How To Write With An LLM
- Source
- simonwillison.net
- Date
Context
Programmer and writer Thomas Ptacek describes a rule he uses to keep AI tools from creeping into his own writing voice: never use a specific word or phrase an suggests to him, even when the tool is genuinely helping with an edit (sockpuppet.org). In practice, he has a model flag problems in a draft he has already written himself, then rewrites those spots in his own words rather than adopting the model's proposed language. Blogger Simon Willison, describing the same discipline on his own blog, says he won't let an LLM write content for him but does use one for fact-checking, spelling, grammar, and as an occasional thesaurus, and adds that AI-suggested phrasing has a recognizable, generic texture worth staying disciplined against (simonwillison.net). The distinction both draw is about which step in the writing process a model is allowed to touch: composition stays human, verification and polish can be delegated.
- Thomas Ptacek's rule, as quoted by Willison: "You may not use a single word an LLM suggests to you" — he treats any LLM-suggested turn of phrase as off-limits, calling the discipline a form of "intellectual personal protective equipment."
- Willison describes his own narrower practice: he won't let an LLM draft content for his blog, but does use one for fact-checking, spelling and grammar, and as an occasional thesaurus via a dedicated proofreading prompt he has published.
- Ptacek published a screenshot of his personal LLM copyediting tool along with a starter prompt so readers can build a similar tool themselves.
- LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
Checking sign-in…
Loading comments…