At-most-once delivery assumption in SQS
An AI assistant wrote this SQS consumer in Python that processes payout messages. It treats each delivery as exactly-once. Review the concurrency/delivery assumptions.
What goes wrong in production, and how do you make it correct?
Implement
apply_payouts(account_ids: list[str], starting_balances: list[int], deliveries: list[str]) → list[int]Examples
in
[["acct-1","acct-2"],[100,50],["p1|acct-1|25","p1|acct-1|25","p2|acct-2|10"]]out[125,60]in
[["acct-1"],[0],["p1|acct-1|5","p2|acct-1|5"]]out[10]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 22 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.