Deploy succeeds without confirming health
An AI wrote this blue-green deploy script. It exits 0 and the pipeline goes green, but one morning prod was serving a broken build while CI reported success. Walk through how this script lies about success.
Implement
deploy_exit_code(artifact_url: str, downloaded_bytes: int, health_body: str) → intExamples
in
["https://artifacts.acme.io/app.tar.gz",10485760,"{\"status\":\"ok\",\"version\":\"42\"}"]out0in
["null",0,""]out2What 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 17 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.