Assistant professor (of mathematics) at the University of Toronto. "Tireless math ronin." Algebraic geometry, number theory, etc. He/him.
Joined August 2010
- Tweets 31,628
- Following 916
- Followers 57,842
- Likes 70,381
2,179 Photos and videos
Pinned Tweet
Jan 14
New paper with Josh Lam, about which I'm really excited! I want to try to briefly explain what the point is in this thread.
13
34
508
162,636
Jun 12
First half or so is meant to be fully accessible to people near the end of an undergrad math major, and the second half somewhat accessible to such. (Not sure if I actually accomplished this--I basically always feel like my talks were too hard after the fact--but I tried!)
Another highlight from the CTNT 2026 conference was @littmath
's talk on the "Arithmetic of Differential Equations". Really, really interesting and a *great talk*. Thanks Daniel for joining us!!
youtu.be/B0E0q6ICK3c?is=hgXZ…
4
5
155
16,884
Daniel Litt retweeted
Jun 12
FrontierMath: Tiers 1–4 (v2) is live.
We concluded an audit that addressed errors in 42% of problems. Rankings are similar but scores are higher across the board. The current leaders are GPT-5.5 (xhigh) with 85% on Tiers 1–3 and Google’s AI co-mathematician with 76% on Tier 4.
27
66
575
113,725
Daniel Litt retweeted
Another highlight from the CTNT 2026 conference was @littmath
's talk on the "Arithmetic of Differential Equations". Really, really interesting and a *great talk*. Thanks Daniel for joining us!!
youtu.be/B0E0q6ICK3c?is=hgXZ…
4
77
28,589
Daniel Litt retweeted
Jun 11
Interesting! There are improvements in certain directions: the best out-of-the-box model (GPT 5.5 Pro) got essentially 4/10 correct versus 2/10 last time, and Submission A should have gotten 7/10 except for some API error (see comments on P6) versus 5ish/10 for the best harness last time. For individual performances this is roughly in line with what I expected.
Collectively, the performance was no better than last time -- this falls well under the threshold that I said would be "disappointing for AI". There were far fewer teams than I thought there would be, though. In the planning stage I personally heard about more parties planning to participate than showed up in the end. Would be great if FirstProof could find a way to facilitate more participation without comprising the standards of transparency.
Important takeaway for the masses: math is far from "done"!
3
6
48
5,919
Daniel Litt retweeted
Jun 10
Contrary to the claim in the video, incompleteness phenomenon does come up in practice, unless you are specifically restricting yourself to Gödel-type sentences.
Whitehead's problem, Suslin's problem, Borel's conjecture, normal Moore space conjecture, Kaplansky's conjecture, existence of outer automorphisms of Calkin algebra,... These problems were asked naturally in their respective fields and they are all independent of ZFC. So these are instances of incompleteness.
I am not claiming that problems mathematicians work on frequently end up being independent, but they sometimes do. Percentagewise, such problems might be rare but the claim that incompleteness is irrelevant to mathematical pratice is simply not true.
4
13
126
8,931
Jun 10
Results from 1st proof, second batch are out! Perhaps surprisingly, no improvement over the first batch (though note that internal models were not tested), despite taking place several months later.
Jun 10
results from the second batch of #1stproof will be posted tomorrow (june 10) at 1stproof.org, and presented by the editorial board at 1pm ET here: cmsa.fas.harvard.edu/first-p…
12
14
160
21,395
Jun 10
I haven't yet had a chance to get a sense of how difficult these problems are, though the report (here: 1stproof.org/assets/docs/rep…) contains preregistered difficulty statements. See e.g. here for the statement on problem 8, which wasn't solved.
1
26
4,311
Jun 10
A bit interesting also that Submission D (a scaffold for Gemini 3.1) seems to have underperformed GPT 5.5 Pro "out of the box," in many cases at ~10x the cost... Submission A, which performed best, spent between $36 and ~$950 per problem.
