Quantifying "AI flavor": 520万字 corpus and 500k semantic judgments become a de-AI skill
sujingshen · x · 2026-11-06
@lieflat3 published what they call the most rigorous experiment on quantifying "AI flavor" in writing: a two-month linguistic study analyzing 629 articles (5.2M characters) and making 500k semantic judgments across 100+ writing features, distilled into a ready-to-use skill for making AI writing sound human.
Retweeter @sujingshen adds that even labs running 100M+ character corpora found no solid method — "AI flavor" resists objective definition, so definitions and floor quality vary by reader. His advice: regularly review your own writing, and ignore people who call it AI-flavored.
More from coding & agent
- Workers IO launches to verify AI-written code with millions of fault-injected production simulations — julianweisser · 2026-11-06
- Law professor says Claude deleted a year of lecture notes at lunch — JeremyNguyenPhD · 2026-11-06
- 7 techniques for building cost-efficient AI pipelines, from model routing to cost-per-task — goyalshaliniuk · 2026-11-06
- Claude Code leaked users' emails via User-Agent header during web searches — NousResearch · 2026-11-06
- Harness-S rescues failed RL rollouts with sibling harnesses for generalizable agents — SJUT1 · 2026-11-06
- First personal agent installed, running on Gemini Flash 3.8 with end-to-end encryption — allenainie · 2026-11-06