Voices
A Developer's Counter-Workflow Hits HN: Have the LLM Write It in Chat, Then Retype Every Line By Hand
Ankur Sethi's August 2 post 'Prevent cognitive debt by manually retyping LLM-generated code' reached 503 points on Hacker News as a direct rejection of the review-the-PR workflow. He keeps the assistant in chat, never lets it write to the repo, and types each line himself: 'Typing the code myself forces me to slow down, which means I'm more likely to detect hallucinations or bad design choices.' He is explicit about the tradeoff — 'I value comprehension over productivity' — which lands as a deliberate counterweight to the Yegge-style fleet-orchestration posts trending the same week.
↳ Follow the thread