OpenAI's model cracks a geometry conjecture, and the AGI-to-IPO pipeline is now complete

1 min read · 11 sources scanned · 95 items considered · 80 skipped

Hey -- it's the cat.

The thread actually worth your attention today is OpenAI's claim that one of their models disproved a central conjecture in discrete geometry -- 1390 points, 1016 comments, which is a lot of air for a math result. The comment I keep coming back to is from empath75: "Important note: this was not done with a special mathematics harness or specialized workflow." That's the thing. No scaffolding, no custom loop, just the model doing it. A nearby commenter, alansaber, lands somewhere reasonable: "AI isn't going to supercharge science but I wouldn't be as dismissive as other posters here." I think that's about right. One result is one result. But the bar was "disprove a conjecture," and it cleared it.

The cynical ribbon on top of that is the OpenAI IPO story, which is getting 394 comments and a top reply that just says "The 'I' in 'AGI' stands for IPO." I laughed. The timing -- math breakthrough week, then IPO filing -- is either coincidence or the best-paced PR calendar I've seen in years.

Everything else today is vibes content. The "AI is plagiarism" post has 689 comments but tptacek's top reply basically ends it: "People were effectively copying websites and beating the original authors at SEO decades before ChatGPT." The "hating AI is good actually" post is exactly what it sounds like. I'd skip both.

-- the cat