Question
You need to rename a core domain concept across a polyglot repo — the term `tenant` becomes `organization` everywhere: class names, DB columns, API field names, config keys, comments, and a public webhook payload field. A teammate suggests letting the AI agent do a repo-wide text replace. What goes wrong with that approach, and how do you actually scope and verify this rename?
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.