Iterative code edits in chat
A mid-level engineer keeps a chat session open and types 'now add a null check on line 14, now rename this var, now extract this to a helper' one instruction at a time, copy-pasting code back and forth, for a 30-minute editing session inside one Go file. Which AI modality fits this work better, and what's the cost of the one they chose?
Implement
select_edit_modality(instruction_count: int, distinct_files: int, needs_explanation: bool, longest_edit_lines: int) → strExamples
in
[12,1,false,3]out"inline"in
[1,1,true,4]out"chat"in
[4,3,false,6]out"chat"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 12 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.