Question
Your team wants to backfill docstrings across a large legacy Python service by running an AI over every function and committing whatever it writes. A staff engineer is uneasy about merging thousands of AI-written docstrings in bulk. When is AI-generated documentation the wrong tool, or at least the wrong process — and how would you make a bulk-docstring effort safe rather than block it outright?
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.