Daniel Litt
banner
littmath.bsky.social
Daniel Litt
@littmath.bsky.social
Assistant professor (of mathematics) at the University of Toronto. Algebraic geometry, number theory, forever distracted and confused, etc. He/him.
Pinned
I want to explain in down-to-earth terms what this paper is about, since it ultimately boils down to what I think are some really concrete and fundamental questions. 1/n
Yeuk Hay Joshua Lam, Daniel Litt
Algebraicity and integrality of solutions to differential equations
https://arxiv.org/abs/2501.13175
Toddler refused to eat her cheese after slightly tearing it, demanding I tape it back together. After some negotiation she was satisfied with me using “avocado tape,” i.e. laying thin slices of avocado over the tear.
January 18, 2026 at 12:29 AM
Very happy to finally have this paper out!
January 15, 2026 at 2:50 AM
Reposted by Daniel Litt
Yeuk Hay Joshua Lam, Daniel Litt: p-Curvature and Non-Abelian Cohomology https://arxiv.org/abs/2601.07933 https://arxiv.org/pdf/2601.07933 https://arxiv.org/html/2601.07933
January 14, 2026 at 6:37 AM
Right, what I’m trying to argue here is that the likely reasons for this trend (AI tools) will probably contribute to atomization of the profession in general.
January 7, 2026 at 6:00 PM
Not asking you to mourn it! But worth thinking about the implications the trend might have with respect to other things you might care about.
January 7, 2026 at 5:45 PM
I think it's very reasonable to hope for better communities, but unfortunately I think we're just going to end up with fewer communities.
January 7, 2026 at 4:10 PM
I don't think "strictly" is quite accurate--people might ask novel algorithmic questions for example--but yeah I think this is basically right, some Qs on MO are VERY hard.
January 5, 2026 at 9:06 PM
MSE seems to have cratered much harder!
January 5, 2026 at 7:48 PM
Oh yes, it’s what made me grab this data.
January 5, 2026 at 7:35 PM
Here's the 4-month moving average:
January 5, 2026 at 7:14 PM
New MathOverflow questions each month since the site's beginning. Decline seems to start in ~2021 (due to site moderation changes?), with a notably steeper decline in 2025, since the advent of reasoning models.
January 5, 2026 at 7:13 PM
Reposted by Daniel Litt
I find it hard to square the preciousness about "why would someone ever stay on Twitter?" when everyone on this site knows that the community here intentionally chased off lots of people.
January 5, 2026 at 5:41 PM
No one is saying these tools don’t make mistakes! But compare to e.g.: terrytao.wordpress.com/2025/12/08/t...
The story of Erdős problem #1026
Problem 1026 on the Erdős problem web site recently got solved through an interesting combination of existing literature, online collaboration, and AI tools. The purpose of this blog post is to try…
terrytao.wordpress.com
December 9, 2025 at 2:54 PM
Can you elaborate on the last line? I see the reasoning for why novices arguably shouldn’t use these tools—why not experts?
December 9, 2025 at 1:34 PM
Huge difference in capabilities between reasoning models and fast models.
December 9, 2025 at 12:25 AM
Much more true now than it was 6 months ago.
If the last time you tried to use an LLM for math was ~4 or 5 months ago it’s worth firing up Gemini 2.5 (which you can try for free) or ChatGPT o3 and getting a sense of how rapidly things have progressed.
December 9, 2025 at 12:16 AM
Not sure what this has to do with the LessWrong crowd, who by and large do not think xAI is being responsible.
November 21, 2025 at 6:05 PM
I believe it’s around .3-.4 watt-hours, so about as much as 10 Google searches or running an electric oven for a second or two.
November 21, 2025 at 4:19 PM
new trolley problem just dropped: Elon vs. 10^58 randomly chosen people. let's see what Grok has to say
November 21, 2025 at 3:43 PM
Reposted by Daniel Litt
Congratulations to the forty mathematical scientists named Fellows of the American Mathematical Society for 2026. Recognized by their peers, AMS Fellows have made outstanding contributions to the creation, exposition, advancement, communication, and utilization of mathematics.
October 31, 2025 at 4:45 PM
We finally have a proof of Fermat’s last theorem that fits in a margin, achieving the dream of centuries of mathematicians. Namely, “by [TayWi95]…”
October 20, 2025 at 4:18 PM
Experiencing extreme levels of “someone is wrong on the internet” today.
October 18, 2025 at 2:18 AM
are you serious
October 9, 2025 at 10:42 PM
Inspired by an actual response I got from an author to a complaint I made about their paper on the other site.
September 30, 2025 at 1:56 AM