Question
Your team is porting a Python service that computes per-customer billing rollups to Go for throughput. You point an AI agent at the Python module and ask it to produce an idiomatic Go port. The math involves summing thousands of Decimal line items and rounding to cents. How do you scope this port for the agent, and how do you guarantee the Go output produces byte-for-byte identical totals to the Python version before you cut over?
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.
Vibe coding: describe the solution in plain language (or narrate it) and the coach grades your approach. Generating runnable code from your description is coming next.