Bryan Cantrill · 2026-09-05 · notable
Bryan Cantrill — readers quit a post the moment they spot AI writing
Bryan Cantrill argues that LLM-written prose repels the readers a writer most wants. He cites a survey of 668 developers by Cynthia Dunlop in which 78% stop reading as soon as they notice AI writing and 71% avoid that author afterwards.
An argument that AI-written prose fails not because it is wrong but because readers detect it and walk away.
What is it?
The revolt of the reader is Bryan Cantrill's essay on what happens after a reader notices that a post was written by a language model. He points at a survey of 668 developers run by Cynthia Dunlop: 78% say they stop reading immediately, 71% say they avoid that author in future, and 98% say they prefer an author's own imperfect writing to an LLM-polished version.
How does it work?
Cantrill frames writing as a contract in which the reader lends attention against an expectation of effort, so recognisable model prose reads as a broken promise rather than a style choice. He compares the resulting detection problem to email spam filtering and points at Pangram Labs' detector, arguing that the Pangram 4 model reaches low enough false-positive and false-negative rates to be used as a check rather than an accusation.
Why does it matter?
The practical proposal is a policy, not a ban on tools: Cantrill says organisations that care about their institutional voice should require public writing to come back from Pangram as human-authored, and reports that Oxide Computer does exactly that. That turns a vague cultural worry into something a team can actually write down, which is why the post drew 568 points and 277 comments on Hacker News.
Who is it for?
engineers and teams writing in public