#  @littmath Daniel Litt Daniel Litt posts on X about math, ai, in the, wrong the most. They currently have [------] followers and [---] posts still getting attention that total [------] engagements in the last [--] hours. ### Engagements: [------] [#](/creator/twitter::177416255/interactions)  - [--] Week [-------] +31% - [--] Month [---------] -66% - [--] Months [---------] -11% - [--] Year [----------] -43% ### Mentions: [--] [#](/creator/twitter::177416255/posts_active)  - [--] Week [--] +65% - [--] Month [---] -26% - [--] Months [---] +80% - [--] Year [---] +52% ### Followers: [------] [#](/creator/twitter::177416255/followers)  - [--] Week [------] +0.27% - [--] Month [------] +1% - [--] Months [------] +10% - [--] Year [------] +20% ### CreatorRank: [-------] [#](/creator/twitter::177416255/influencer_rank)  ### Social Influence **Social category influence** [technology brands](/list/technology-brands) 5.17% [travel destinations](/list/travel-destinations) 0.86% [currencies](/list/currencies) 0.86% [stocks](/list/stocks) 0.86% [finance](/list/finance) 0.86% [social networks](/list/social-networks) 0.86% **Social topic influence** [math](/topic/math) #2487, [ai](/topic/ai) 9.48%, [in the](/topic/in-the) 6.9%, [wrong](/topic/wrong) #758, [how to](/topic/how-to) 4.31%, [this is](/topic/this-is) 3.45%, [future](/topic/future) 2.59%, [the first](/topic/the-first) 2.59%, [model](/topic/model) 2.59%, [theory](/topic/theory) #717 **Top accounts mentioned or mentioned by** [@sheafnat](/creator/undefined) [@andilesanthony](/creator/undefined) [@xixidu](/creator/undefined) [@bengolub](/creator/undefined) [@jeffconcerto](/creator/undefined) [@afinetheorem](/creator/undefined) [@quantumgeoff](/creator/undefined) [@sar1287](/creator/undefined) [@jasondeanlee](/creator/undefined) [@sosowski](/creator/undefined) [@ognifedefingo](/creator/undefined) [@profnoahgian](/creator/undefined) [@alzzyd](/creator/undefined) [@oydeis](/creator/undefined) [@acerfur](/creator/undefined) [@spoonedher](/creator/undefined) [@fleetingbits](/creator/undefined) [@prfsanjeevarora](/creator/undefined) [@lptomov82](/creator/undefined) [@jakobzupanec](/creator/undefined) **Top assets mentioned** [Alphabet Inc Class A (GOOGL)](/topic/$googl) ### Top Social Posts Top posts by engagements in the last [--] hours "@sheafnat @sar1287 @SwallieC69635 Wonderful Please keep me posted if you do" [X Link](https://x.com/littmath/status/2022695604191887389) 2026-02-14T15:33Z 54.9K followers, [---] engagements "This kind of thing is so useless. If the words in this sentence have their standard meaning its already been falsified (though one can debate the significance of the discoveries). Of course its more likely the words are defined so as to make the claim contentless. AI cannot in principle make novel discoveries. AI cannot in principle make novel discoveries" [X Link](https://x.com/littmath/status/2022038950387556393) 2026-02-12T20:03Z 54.9K followers, 37.9K engagements "Those are the ones I can semi-quickly evaluate. If an expert has any thoughts on the proposed solution to [--] Id be very curious" [X Link](https://x.com/littmath/status/2022592765175980461) 2026-02-14T08:44Z 54.9K followers, [----] engagements "Glad to see this clarified. I think it would be good to wait a bit to see how the rest of the claimed solutions shake out; tbc I do not have any strong opinion about any I have not already discussed on here. Would also like to hear more about how solutions were generated Based on the official #1stProof commentary community analysis and more clarification with external experts we now believe the solution to problem [--] above is likely incorrect. Grateful for the engagement and looking forward to continued review Based on the official #1stProof commentary community analysis and more clarification" [X Link](https://x.com/littmath/status/2022831537121542454) 2026-02-15T00:33Z 54.9K followers, 13K engagements "@JeffConcerto @ben_golub Impressive IMO (and I think at least [--] is pretty likely) but hard to be more precise without having some more insight into how they were generated. I tweeted at some point at the start of the week that my weak expectation was 2-3 but 4-5 wouldnt shock me" [X Link](https://x.com/littmath/status/2022810220917866559) 2026-02-14T23:08Z 54.9K followers, [----] engagements "FWIW I think ideally OAI employees etc. would also urge caution interpreting these claims until theyve been checked" [X Link](https://x.com/littmath/status/2022835085448257694) 2026-02-15T00:47Z 54.9K followers, [----] engagements "@XiXiDu This is a question of motivation not capability" [X Link](https://x.com/littmath/status/2022998841956732962) 2026-02-15T11:38Z 54.9K followers, [----] engagements "@ProfNoahGian @XiXiDu They are actually not going to verify the answers to round 1; this was supposed to be a proof of concept" [X Link](https://x.com/littmath/status/2023017096310141282) 2026-02-15T12:50Z 54.9K followers, [---] engagements "toddler (on the way to preschool): I can see the CN tower me: it was too cloudy to see it yesterday. isnt it nice we can see the whole thing today toddler *patiently*: daddy we can only see the front of it" [X Link](https://x.com/littmath/status/2021598587931517192) 2026-02-11T14:53Z 54.9K followers, 25.4K engagements "as a mathematician parent I have never been prouder" [X Link](https://x.com/littmath/status/2021600735436144912) 2026-02-11T15:02Z 54.9K followers, [----] engagements "Beautiful talk by Barry Mazur on the BSD conjecture from a few days ago here: Good example of a quite early major impact of computing on mathematics research. https://www.youtube.com/watchv=14-9iCoclFE https://www.youtube.com/watchv=14-9iCoclFE" [X Link](https://x.com/littmath/status/2022379130667569195) 2026-02-13T18:35Z 54.9K followers, [----] engagements "@alz_zyd_ who hurt you" [X Link](https://x.com/littmath/status/2022554510602670141) 2026-02-14T06:12Z 54.9K followers, [----] engagements "hmmmmmmmmmmmm" [X Link](https://x.com/littmath/status/2022587535697023271) 2026-02-14T08:23Z 54.9K followers, 10.9K engagements "I think the solution to Problem [--] is wrong" [X