There are no lossless transformations of natural-language text

Per Simon Willison's Weblog, engineer Sophie Alpert has established a policy on using AI to help with writing: you must stand behind every sentence in your documentation. If someone questions a line, you can't say 'the AI wrote that, ignore it'—that approach confuses readers and wastes their time. Alpert's reasoning is that there are no truly lossless transformations of natural-language text. Every rewrite changes meaning slightly, and when an AI does the rewriting, it lacks the writer's full context, so information gets lost. The upshot: if you're using an LLM to help with docs, you fully own the result.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton