I think a good practice for frontier model developers, in pre-release testing, would be to spend $10M of compute on easily-verifiable “destabilizing” innovations — like polynomial factorization, for instance (which this isn’t, see thread)
Show more
636606729769440499166579950236036751749912014371509557713570027508971809534551913252252094954941974952859310861988904737359709200557919
is a factor of RSA-896
A big question in AI right now is whether math is more idiosyncratic or more of a leading indicator. Either way, AI is certainly remaking how math is done.
The share of math preprints on arXiv that acknowledge AI has risen rapidly, from 4% in April to 25% in August. This increase holds even when filtering to papers with at least one author who published regularly before 2023.
Show more
I have now bet @GregHBurnham at 1:3 odds that P vs NP or the Riemann Hypothesis will be solved within a year (EOD 2027-09-17).
Some discussion of this result — amusing and all, but also note the by-now familiar comment, "I originally thought it might require much more theory-building to solve the problem in full, and I had been working on it for quite some time."
Show more
Another problem from FrontierMath: Open Problems has been solved! The solution was elicited by Becker, Greger, and @DominikPeters in an interactive session with GPT-6 Astra. Peters originally suggested the problem for the benchmark. He had this to say.
Show more
My wife finds romanesco unsettling. I mentioned this to the 3yo once, and he’s been very curious to see “scary broccoli” ever since. Now it’s back in season and I’m sure one way or another it wont disappoint.
Show more
For Your Consideration: Everything in the world is about AI except AI. AI is about math.
@max_sixty Very nice, but I don't feel like this one has been cracked yet. Open Problems in humor... AI can't solve it either.
“Aquiline” is clearly an op by Big Nose, but good for them 🦅
Going to make my exclamations geology themed: gneiss and schist
@max_sixty Very nice, but I don't feel like this one has been cracked yet. Open Problems in humor... AI can't solve it either.
@max_sixty Very nice, but I don't feel like this one has been cracked yet. Open Problems in humor... AI can't solve it either.
Thoughts a day after Navier-Stokes.
The anti-hype take: mathematicians devised a promising strategy and the last mile turned out to be well-suited to AI.
The usual pro-hype take is “progress is fast” — true! — but there’s another angle on my mind…
Show more
Checking in on the benchmarks in the AI 2026 forecasting survey, halfway through the year...
I realized I'd never seen "# of Hilbert Problems Solved Over Time", so here's a cut thanks to GPT-5.5 and some guesses on my part. Feel free to correct specifics.