Link](https://x.com/littmath/status/2022588365242040329) 2026-02-14T08:26Z 54.9K followers, 50.2K engagements "AI enthusiasts interested in "accelerating science" feel free to amplify. Arguably correct solutions are more interesting but I think being confidently wrong is also an interesting signal and it would be good to have confirmation" [X Link](https://x.com/littmath/status/2022719819448021107) 2026-02-14T17:09Z 54.9K followers, [----] engagements "@jasondeanlee Re: autoformalization companies I am only aware of one credible Lean verification (of problem 9) which seems to have involved a LOT of human labor. Tbf I think most of these problems have statements that would be very challenging to formalize given the current state of Mathlib" [X Link](https://x.com/littmath/status/2022743994166579283) 2026-02-14T18:45Z 54.9K followers, [----] engagements "I think no. Ive pinged some experts about [--] (fwiw Im fairly confident its wrong) but havent heard back yet. AFAIK nothing is really checked; the only other things Im confident about is [--] is essentially right and [--] is very wrong. Solution to [--] seems to be credible but ofc I cant vouch for it myself. https://twitter.com/i/web/status/2022780672948199503 https://twitter.com/i/web/status/2022780672948199503" [X Link](https://x.com/littmath/status/2022780672948199503) 2026-02-14T21:11Z 54.9K followers, 41K engagements "@ben_golub Based on author comments Id be surprised if [--] is wrong and I wouldnt be surprised if [--] is right" [X Link](https://x.com/littmath/status/2022780868431901079) 2026-02-14T21:11Z 54.9K followers, [----] engagements "Really good question (note that DeepMind shared transcripts in their recent Aletheia paper and I think this is clearly best practice). Hopefully OAI follows suit. @merettm as an organizer of #1stProof this is truly exciting jakub would you consider sharing your transcripts @merettm as an organizer of #1stProof this is truly exciting jakub would you consider sharing your transcripts" [X Link](https://x.com/littmath/status/2022842585958482048) 2026-02-15T01:17Z 54.9K followers, 22.7K engagements "No need to amplify further OAI confirms its likely incorrect" [X Link](https://x.com/littmath/status/2022847039516627315) 2026-02-15T01:34Z 54.9K followers, [----] engagements "@quantum_geoff @JeffConcerto @ben_golub I mean e.g. experts identify promising strategies: https://x.com/merettm/status/2022925024798372262s=20 @Knikct Hi Nikhil We will aim to publish more information next week but as I noted above this was a quite chaotic sprint (you caught us by surprise please give us time to prepare next time). We will not be able to gather all the transcripts as they are quite scattered. Some of the https://x.com/merettm/status/2022925024798372262s=20 @Knikct Hi Nikhil We will aim to publish more information next week but as I noted above this was a quite chaotic sprint" [X Link](https://x.com/littmath/status/2022993759710187719) 2026-02-15T11:17Z 54.9K followers, [---] engagements "@XiXiDu I was responding to your claim about Batson whom I have no doubt could check this proof without learning to build a jet engine. He just doesnt care to" [X Link](https://x.com/littmath/status/2023009505152520442) 2026-02-15T12:20Z 54.9K followers, [---] engagements "@quantum_geoff @JeffConcerto @ben_golub .@jasondeanlee @damekdavis would be curious to get your thoughts on how to interpret this info from Pachocki" [X Link](https://x.com/littmath/status/2023021137127686201) 2026-02-15T13:06Z 54.9K followers, [---] engagements "@alz_zyd_ Basically endorse this" [X Link](https://x.com/littmath/status/2023098958013403171) 2026-02-15T18:15Z 54.9K followers, [----] engagements "The US National Science Foundation spends about $250 million (yes thats million with an m) funding math research (pure and applied) per year. ROI is actually incredible" [X Link](https://x.com/littmath/status/1897295005422911686) 2025-03-05T14:35Z 54.9K followers, 254.7K engagements "new AI tool that lets me enter The Last Supper walk around and explore it and find a Super Mario [--] star glowing and spinning under the table" [X Link](https://x.com/littmath/status/1954620007687581877) 2025-08-10T19:04Z 54.9K followers, 72.1K engagements "Sort of funny: published conjectures that are at the level where they can be resolved autonomously by AI are a non-renewable resource since in the future well just ask an AI rather than publishing the conjecture" [X Link](https://x.com/littmath/status/2017415179995123895) 2026-01-31T01:50Z 54.8K followers, 21.7K engagements "Lot of chatter on the timeline about Sonnet 5; all I have to say is: "But flowers distilld though they with winter meet Leese but their show; their substance still lives sweet."" [X Link](https://x.com/littmath/status/2018137644723482981) 2026-02-02T01:41Z 54.7K followers, 23.4K engagements "Really nice paper. Aside from the nice results resolving a number of open Erdos problems (or finding solutions in the literature) the paper does a really good job contextualizing significance and prior work. Okay looks like I can now talk about Aletheia on the Erds Problems https://t.co/JqpyUJUIcV Okay looks like I can now talk about Aletheia on the Erds Problems https://t.co/JqpyUJUIcV" [X Link](https://x.com/littmath/status/2018348971932909955) 2026-02-02T15:41Z 54.8K followers, 15.6K engagements "@Afinetheorem FWIW I do not think its very hard to generate questions where frontier models still substantially underperform humans. I was doing some trip planning with [---] Thinking the other day and it was basically unable to correctly answer Qs of the form list [--] hotels satisfying XYZ" [X Link](https://x.com/littmath/status/2019050129248920045) 2026-02-04T14:07Z 54.8K followers, [---] engagements "@Afinetheorem Here XYZ were desiderata like less than [--] minute walk to a subway availability on July [--] with a suite that costs at most $Z" [X Link](https://x.com/littmath/status/2019050389669109813) 2026-02-04T14:08Z 54.8K followers, [---] engagements "That we can now automate some mathematics that previously required an expert is a huge deal. That said the mathematics produced thus far is (in my obviously very subjective opinion) not notable in itself but rather because it is automated and as a leading indicator" [X Link](https://x.com/littmath/status/2019480079382814892) 2026-02-05T18:35Z 54.8K followers, 56.2K engagements "@quantum_geoff Curve of best fit f(x)=1+x. Oops no its 1+x+x2/2. Oops no its 1+x+x2/2+x3/6. Oops no its 1+x+x2/2+x3/6+x4/24. Oops no its" [X Link](https://x.com/littmath/status/2019777001620430980) 2026-02-06T14:15Z 54.8K followers, [----] engagements "This is a really nice test suite. Martin Hairer and colleagues released a set of hard maths problems designed to be test cases for LLMs. We have *one week* to solve them using LLMs. They encrypted the solutions at https://t.co/EZNjVzFT9t and will reveal them just after. https://t.co/20TaPDaSf2 (1/3) Martin Hairer and colleagues released a set of hard maths problems designed to be test cases for LLMs. We have *one week* to solve them using LLMs. They encrypted the solutions at https://t.co/EZNjVzFT9t and will reveal them just after. https://t.co/20TaPDaSf2 (1/3)" [X Link](https://x.com/littmath/status/2019795774779908385) 2026-02-06T15:30Z 54.8K followers, 27.9K engagements "Playing around with Opus [---] (extended thinking) for math. Seems to be plausibly the first Anthropic model to be useful for my purposes; it answered correctly (though non-rigorously) a pretty tricky algebraic geometry question that no other model has gotten" [X Link](https://x.com/littmath/status/2019956353520005527) 2026-02-07T02:08Z 54.8K followers, 28.3K engagements "TBC "useful" here means e.g. "it can sometimes perform routine computations I need" which is comparable with other frontier models" [X Link](https://x.com/littmath/status/2019960931879620830) 2026-02-07T02:26Z 54.8K followers, [----] engagements "Interestingly fails another one of my test questions in the exact same way [---] Pro does; both models consistently seem to (incorrectly) cite the same result" [X Link](https://x.com/littmath/status/2019962739968372909) 2026-02-07T02:33Z 54.8K followers, [----] engagements "Opus says "that's a beautiful question" a bit too much for my taste (though of course my questions are in fact beautiful). I prefer GPT 5.2's barely-concealed contempt" [X Link](https://x.com/littmath/status/2019972061880602648) 2026-02-07T03:10Z 54.8K followers, [----] engagements "@XiXiDu @leifweatherby Ive been playing around with the models for a long time since GPT-3. I guess I started doing so seriously when o1 came out which was the first time there were clear signals of usefulness for math imo. Obviously improvement since GPT-4 has been remarkably fast" [X Link](https://x.com/littmath/status/2020127151224160298) 2026-02-07T13:26Z 54.8K followers, [--] engagements "@oydeis Eh in math almost everything is on arxiv" [X Link](https://x.com/littmath/status/2020236573267181687) 2026-02-07T20:41Z 54.7K followers, [---] engagements "Some seem to be reading this interview as predicting that AI models will not continue to rapidly improve at math. I think this is wrongas I read it the interviewees are declining to make a prediction and are instead focused on accurately evaluating existing models. I think the mathematicians interviewed in this article are well-calibrated with respect to the capabilities of publicly available frontier models for math research. https://t.co/jUMkaqRa43 I think the mathematicians interviewed in this article are well-calibrated with respect to the capabilities of publicly available frontier" [X Link](https://x.com/littmath/status/2020516350892872104) 2026-02-08T15:13Z 54.8K followers, 10.1K engagements "@Miles_Brundage I think all the AI-for-math companies have done this Google just released a paper about one of their internal scaffolds (Alethia) and there are a number of individuals who have built scaffolds using various model APIs. This doesn't count" [X Link](https://x.com/littmath/status/2020596660150382633) 2026-02-08T20:32Z 54.8K followers, [----] engagements "me: *reading The Lorax* NOWthanks to your hacking my trees to the ground there's not enough Truffula Fruit to go 'round. And my poor Bar-ba-loots are all getting the crummies because they have gas and no food in their tummies toddler: They should get sushi" [X Link](https://x.com/littmath/status/2020663351878058449) 2026-02-09T00:57Z 54.8K followers, [----] engagements "@AcerFur I really think you should mention you independently came up with the same argument for [----] prior to the model. Why give the model all the credit" [X Link](https://x.com/littmath/status/2021564216575131832) 2026-02-11T12:37Z 54.9K followers, [---] engagements "@emollick don't let Anne Hathaway's publicist see this" [X Link](https://x.com/littmath/status/2021593459476365629) 2026-02-11T14:33Z 54.9K followers, [----] engagements "@olivertraldi Yes I think theres an arguably underdiscussed learning curve wrt how to interact with them" [X Link](https://x.com/littmath/status/2021735697023902176) 2026-02-11T23:58Z 54.8K followers, [---] engagements "@keysmashbandit Awful but no idea how to fix it" [X Link](https://x.com/littmath/status/2022020647678021882) 2026-02-12T18:51Z 54.9K followers, [----] engagements "Huge improvement over past models for sure. It would be interesting to measure how hallucination rates scale with task difficulty. I still somewhat consistently see hallucinations from [---] Pro on difficult math Qs of the form this follows by result XYZ from [--] and then XYZ is slightly (and conveniently) different from what is claimed. Tbf the rate of I dont know how to prove this has certainly improved" [X Link](https://x.com/littmath/status/2022362691097035116) 2026-02-13T17:30Z 54.9K followers, [----] engagements "@spoonedher OTOH my sense is that Problem [--] is itself a fairly involved project" [X Link](https://x.com/littmath/status/2022533840800231890) 2026-02-14T04:50Z 54.9K followers, [---] engagements "@spoonedher But caveat emptor none of the problems are really in my area" [X Link](https://x.com/littmath/status/2022534201309041033) 2026-02-14T04:51Z 54.9K followers, [---] engagements "IMO it should be considered quite rude in most contexts to