free for personal repos, $15/user/month team tier with audit reports

Why now

AI-written code share is rising fast and understanding debt is a new team pain; existing review tools check correctness, not author comprehension