AI Code Review Roulette is a small QA survival test for the moment after an agent writes the pull request and before everyone pretends the green checkmark means safety. Paste a diff, or describe the change in plain English, then describe the test setup honestly: Playwright, unit tests, CI gates, visual checks, contract tests, and how much of the code came from an AI coding agent.
The tool returns a QA Survival Score from 0 to 100, a letter grade, and the three failure modes most likely to escape to production. It also pulls one randomized horror scenario from a bank of real bug classes: selector rot, missing production env vars, prompt-injected cleanup, retry storms, access-control gaps, migration risk, stale state, and the other classics that make QA teams sound cynical because they are usually correct.
The score is intentionally legible. You can see the points gained for end-to-end coverage, CI gates, unit tests, contract checks, and accessibility passes, plus the penalties for large diffs, high AI-written percentages, auth-sensitive changes, schema edits, and thin test descriptions. That makes the loop useful instead of random: improve the coverage story, spin again, and watch the survival rate move.
Use it as an AI code review checklist before merging a vibe-coded PR, a quick prompt for deciding which tests to add, or a shareable risk card when a teammate says the agent already tested it. The wheel is sardonic, but the output is practical: merge, add tests, or send the PR back to the agent with a better assignment.