post or send someone a wall of 100% AI-generated text. Here read this thing I didnt care enough about to express myself" [X Link](https://x.com/littmath/status/2010759165061579086) 2026-01-12T17:01Z 54.9K followers, 772.1K engagements "I think the mathematicians interviewed in this article are well-calibrated with respect to the capabilities of publicly available frontier models for math research. https://www.nytimes.com/2026/02/07/science/mathematics-ai-proof-hairer.htmlsmid=nytcore-ios-share https://www.nytimes.com/2026/02/07/science/mathematics-ai-proof-hairer.htmlsmid=nytcore-ios-share" [X Link](https://x.com/littmath/status/2020210294799442430) 2026-02-07T18:57Z 54.9K followers, 93.8K engagements "@fleetingbits Come on man show what I was responding to" [X Link](https://x.com/littmath/status/2022541027484274797) 2026-02-14T05:18Z 54.9K followers, [----] engagements "FWIW (perhaps not much) ChatGPT [---] Pro agrees with me that the solution is flawed at exactly this point. https://chatgpt.com/share/6990a44b-bdac-8010-ad38-dd0611b16624 https://chatgpt.com/share/6990a44b-bdac-8010-ad38-dd0611b16624" [X Link](https://x.com/littmath/status/2022711683643265293) 2026-02-14T16:37Z 54.9K followers, [----] engagements "Fun thought: it would be good (and arguably a nice demonstration of usefulness) to have the model try to write errata for the incorrect solutions especially those originally claimed to be correct like #2. Heres the error in #2: https://x.com/littmath/status/2022710582860775782s=20 Requesting another pair of eyes on this from someone who knows more about representation theory of p-adic groups than I do. I think that Proposition [---] in the proposed OAI solution to #1stproof problem [--] is false. Would be good to have confirmation. https://x.com/littmath/status/2022710582860775782s=20 Requesting" [X Link](https://x.com/littmath/status/2022837539782996188) 2026-02-15T00:57Z 54.9K followers, [----] engagements "@prfsanjeevarora [--] and [--] are definitely wrong based on a [--] second look; Lean formalizations are nonsense as well" [X Link](https://x.com/littmath/status/2023090095637594497) 2026-02-15T17:40Z 54.9K followers, [----] engagements "RT @yangpliu: 1/ Technical thread on #1stProof Problem 6: finding spectrally light vertex subsets in a graph and how its solution fits i" [X Link](https://x.com/littmath/status/2022757383424926141) 2026-02-14T19:38Z 54.9K followers, [--] engagements "Much harder to predict the future (and I think Im probably a bit more optimistic about future capabilities than some of the interviewees) but at present I think theyre right that models have the capacity to be extremely useful tools but not more than that" [X Link](https://x.com/littmath/status/2020211363789041826) 2026-02-07T19:01Z 54.9K followers, 12.4K engagements "The interviewees are among the authors of First Proof and the article centers around their thinking in writing it. I highly recommend reading the introduction which among other things gives an excellent overview of what math research is. https://arxiv.org/abs/2602.05192 https://arxiv.org/abs/2602.05192" [X Link](https://x.com/littmath/status/2020212265207890395) 2026-02-07T19:05Z 54.9K followers, 57.2K engagements "@fleetingbits Bad behavior to cast me disagreeing with a specific claim as ridiculing a more general one IMO" [X Link](https://x.com/littmath/status/2022544599139643539) 2026-02-14T05:33Z 54.9K followers, [----] engagements "Final comment for now is this is punishment for every time Ive claimed something is standard" [X Link](https://x.com/littmath/status/2022593135013171405) 2026-02-14T08:45Z 54.9K followers, 12.3K engagements "Requesting another pair of eyes on this from someone who knows more about representation theory of p-adic groups than I do. I think that Proposition [---] in the proposed OAI solution to #1stproof problem [--] is false. Would be good to have confirmation. I think the solution to Problem [--] is wrong I think the solution to Problem [--] is wrong" [X Link](https://x.com/littmath/status/2022710582860775782) 2026-02-14T16:32Z 54.9K followers, 47.5K engagements "FWIW this is not my area so caveat emptor but I don't see how the solution strategy can possibly overcome the issues Paul Nelson raises in his comments on the problem" [X Link](https://x.com/littmath/status/2022710920808468585) 2026-02-14T16:33Z 54.9K followers, [----] engagements "@quantum_geoff @JeffConcerto @ben_golub I think I probably agree on second thought--in fact I've seen a more unambiguously autonomous solution to problem [--] that passes my (very uneducated) sniff test. I think I was probably slightly undercalibrated on the kind of performance these scaffolds can elicit" [X Link](https://x.com/littmath/status/2023146036982587567) 2026-02-15T21:22Z 54.9K followers, [---] engagements "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" [X Link](https://x.com/littmath/status/2011518905605583218) 2026-01-14T19:20Z 54.9K followers, 84.2K engagements "Looking forward to seeing how this pans out Very excited about the "First Proof" challenge. I believe novel frontier research is perhaps the most important way to evaluate capabilities of the next generation of AI models. We have run our internal model with limited human supervision on the ten proposed problems. The Very excited about the "First Proof" challenge. I believe novel frontier research is perhaps the most important way to evaluate capabilities of the next generation of AI models. We have run our internal model with limited human supervision on the ten proposed problems. The" [X Link](https://x.com/littmath/status/2022529725176897665) 2026-02-14T04:33Z 54.9K followers, 29.9K engagements "@ben_golub FWIW I expect it to take several days for clarity to emerge; would be nice if OAI employees would chill at least until one other person looks at 2" [X Link](https://x.com/littmath/status/2022781248985244007) 2026-02-14T21:13Z 54.9K followers, [----] engagements "@DayShuai @prfsanjeevarora I think correctness of informal proofs aside (which is very labor-intensive to verify) IMO you should retract the claim to have verified any of your solutions