Colin Breck's 'I don't want to read what you didn't write' tops 1,000 points on Hacker News
Colin Breck / Hacker News·medium signal
Breck's Sep 20 essay argues that AI-drafted design docs, PRs and summaries fail because the prompter has context the reader lacks. He cites figures that 78% of readers stop reading when they suspect AI authorship and 98% prefer flawed human writing. He still uses AI for verification, citations, grammar and diagrams, never for first drafts, and the HN thread (1,018 points) added the information-theory point that an LLM cannot supply the 700 bits the sender never gave it.