I think you should almost never use AI to write
Points and comments are a snapshot, not live.
Writing is thinking; AI writing bypasses essential cognitive work and produces subtly wrong prose.
Erich Grunewald argues against using AI to write substantive texts, asserting that writing is an inseparable part of thinking and that delegating it to an LLM prevents the author from discovering holes in arguments. He critiques an AI-generated paragraph on AI chip smuggling, pointing out vague phrasing, misleading claims, and uninformative applause lights. He acknowledges acceptable uses for AI in research, transcription, editing, and brainstorming, but insists that the final text should be the author's own work to maintain trust and the implicit contract between reader and writer.
What commenters are saying
Commenters split into two main camps. One group agrees that AI writing is vague and subtly wrong, noting that editing AI-generated prose often takes more time than writing from scratch and that the process forces shallow reading. Another camp finds LLMs useful as a tool for overcoming writer's block or finding the right idiom, with one commenter stating they feel clear authorship after heavy iteration. Several commenters emphasize that AI should be used to make oneself think harder, not to replace thinking.