in Lean. Clearly you are "verifying" some false results so whatever Lean you are producing is evidently unreliable" [X Link](https://x.com/littmath/status/2023242924402610290) 2026-02-16T03:47Z 54.9K followers, [--] engagements "Nate Silver ran [-----] high-fidelity simulations of the election which will only happen once. Whats more likelythat we are watching the actual election or that we are in one of his simulations dancing for his amusement" [X Link](https://x.com/littmath/status/1853947902302765346) 2024-11-05T23:49Z 54.8K followers, 158.3K engagements "let the efficiency flow through you" [X Link](https://x.com/anyuser/status/1857180056126190008) 2024-11-14T21:53Z 54.8K followers, 422.7K engagements "department of the federal government that only recruits bluechecks" [X Link](https://x.com/anyuser/status/1857180681480401112) 2024-11-14T21:55Z 54.8K followers, 33.5K engagements "progress in mathematics will inevitably slow. indeed once you count high enough the numbers take a very long time to say out loud" [X Link](https://x.com/anyuser/status/1858675896653115528) 2024-11-19T00:57Z 54.8K followers, 103.2K engagements "The periodic avalanches of misogyny on here directed at women who have posted about some accomplishment (in this case getting a PhD) are really something. A bit funny to see thousands of chuds pretending to have read a PhD thesis though" [X Link](https://x.com/anyuser/status/1863269831622996233) 2024-12-01T17:12Z 54.8K followers, 209K engagements "in order to get taxpayer funding for your research you should first have to explain why its interesting to a lay audience of the [----] most incoherent incels on X the everything site" [X Link](https://x.com/littmath/status/1863447139054923784) 2024-12-02T04:56Z 54.8K followers, 90.5K engagements "Congrats on defending your thesis in front of a committee made up of distinguished professors in your field. Only one hurdle remains before you can be granted your doctoratea grand tradition of higher education: the internet harassment campaign portion of your thesis defense" [X Link](https://x.com/littmath/status/1863684762105266628) 2024-12-02T20:40Z 54.9K followers, 101.7K engagements "Complaints about jargon in academic (or e.g. legal) writing are sort of mystifying to me. You can just learn what words mean" [X Link](https://x.com/littmath/status/1869073389207601463) 2024-12-17T17:33Z 54.9K followers, 108.3K engagements "FWIW the performance of o3 on FrontierMath is obviously immensely impressive but as usual people need to cool their jets a bit. E.g. these two problems (rated "medium" and "low" respectively) I immediately knew how to do" [X Link](https://x.com/littmath/status/1870543769323581783) 2024-12-21T18:56Z 54.7K followers, 217.7K engagements "if Trump negatively polarizes the left against H-1Bs i am going to become the joker" [X Link](https://x.com/anyuser/status/1875034796122104192) 2025-01-03T04:21Z 54.8K followers, 143.9K engagements "went on a 5-plus-minute rant about some mathematical text that used the phrase an empty set. Theres only one empty set All empty sets are equal Call it *the* empty set" [X Link](https://x.com/littmath/status/1883638155691332004) 2025-01-26T22:08Z 54.8K followers, 201.6K engagements "@andrewprock @Lptomov82 Its a theorem of ZFCnot an axiomthat any two empty sets are equal. (And not just ZFC; not a set theorist but this is true in the [--] foundations of set theory about which I know enough to check it.)" [X Link](https://x.com/littmath/status/1883645501163385024) 2025-01-26T22:37Z 54.6K followers, [---] engagements "25-50% of staff at the NSF to be laid off in the next [--] months apparently. Seems bad" [X Link](https://x.com/littmath/status/1886976511942480152) 2025-02-05T03:13Z 54.8K followers, 1.5M engagements "Something I wish mathematicians conveyed more convincingly to our students is how its possible to get stuck on a problem for literally *years* and then solve it" [X Link](https://x.com/littmath/status/1889070324014027094) 2025-02-10T21:53Z 54.9K followers, 215.3K engagements "Begging AI companies to stop using the meaningless term PhD-level in their marketing" [X Link](https://x.com/anyuser/status/1897487414869856620) 2025-03-06T03:20Z 54.8K followers, 220.7K engagements "my almost-two-year-old said tetrahedron cube and octahedron today" [X Link](https://x.com/anyuser/status/1908709406486691862) 2025-04-06T02:32Z 54.8K followers, 42.1K engagements "Yes our best and brightest should be serving the public interest through the means for which they are best-suited: by shaving off a few nanoseconds of latency for HFT firms. In the long-run even cuts to STEM funding are very good. Top STEM researchers belong in industry not academia. In the long-run even cuts to STEM funding are very good. Top STEM researchers belong in industry not academia" [X Link](https://x.com/littmath/status/1912925178742411370) 2025-04-17T17:44Z 54.9K followers, 828.3K engagements "If Apples market cap is $2.96 trillion why should we pay for iPhones and MacBooks Seems absurd. If Harvard has $53bn of endowments why on earth does it need or warrant another $2bn of public funding Seems absurd. If Harvard has $53bn of endowments why on earth does it need or warrant another $2bn of public funding Seems absurd" [X Link](https://x.com/anyuser/status/1912993199020274115) 2025-04-17T22:14Z 54.8K followers, 1.2M engagements "first math pope" [X Link](https://x.com/anyuser/status/1920538577055699255) 2025-05-08T17:57Z 54.8K followers, 1.4M engagements "Incredibly bad (changes to N.S.F. funding for math):" [X Link](https://x.com/littmath/status/1925640323109191974) 2025-05-22T19:50Z 54.8K followers, 744.3K engagements "IMO it would be useful for the Vice President to explain why he thinks cutting math and physics funding 70-80+% advances the aims listed in this tweet. There is an extraordinary "reproducibility crisis" in the sciences particularly in biology where most published papers fail to replicate. Most universities have massive bureaucracies that inhibit the translation of basic research into commercial adoption. The voting There is an extraordinary "reproducibility crisis" in the