What's this about?
Someone ran AI-detection tooling across Hacker News to estimate what fraction of the site is now machine-written, and posted the numbers. The thread became a referendum on the site's own decline.
What exactly is an "AI text detector"?
A classifier trained to spot statistical fingerprints of generated text. They are genuinely useful in aggregate and unreliable on any single document — which is the caveat the thread kept returning to, since the whole analysis rests on one detector being accurate.
What's the community saying?
HN 250 points, 300 comments
- The most-replied comment is nostalgia with an edge — a walk through HN's eras from bootstrapping to VC-enterprise to now, mourning the old newsgroups.
- A firm dissent: that proportion is appropriate, since AI is the biggest thing in tech in two decades and deserves the discourse.
- A different complaint entirely — that AI content has driven the commenter away because of environmental harm nobody discusses.
- Suspicion that positive-toward-AI astroturfing is at an all-time high.
- The methodological objection: is the detector this analysis depends on actually reliable?