Verify untrusted code fast
An agent produced 400 lines of Rust implementing a new rate limiter you must ship by end of day. You can't deeply read and reason about all 400 lines in the time you have, but you own the merge. How do you verify enough to ship responsibly under the deadline — what do you read closely, what do you test, and what do you accept on faith?
Implement
triage_review_effort(risk_levels: list[str], covered_by_tests: list[bool], read_budget: int) → list[str]Examples
in
[["irreversible","degrades","cosmetic"],[false,true,true],5]out["read","test","trust"]in
[["irreversible","irreversible","irreversible"],[true,true,true],1]out["read","gate","gate"]in
[["cosmetic","cosmetic"],[true,false],0]out["trust","test"]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 20 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.