Code RoomCustomer data to AI support
MediumPrep Room Coding #4176

Customer data to AI support

Vibe & agenticAlgorithms & data structuresMid–Senior~16 min

Support engineers keep hitting a bug only reproducible with a specific customer's account state. One proposes a slick workflow: when a ticket is stuck, export the customer's full record (settings, message history, transaction list) into an AI agent and ask it to write a repro script and suggest a fix. It would massively speed up support. As tech lead, how do you respond, and what would a safe version look like?

Implement
sanitize_account_state(field_lines: list[str], safe_keys: list[str]) → list[str]
Examples
in[["plan=pro","email=ada@example.com","message_count=412","beta_flag=true"],["plan","beta_flag"]]out["plan=pro","email=<str>","message_count=<int>","beta_flag=true"]
in[["email=x@y.z","retry_count=-3","coupon="],["email"]]out["email=<str>","retry_count=<int>","coupon=<empty>"]
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.