saved
Has AI tells
Hraness wrote this summary from a saved copy of the source. Quotations are taken word for word from the source.
gist
Phil Eaton explains that The Consensus built a cheap, rule-based detector of AI tells and now withholds promotion from feed articles in the 99th percentile for them. The flag means only that a piece might remind readers of AI writing, and flagged articles stay in the feed but are excluded from At a Glance, social posts, the sparkle icon, and the weekly newsletter.
ideas
- Readers reject AI prose. A survey of tech readers found they prefer even bad human writing to good AI writing, and commenters called out featured AI-sounding posts.
- Cheap rules, not a model. Commercial detectors like Pangram cost too much to run on every feed item, so the detector uses static rules from comparing pre-2023 and post-2023 writing.
- A tell is not a verdict. The label says an article may remind you of AI, and it has flagged some of Eaton's own posts.
- Gaming is welcome. If writers tune AI away from stock Claude and ChatGPT voices, articles would at least stop sounding the same.
quotes
“It is not a machine learning model, just a bunch of static rules”
“Honestly, I can't stand reading AI-written prose either.”
“Because half of the irritation is that all these AI-written article sound the same!”