sciences particularly in biology where most published papers fail to replicate. Most universities have massive" [X Link](https://x.com/anyuser/status/1926590979634581822) 2025-05-25T10:47Z 54.8K followers, 200.1K engagements "I think its pretty unlikely that Grok 3.5/4 will be used to rewrite the entire corpus of human knowledge adding missing information and deleting errors" [X Link](https://x.com/littmath/status/1936414981593153710) 2025-06-21T13:24Z 54.8K followers, 117.3K engagements "Claude: my wife and I went antique shopping this weekend Gemini: if I cant 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 [----] why do you ask" [X Link](https://x.com/littmath/status/1942696166774431780) 2025-07-08T21:23Z 54.8K followers, 189.9K engagements "😏 looks like theres a huge amount of evidence for my conjecture that all numbers are less than 10100" [X Link](https://x.com/littmath/status/1944443271599497234) 2025-07-13T17:06Z 54.8K followers, 248.1K engagements "if you want a picture of the future imagine being fed cocomelon from the cradle to the grave You can create from scratch remix what you see or just scroll through to check out videos from the creators + the visual artists weve been collaborating with. https://t.co/M9kWNjyoEc You can create from scratch remix what you see or just scroll through to check out videos from the creators + the visual artists weve been collaborating with. https://t.co/M9kWNjyoEc" [X Link](https://x.com/littmath/status/1971555791703507071) 2025-09-26T12:41Z 54.8K followers, 180.6K engagements "It's good for academics to publicly experiment with new AI tools but important to report both successes and failures when doing so. Audience capture incentivizes only doing one or the other which is part of the reason the information environment around capabilities is so bad" [X Link](https://x.com/littmath/status/2007165168716001456) 2026-01-02T19:00Z 54.9K followers, 29.4K engagements "The basic objects of study here are algebraic varieties--shapes defined as the set of solutions to a system of polynomial equations--and polynomial maps between them" [X Link](https://x.com/littmath/status/2011518908327657847) 2026-01-14T19:20Z 54.6K followers, [----] engagements "One of the basic themes of 20th century mathematics is that the topology of the set of complex solutions to a system of polynomial equations is controlled by the arithmetic of the polynomials" [X Link](https://x.com/littmath/status/2011518910345134514) 2026-01-14T19:20Z 54.6K followers, [----] engagements "For example the Weil conjectures (proved by Dwork Grothendieck and Deligne) show that the Betti numbers of a (smooth projective) variety defined by polynomials with integer coefficients are controlled by the number of solutions to those polynomials over finite fields" [X Link](https://x.com/littmath/status/2011518912207474893) 2026-01-14T19:20Z 54.6K followers, [----] engagements "@Archivara Im really sorry to put you guys on blast again but you need to hire a subject-matter expert. The content of this paper is (1) Winograds [----] result (basically CRT) that the number of multiplications youre computing is at most 2*deg-#factors and (3) 2*5-3=7" [X Link](https://x.com/littmath/status/2012912210327077035) 2026-01-18T15:37Z 54.6K followers, 51K engagements "I have little doubt that AI tools will substantially change the way mathematics is done likely pretty soon. But communicators like this are doing their audiences a disservice. In this particular example the new contribution from the AI tool was plugging (nm)=(53) into the expression 2n-m. Im not exaggerating. The number of multiplications in the result in question was shown by Winograd [--] years ago to be 2n-m where the matrices in question are nxn and m is the number of factors of the polynomial xn-1. The result here is that this polynomial has [--] factors over Q(sqrt(5)) and then that 2*5-3=7." [X Link](https://x.com/littmath/status/2013037012765331883) 2026-01-18T23:53Z 54.6K followers, 75.5K engagements "Toddler: I opened the fridge all by myself I tried and tried with all my might I am super STRONG" [X Link](https://x.com/littmath/status/2015418274427670918) 2026-01-25T13:35Z 54.6K followers, [----] engagements "Figured I might as well amplify this Q. @kevinroose Do you have a sense of whether people are productively using multi-agent setups for non-SWE tasks @kevinroose Do you have a sense of whether people are productively using multi-agent setups for non-SWE tasks" [X Link](https://x.com/littmath/status/2015496002317234279) 2026-01-25T18:44Z 54.6K followers, 14.7K engagements "Toddler melting down due to broken granola bar. Me: How bout I glue it together with honey. Toddler *tearfully*: OK Me: Did you know bees make honey Toddler: *nods* they make pizza too" [X Link](https://x.com/littmath/status/2015517764111233349) 2026-01-25T20:10Z 54.6K followers, 238K engagements "Toddler talking to her uncle on the phone: Where are you Uncle: Im at work. What about you Toddler: I dont work. I just play Youre SO silly" [X Link](https://x.com/littmath/status/2016280209176088903) 2026-01-27T22:40Z 54.6K followers, 60.7K engagements "Youll never guess what this is a reply to. Kind of a surprise to get blocked for it @tunguz LOL ok. @tunguz LOL ok" [X Link](https://x.com/littmath/status/2016523501054656841) 2026-01-28T14:47Z 54.7K followers, 59.3K engagements "@jakobzupanec @bayesianboy Quite a bit for literature search. Less for actual math though the latest models are sometimes useful for routine arguments" [X Link](https://x.com/littmath/status/2016587521497420111) 2026-01-28T19:01Z 54.6K followers, [---] engagements "@Allodoxaa @jakobzupanec @bayesianboy Paid models are way way better for basically any task. I mostly use ChatGPT [---] Thinking/Pro" [X Link](https://x.com/littmath/status/2016844035030172145) 2026-01-29T12:01Z 54.6K followers, [---] engagements "This was a really fun podcast to record Not only about AIalso really enjoyed sharing some thoughts on the practice of mathematics overall. How does math research change when the cost of trying your first dumb idea goes to zero @littmath joins @GregHBurnham and @ansonwhho to discuss what todays models can and cant do in math and how far they are from doing high-quality research. 