Debounce test flakes with sleep
An AI agent's async JS test for a debounced search was flaky in CI. The agent 'fixed' it by adding a sleep:
It now passes locally and 'usually' passes in CI. Why is this fix wrong, and how should the test be made reliable?
Implement
debounce_calls(events: list[str], wait_ms: int) → list[str]Examples
in
[["0|a","100|ab","200|abc"],300]out["abc"]in
[["0|a","300|b"],300]out["a","b"]in
[["0|a","299|b"],300]out["b"]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 16 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.