How to Write with an LLM: Thomas Ptacek's copyediting rules, endorsed by Simon Willison

Thomas Ptacek published a method for LLM-assisted writing built on two rules — never adopt a single word or phrase a model suggests, and forbid model encouragement — and Simon Willison endorsed the approach the same day, saying he uses LLMs only for…

Thomas Ptacek published an essay, 'How to Write with an LLM,' laying out two rules for using LLMs as copyeditors rather than ghostwriters: (1) never use any word or phrase a model suggests, and (2) forbid model encouragement, which reinforces first-draft impulses and erodes author voice. He recommends running model passes to flag passive voice, repetition, and misplaced paragraphs, and comparing rewrites with a fresh-context model to avoid bias. He also recommends the book 'Style: Lessons in Clarity and Grace' for systematic editing and mentions building a small tool to manage context-free copyediting comparisons. Simon Willison highlighted the essay on 2026-09-17, endorsing the no-suggested-phrasing rule as useful discipline against recognizable AI output, and disclosed that he himself uses LLMs only for fact-checking, spelling, grammar, and as an occasional thesaurus — never to draft blog content. The two reports are complementary with no factual disagreements; Willison's post corroborates Ptacek's core rule.

• Thomas Ptacek published an essay titled 'How to Write with an LLM' (Report 1, Hacker News, 2026-09-17).
• Rule 1: never use a single word or phrase an LLM suggests in published prose (stated in both reports).
• Rule 2: forbid model encouragement to avoid reinforcing first-draft impulses (Report 1).
• Ptacek recommends model passes to flag passive voice, repetition, and misplaced paragraphs (Report 1).
• He recommends comparing rewrites with a fresh-context model to avoid bias (Report 1).
• Ptacek recommends the book 'Style: Lessons in Clarity and Grace' for systematic editing (Report 1).
• Ptacek built a small tool to manage context-free copyediting comparisons (Report 1).
• Simon Willison endorsed the essay on 2026-09-17, framing the no-suggested-phrasing rule as protection against AI-sounding prose (Report 2, 2026-09-17T23:37Z).