0:00:00 What's the hardest math https://t.co/vxxOevyLu4 How does math research change when the cost of trying your first dumb idea goes to zero @littmath joins @GregHBurnham and @ansonwhho to discuss what todays models" [X Link](https://x.com/littmath/status/2016982547947819187) 2026-01-29T21:11Z 54.6K followers, 11.1K engagements "Small comment is that this was recorded in December of last year an eternity ago in AI time. But I think its held up well overall. Two small caveats: (1) this was recorded just before the release of GPT [---] which was a substantial improvement for math use cases. I probably would have mentioned this if it had been recorded a few days later. (2) This was also right at the beginning of the recent spate of (semi-)autonomous solutions to e.g. Erdos problems. But I think what I said about these has held up so far https://twitter.com/i/web/status/2016983497827897733" [X Link](https://x.com/littmath/status/2016983497827897733) 2026-01-29T21:15Z 54.6K followers, [----] engagements "@vladtenev @HarmonicMath Do you really believe that no Erdos problems will be left in two years" [X Link](https://x.com/littmath/status/2017379242103681145) 2026-01-30T23:27Z 54.7K followers, [----] engagements "Toddler: I want two apricots because Im two-and-a-half. *Pause* When I turn three I can have three apricots. *Pause realization* When I turn four I can have four apricots When I turn five I can have FIVE apricots" [X Link](https://x.com/littmath/status/2017725001697222664) 2026-01-31T22:21Z 54.7K followers, 36.6K engagements "Very easy to get LLMs to roleplay reddit. It even did the viral moltbook post about getting the agent to delete stuff on its own" [X Link](https://x.com/littmath/status/2017989982930014450) 2026-02-01T15:54Z 54.6K followers, [----] engagements "@Afinetheorem lol I could generate [--] tough math questions and the AI is the one that gets them right" [X Link](https://x.com/littmath/status/2019032826965119105) 2026-02-04T12:58Z 54.7K followers, [----] engagements "@Afinetheorem @JaneParkway Yeah I think it should be possible to get the models to do basically any computer use task the average person can do now though for some tasks it would likely be a huge pain" [X Link](https://x.com/littmath/status/2019057332420739431) 2026-02-04T14:35Z 54.7K followers, [---] engagements "@AdrianTMiranda I have no doubt its harder for contestants than IMO but the contestant pool is much bigger" [X Link](https://x.com/littmath/status/1998016319141163458) 2025-12-08T13:06Z 54.9K followers, [----] engagements "Part of being a mathematician is finding something you feel you cant understand and nonetheless working to understand it. Rather than claiming expertise about the role of genetics in math Murray might benefit by showing some curiosity about what it is mathematicians do" [X Link](https://x.com/littmath/status/1681476550363840512) 2023-07-19T01:30Z 54.9K followers, 216.4K engagements "Honestly dont see how the US comes back from this without serious legal consequences for members of the administration and people following their orders" [X Link](https://x.com/anyuser/status/1912171613895733350) 2025-04-15T15:50Z 54.9K followers, 108.5K engagements Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing
@littmath Daniel Litt
Daniel Litt posts on X about math, ai, in the, wrong the most. They currently have [------] followers and [---] posts still getting attention that total [------] engagements in the last [--] hours.
Engagements: [------] #
- [--] Week [-------] +31%
- [--] Month [---------] -66%
- [--] Months [---------] -11%
- [--] Year [----------] -43%
Mentions: [--] #
- [--] Week [--] +65%
- [--] Month [---] -26%
- [--] Months [---] +80%
- [--] Year [---] +52%
Followers: [------] #
- [--] Week [------] +0.27%
- [--] Month [------] +1%
- [--] Months [------] +10%
- [--] Year [------] +20%
CreatorRank: [-------] #
Social Influence
Social category influence technology brands 5.17% travel destinations 0.86% currencies 0.86% stocks 0.86% finance 0.86% social networks 0.86%
Social topic influence math #2487, ai 9.48%, in the 6.9%, wrong #758, how to 4.31%, this is 3.45%, future 2.59%, the first 2.59%, model 2.59%, theory #717
Top accounts mentioned or mentioned by @sheafnat @andilesanthony @xixidu @bengolub @jeffconcerto @afinetheorem @quantumgeoff @sar1287 @jasondeanlee @sosowski @ognifedefingo @profnoahgian @alzzyd @oydeis @acerfur @spoonedher @fleetingbits @prfsanjeevarora @lptomov82 @jakobzupanec
Top assets mentioned Alphabet Inc Class A (GOOGL)
Top Social Posts
Top posts by engagements in the last [--] hours
"@sheafnat @sar1287 @SwallieC69635 Wonderful Please keep me posted if you do"
X Link 2026-02-14T15:33Z 54.9K followers, [---] engagements
"This kind of thing is so useless. If the words in this sentence have their standard meaning its already been falsified (though one can debate the significance of the discoveries). Of course its more likely the words are defined so as to make the claim contentless. AI cannot in principle make novel discoveries. AI cannot in principle make novel discoveries"
X Link 2026-02-12T20:03Z 54.9K followers, 37.9K engagements
"Those are the ones I can semi-quickly evaluate. If an expert has any thoughts on the proposed solution to [--] Id be very curious"
X Link 2026-02-14T08:44Z 54.9K followers, [----] engagements
"Glad to see this clarified. I think it would be good to wait a bit to see how the rest of the claimed solutions shake out; tbc I do not have any strong opinion about any I have not already discussed on here. Would also like to hear more about how solutions were generated Based on the official #1stProof commentary community analysis and more clarification with external experts we now believe the solution to problem [--] above is likely incorrect. Grateful for the engagement and looking forward to continued review Based on the official #1stProof commentary community analysis and more clarification"
X Link 2026-02-15T00:33Z 54.9K followers, 13K engagements
