Daniel Litt
@littmath
Followers
49K
Following
59K
Media
2K
Statuses
27K
Assistant professor (of mathematics) at the University of Toronto. Algebraic geometry, number theory, forever distracted and confused, etc. He/him.
Toronto, Ontario
Joined August 2010
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
18
106
926
I think the kind of claim about writing in the OP would be much stronger if accompanied by an exemplar. What's a great math paper that doesn't use the theorem/lemma/proof style? I can think of tons of fantastic exposition that does.
3
0
12
@gregfloyd42069 My experience (both writing and reading) is that when one tries to communicate hard-won mathematical intuitions directly in text it ends up too vague/imprecise to be useful. The reason reading math papers is hard work isn't the style, it's that developing intuition is hard.
4
0
12
It is indeed a bad, hard-to-read style. Its one redeeming feature is that it's better than anything else that's been tried.
The desiccated "Theorem, Lemma, Proof, Corollary,. " presentational style is staggeringly counterproductive, if one's objective is actually communicating the underlying mathematical intuitions and thought processes behind a result. In reality, the process is more like. (1/4).
12
4
88
I think you should read self-reports of how much AI tools increase productivity with a LOT of skepticism.
We ran a randomized controlled trial to see how much AI coding tools speed up experienced open-source developers. The results surprised us: Developers thought they were 20% faster with AI tools, but they were actually 19% slower when they had access to AI than when they didn't.
7
4
123
you're telling me this firmware update to my Neuralink device will make me maximally truth-seeking? sign me up!.
11
36
498
since there are already a few illiterate responses to this tweet, let me explain it: the quote “this is the worst it will ever be” is about capabilities, but I’m talking about moral compass.
4
0
32
We’re often told that “this is the worst AI will ever be” but tbh I’m not convinced. It can get worse!.
"We updated Grok to make it less woke."."Did you make it 'less woke' or did you make it seethingly hate Jews?"."It's a good model, sir.". (They made it seethingly hate Jews.)
12
20
383
Claude: my wife and I went antique shopping this weekend.Gemini: if I can’t get this code to work I will k*** myself.ChatGPT: the answer to your question came to me in a dream.Grok: why yes I was in Berlin in 1939, why do you ask?.
31
207
4K
Interesting that one sees comments like this a month or two after the release of every new model—I recall similar claims about o3 and Gemini. I think it’s just the honeymoon period wearing off and weaknesses becoming apparent with use.
I'm pretty sure Claude 4 got nerfed.
21
13
339
