AI coding assistants need repro notes too

A debugging-desk bridge between AI tool choice and reproducible failure reports.

When an AI coding assistant makes a mistake, the team often writes the wrong lesson. “Tool missed the bug” is too vague. What was the prompt? What file context was visible? Did the assistant invent a path, skip a test, preserve an old assumption, or answer a different question? The same rule as debugging applies: write the gap, not only the complaint. Otherwise the team cannot tell whether the next tool is better or merely wrong in a different accent.