LessWrong (Curated & Popular) "Don’t Let LLMs Write For You" by JustisMills
7 snips
Mar 12, 2026 A critique of AI-written prose and why attentive readers can tell when text smells like an LLM. Discussion of how polished machine text can mask shallow thinking and fabricated citations. Lists of subtle stylistic giveaways to watch for. Advice on using AI as a tool for sourcing, then verifying and rewriting to reclaim human judgment.
AI Snips
Chapters
Transcript
Episode notes
AI Prose Breaks The Link Between Clarity And Thought
- Human readers can often detect AI-generated prose and it reduces trust in the writer.
- Justice Mills says clear human writing signals refined thought, while LLM prose can make weak ideas look polished without real understanding.
Polished Prose Can Fake Rigor
- LLMs can make vague or poor ideas appear authoritative by producing polished prose and manufactured citations.
- Mills notes Claude will find D-tier writeups and invent hedges and caveats that give the illusion of rigor.
Style Tells Reveal AI Authorship
- AI-generated writing often contains unnecessary framing, lists, and flashy formatting that slow readers down.
- Mills describes stylistic telltales like frequent lists, bold headers, and contrastive disclaimers that make reading a slog.