1
25
3,633
Jun 10
IMO the common-sense view is that you should use AI tools if, on balance, they improve your outputs--e.g. for many coding tasks, if like me you're not a great coder--and you shouldn't use them if they harm your outputs (for example, for writing if you're a competent writer).
19
7
199
15,867
Jun 10
Obviously as capabilities improve some tasks move from the second bucket to the first, and exactly when that happens depends a lot on one's own distribution of skills, etc.
2
1
28
2,628
Daniel Litt retweeted
29 Dec 2020
i find the unreasonable effectiveness of mathematics in the natural sciences pretty suspicious, frankly
35
63
974
Daniel Litt retweeted
I recently had the chance to record a convo with @littmath about math and AI. It was great hearing a little about his workflow, some current best practices, and more. Excited to try out some new things myself, now!
1
2
26
3,892
Jun 7
Just saw a paper on here by an apparently reputable scientist which appears to be entirely AI-generated, with the tweet announcing it also AI-generated. According to @pangram, his responses to substantive comments are also 100% AI-generated. What are we doing here?
45
10
483
59,432
Jun 8
Insane to me that people in the replies are defending the act of turning oneself into an appendage.
3
1
39
3,499
Daniel Litt retweeted
Look who stopped by UConn to give an amazing talk at CTNT and to sit down with me to record an interview about Math and AI! Thanks @littmath for a really fun visit! Stay tuned for a deluge of CTNT videos and clips from the interview!
4
3
77
5,306
Jun 6
My general view has always been that it's not worth rushing to be the first to do work that someone else could do next week; clearly the marginal impact of such work is low. Worth thinking about this in the rush to publish results where the intellectual labor was done by AI.
18
36
414
24,644
Daniel Litt retweeted
Jun 6
I'm extremely bullish on there being a place where the young and interested will spend time, goofing off and learning and experimenting, helped along by those older and often wiser, in the pursuit of knowledge. This might well be a University.
Jun 6
This is what ultimately makes me bullish about higher ed in the long run. These statements from leaders of large schools, the Yale report, Berkeley CS failing sub-par performance, etc are all signs of a badly needed course correction.
These are early days, but I hope that the correction continues, and that US higher ed emerges with a renewed sense of purpose, value, and dedication to the students they are educating.
3
9
123
14,116
A lot of AI discussions here are people gesturing at how good/bad AI will get grand mystical pronouncements. It's sooo boring. The eternal it's so over vs we're so back fluctuations via predictable low perplexity takes. Kinda tired of Temu prophets, I want grass-toucher tales
22
12
188
17,122
Daniel Litt retweeted
Avi is not only an amazing scientist but also a wise and great person. Worth watching this interview
Jun 1
Avi Wigderson is the only person in history to have won both a Turing Award (computer science) and Abel Prize (math). I interviewed him all about his field. We discussed:
• His intuition on a proof of P vs NP
• Why we use SAT solvers for most NP problems
• Zero knowledge proofs and their impact
• Quantum computation and implications
• Math and computer science's relationship
Where to watch:
• YouTube: youtu.be/5GUcvSAJcJw
• Spotify: open.spotify.com/episode/4JZ…
• Apple Podcasts: podcasts.apple.com/us/podcas…
• Transcript: developing.dev/p/turing-awar…
Thank you to this episode's sponsors for supporting my work:
• WorkOS: makes your app Enterprise Ready with easy to use APIs to add SSO, SCIM, RBAC, and more in just a few lines of code, check them out at workos.com/
Timestamps:
00:00 - Intro
01:08 - P vs NP
14:51 - What if you relaxed correctness
25:38 - Why NP complete problems are equivalent
30:33 - Space vs time complexity
43:06 - Why people use SAT solvers
45:53 - Randomness is a resource
55:48 - Randomness depends on computational power
01:21:20 - Zero knowledge proofs and their significance
01:38:30 - Quantum computation and why it matters
01:56:24 - Math vs computer science
02:08:16 - Major breakthroughs and his experience
02:12:31 - Advice for his younger self
02:14:48 - Outro
2
15
177
29,936