Question
You ask an AI to add caching to a TypeScript function that fetches a user's feature flags from an API. It wraps it in a module-level `Map` cache keyed by userId — fast, but now flags never update until redeploy, and in your serverless runtime each cold invocation has an empty cache so the hit rate is ~0 anyway. You re-prompt 'add cache expiry'; it adds a 1-hour TTL, which fixes staleness but not the cold-start miss or a thundering herd on expiry. How do you re-steer?
Treat the AI’s output as a draft to verify, not an answer to trust. Name the specific flaw and the input that triggers it, say how you’d catch it — tests, edge cases, reading critically — and how you’d re-prompt or decompose to get it right.
Vibe coding: describe the solution in plain language (or narrate it) and the coach grades your approach. Generating runnable code from your description is coming next.