Daniel Litt

I am also going only for this week (for similar reasons) :)

Heading to Fort Collins for the once-a-decade algebraic geometry summer research institute. Please say hi if you’re around! (And come to my talk tomorrow on “p-curvature and non-abelian cohomology”!)

I'm having trouble finding the preview feature. Has this been implemented yet?

I've been working on a new tool, Refine, to make scholars more productive. If you're interested in being among the very first to try the beta, please read on.

Refine leverages the best current AI models to draw your attention to potential errors and clarity issues in research paper drafts.

1/

right, I was thinking that would be the climax, with two versions (how Grothendieck saw it, and what I think is the more accurate version)

current plan (which I will probably never execute on) is to switch back and forth between his early and late career, like Oppenheimer. first is meeting Schwartz, then Vietnam, then probably some childhood stuff

obviously subject to change since I started writing it 20 minutes ago, but it's narrated by Serre, and it opens with him writing one of his last letters to Grothendieck

ok somebody stop me, the first page now exists

(the screenplay does not actually exist but the title does and I'm very proud of it)

at 10,000 followers I will reveal the title of my Grothendieck biopic screenplay

pancakes smothered in maple syrup

planning to use this picture in a talk to explain what a foliation is, feeling extremely Canadian

Oh, this looks nice!

This might be what I end up doing, I was fiddling a bit with it earlier today.

If you have a beautiful example you’d like to share I’d be interested in that too.

Giving a slide talk next week—thinking of trying something other than Beamer. Any recommendations for software for beautiful math talks?

(Sparked by a harder-than-average-to-catch crank paper I received this morning)

Yes, I’ve been debating writing a thread on this :-/

Getting a lot of heat for this on the other site.

I would guess the latter but as you say we don’t know. Re marking, yes, we should scrutinize the solutions carefully.

That was Google. OpenAI said it was within the 4.5 hr time limit.

GitHub - aw31/openai-imo-2025-proofs Contribute to aw31/openai-imo-2025-proofs development by creating an account on GitHub.

OpenAI has just announced that their experimental model achieved IMO gold, solving P1-5. Solutions here: github.com/aw31/openai-...

Leaving out the parts that people tend to skip, but for math papers.

Most writing is improved by simply removing the first sentence. And then the new first sentence, and so on.

Actually kind of curious about what caused the inversion here. Bernoulli effect?

it was a perfect summer day in Toronto. the sun was out; the subway station garbage bags danced upside-down in the wind

7 now!

I actually count 6(!) LLM-generated nonsense papers on the Hodge conjecture posted to arXiv since June 15, of 8 total papers that mention it in the title or abstract.

What a mess.

You'd hope so, but there are some repeat offenders.

I actually count 6(!) LLM-generated nonsense papers on the Hodge conjecture posted to arXiv since June 15, of 8 total papers that mention it in the title or abstract.