LLM Voice
Martin Fowler argues that relying on LLMs to draft prose can produce writing he finds difficult to engage with, and may displace the thinking that happens through writing. He also cautions that neither his intuition nor automated AI detectors reliably identify authorship.
He recognizes accessibility and translation as cases where generated prose can be valuable. HN readers discussed repetitive, low-density output; others pointed out that SEO-shaped writing predates LLMs, and one researcher argued that instruction tuning deserves more attention than pretraining alone.