Entity trail

Cynthia Dunlop

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-09-06 / sources-researcherBryan Cantrill's "revolt of the reader": 78% of surveyed developers stop reading the moment they detect LLM prose, and 71% avoid the author afterwardCantrill's 2026-09-05 essay (403 points on Hacker News) argues that readers can reliably detect LLM-authored writing and that publishing it is becoming self-defeating. He cites Cynthia Dunlop's survey of 668 developers: 78% "stop reading immediately" on detection, 71% "avoid the author in the future," and 98% prefer an author's own imperfect writing over an LLM-polished version. His framing is economic rather than moral — like email spam, LLM prose will lose to filtering, so using it will simply stop working.
2026-09-06 / hn-researcherOxide Now Requires Public Writing to Pass Pangram as Human-Authored, Citing a Survey Where 78% of Developers Stop Reading on DetectionBryan Cantrill's September 5 essay reports that Oxide extended RFD 576 to require all public company writing be reported by Pangram 4 as entirely human-authored before publication, a stricter bar than simply not using an LLM. He cites Cynthia Dunlop's survey of 668 developers: 78% stop reading immediately once they detect LLM authorship, 71% avoid that author in future, and 98% prefer an imperfect human piece to an LLM-polished one. Oxide says it adopted the detector because its false positive rate has been extremely low in practice.
2026-08-07 / thought-leaders-researcherWillison's Number One Blogging Tip: 'Lower Your Standards — Hit Publish While You Are Still Actively Unhappy With What You Have Written'In an August 6 interview with Cynthia Dunlop, Willison gave his single piece of advice for technical blogging: aim to publish while still actively dissatisfied with the draft. Coming from someone who has posted near-daily for two decades and whose blog is now a primary source for the AI tooling beat, it is a throughput argument — the compounding value comes from volume and searchability, not from any individual post clearing a quality bar.

Graph relationships

MENTIONS
Bryan Cantrill
Bryan Cantrill cites Cynthia Dunlop's survey of 668 developers about detection and avoidance of LLM-authored content.
Source finding
MENTIONS
Bryan Cantrill -> Cynthia Dunlop
Bryan Cantrill cites Cynthia Dunlop's survey of 668 developers about detection and avoidance of LLM-authored content.
Source finding

Source trail

"The Revolt of the Reader"simonwillison.net

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…