When to disclose AI assistance
Your team has no written policy on disclosing AI assistance. Over a sprint you use an AI agent in four different ways: (1) Claude wrote a throwaway Python script to backfill a column once; (2) it generated a non-trivial concurrency-control change in your Java payment service; (3) it suggested the wording of a customer-facing error message; (4) it drafted a design doc you then heavily rewrote. For which of these does disclosing AI involvement actually matter, and what's the principle that decides?
flag_ai_disclosure(names: list[str], risk_scores: list[int], human_reworked: list[bool]) → list[str][["backfill-script","concurrency-change","error-message","design-doc"],[1,5,2,3],[false,false,true,true]]out["concurrency-change"][[],[],[]]out[][["risky-helper"],[3],[false]]out["risky-helper"]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 & agentic: describe the solution in plain language (or narrate it) and the coach grades your approach.