I built an AI that writes like me, and I'm measuring how much it fails
Everyone can now spot machine text on sight, the em-dashes, the "you're absolutely right", the invented technical terms nobody at any job ever used. I write a lot with agents, and I did not want my published words to smell like that. So I built hyphos , a small open tool that learns how I actually write, from my own words, and rewrites AI drafts to sound like me. This is the build log, including…
The author built an AI tool called hyphos to rewrite AI-generated drafts to mimic their own writing style. They collected data from their own coding agent conversations and email correspondence over a decade, extracting about 1.27 million words and filtering out lengthy, machine-like text. After processing, they were left with 80,470 words that resembled their actual typing style.
Hyphos uses a two-step process: first, a model rewrites AI drafts, then a set of rules removes any fingerprint-like elements, such as em-dashes or unusual word usage. The tool scores the output based on how much it resembles the author's writing style and counts any remaining AI-isms. In testing, the tool achieved a six-twelve accuracy rate, indicating it sometimes fails to distinguish its own writing from AI-generated text. Hyphos is open-source and runs locally, ensuring user privacy.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
