Code RoomDead code verification
MediumPrep Room Coding #4333

Dead code verification

Vibe & agenticAlgorithms & data structuresMid–Senior~17 min

Onboarding to a service, you ask an AI to identify which code paths handle the /checkout endpoint so you can find where to add a discount step. The agent traces it confidently to CheckoutController.process() and says "that's the only handler; the legacy CheckoutV1 module is dead code you can ignore." You're tempted to take that at face value. How do you verify the V1 module is actually dead before designing around it?

Implement
classify_dead_code(module_names: list[str], static_refs: list[int], request_counts: list[int], window_days: int) → list[str]
Examples
in[["CheckoutV1","CheckoutV2","LegacyPromo"],[0,5,0],[0,0,3],60]out["CheckoutV1=dead","CheckoutV2=referenced","LegacyPromo=live"]
in[["CheckoutV1"],[0],[0],29]out["CheckoutV1=unproven"]
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 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.