Verify report query explanation
An analyst pastes a slow report query into an AI assistant and asks what it does. The agent says: "It returns one row per customer with their total order count and lifetime revenue, excluding customers with no orders." You're about to use this query as the source of truth for a churn dashboard. How do you verify the explanation, and what would you specifically distrust?
Implement
summarize_customer_orders(customer_ids: list[str], order_lines: list[str], refund_lines: list[str]) → list[str]Examples
in
[["c1","c2","c3"],["o1|c1|1000","o2|c1|2500","o3|c2|4000"],["r1|o1|250","r2|o1|250"]]out["c1|2|3000","c2|1|4000","c3|0|0"]in
[["c9"],[],[]]out["c9|0|0"]What a strong answer looks like
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.
0:00 of about 22 min
Vibe & agentic: describe the solution in plain language (or narrate it) and the coach grades your approach.
Which questions mattered is sealed until you submit. Telling you now would just be handing over the edge cases.
Run or narrate your approach, then ask the coach.