Post

Now mathematicians are gonna have to deal with the same issue that people reviewing vibecoded slop are currently dealing with. How many of these "proofs" are utter nonsense that will consume dozens if not hundreds of hours to verify?

Zach Weinersmith@zachweinersmith.bsky.social

OpenAI just blooped hundreds of proofs and counterexamples to open problems: github.com/openai/math/... Boggling.

Without an actual mathematician looking at these before OpenAI shits them out there's no real guarantee that any, much less all of them, are better than crank results. Reminder that OpenAI's Navier-Stokes claims have yet to be formally verified and will take months to do so. And uh, oops.

If this gets to a point where people start taking these things as true and building on them for later math we'll start having endlessly buggy math the same way vibecoded software is endlessly buggy, but worse. The moment someone finds an error in something there's a cascading failure.

0 reposts0 likes

This could legitimately taint entire branches of mathematics if not reviewed with more rigor than is likely possible at the rate OpenAI seems to be churning these out. Yet another realm the flood of slop is pouring into.