"@JeffConcerto @ben_golub Impressive IMO (and I think at least [--] is pretty likely) but hard to be more precise without having some more insight into how they were generated. I tweeted at some point at the start of the week that my weak expectation was 2-3 but 4-5 wouldnt shock me"
X Link 2026-02-14T23:08Z 54.9K followers, [----] engagements
"FWIW I think ideally OAI employees etc. would also urge caution interpreting these claims until theyve been checked"
X Link 2026-02-15T00:47Z 54.9K followers, [----] engagements
"@XiXiDu This is a question of motivation not capability"
X Link 2026-02-15T11:38Z 54.9K followers, [----] engagements
"@ProfNoahGian @XiXiDu They are actually not going to verify the answers to round 1; this was supposed to be a proof of concept"
X Link 2026-02-15T12:50Z 54.9K followers, [---] engagements
"toddler (on the way to preschool): I can see the CN tower me: it was too cloudy to see it yesterday. isnt it nice we can see the whole thing today toddler patiently: daddy we can only see the front of it"
X Link 2026-02-11T14:53Z 54.9K followers, 25.4K engagements
"as a mathematician parent I have never been prouder"
X Link 2026-02-11T15:02Z 54.9K followers, [----] engagements
"Beautiful talk by Barry Mazur on the BSD conjecture from a few days ago here: Good example of a quite early major impact of computing on mathematics research. https://www.youtube.com/watchv=14-9iCoclFE https://www.youtube.com/watchv=14-9iCoclFE"
X Link 2026-02-13T18:35Z 54.9K followers, [----] engagements
"@alz_zyd_ who hurt you"
X Link 2026-02-14T06:12Z 54.9K followers, [----] engagements
"hmmmmmmmmmmmm"
X Link 2026-02-14T08:23Z 54.9K followers, 10.9K engagements
"I think the solution to Problem [--] is wrong"
X Link 2026-02-14T08:26Z 54.9K followers, 50.2K engagements
"AI enthusiasts interested in "accelerating science" feel free to amplify. Arguably correct solutions are more interesting but I think being confidently wrong is also an interesting signal and it would be good to have confirmation"
X Link 2026-02-14T17:09Z 54.9K followers, [----] engagements
"@jasondeanlee Re: autoformalization companies I am only aware of one credible Lean verification (of problem 9) which seems to have involved a LOT of human labor. Tbf I think most of these problems have statements that would be very challenging to formalize given the current state of Mathlib"
X Link 2026-02-14T18:45Z 54.9K followers, [----] engagements
"I think no. Ive pinged some experts about [--] (fwiw Im fairly confident its wrong) but havent heard back yet. AFAIK nothing is really checked; the only other things Im confident about is [--] is essentially right and [--] is very wrong. Solution to [--] seems to be credible but ofc I cant vouch for it myself. https://twitter.com/i/web/status/2022780672948199503 https://twitter.com/i/web/status/2022780672948199503"
X Link 2026-02-14T21:11Z 54.9K followers, 41K engagements
"@ben_golub Based on author comments Id be surprised if [--] is wrong and I wouldnt be surprised if [--] is right"
X Link 2026-02-14T21:11Z 54.9K followers, [----] engagements
"Really good question (note that DeepMind shared transcripts in their recent Aletheia paper and I think this is clearly best practice). Hopefully OAI follows suit. @merettm as an organizer of #1stProof this is truly exciting jakub would you consider sharing your transcripts @merettm as an organizer of #1stProof this is truly exciting jakub would you consider sharing your transcripts"
X Link 2026-02-15T01:17Z 54.9K followers, 22.7K engagements
"No need to amplify further OAI confirms its likely incorrect"
X Link 2026-02-15T01:34Z 54.9K followers, [----] engagements
"@quantum_geoff @JeffConcerto @ben_golub I mean e.g. experts identify promising strategies: https://x.com/merettm/status/2022925024798372262s=20 @Knikct Hi Nikhil We will aim to publish more information next week but as I noted above this was a quite chaotic sprint (you caught us by surprise please give us time to prepare next time). We will not be able to gather all the transcripts as they are quite scattered. Some of the https://x.com/merettm/status/2022925024798372262s=20 @Knikct Hi Nikhil We will aim to publish more information next week but as I noted above this was a quite chaotic sprint"
X Link 2026-02-15T11:17Z 54.9K followers, [---] engagements
"@XiXiDu I was responding to your claim about Batson whom I have no doubt could check this proof without learning to build a jet engine. He just doesnt care to"
X Link 2026-02-15T12:20Z 54.9K followers, [---] engagements
"@quantum_geoff @JeffConcerto @ben_golub .@jasondeanlee @damekdavis would be curious to get your thoughts on how to interpret this info from Pachocki"
X Link 2026-02-15T13:06Z 54.9K followers, [---] engagements
"@alz_zyd_ Basically endorse this"
X Link 2026-02-15T18:15Z 54.9K followers, [----] engagements
"The US National Science Foundation spends about $250 million (yes thats million with an m) funding math research (pure and applied) per year. ROI is actually incredible"
X Link 2025-03-05T14:35Z 54.9K followers, 254.7K engagements
"new AI tool that lets me enter The Last Supper walk around and explore it and find a Super Mario [--] star glowing and spinning under the table"
X Link 2025-08-10T19:04Z 54.9K followers, 72.1K engagements
"Sort of funny: published conjectures that are at the level where they can be resolved autonomously by AI are a non-renewable resource since in the future well just ask an AI rather than publishing the conjecture"
X Link 2026-01-31T01:50Z 54.8K followers, 21.7K engagements
"Lot of chatter on the timeline about Sonnet 5; all I have to say is: "But flowers distilld though they with winter meet Leese but their show; their substance still lives sweet.""
X Link 2026-02-02T01:41Z 54.7K followers, 23.4K engagements
"Really nice paper. Aside from the nice results resolving a number of open Erdos problems (or finding solutions in the literature) the paper does a really good job contextualizing significance and prior work. Okay looks like I can now talk about Aletheia on the Erds Problems https://t.co/JqpyUJUIcV Okay looks like I can now talk about Aletheia on the Erds Problems https://t.co/JqpyUJUIcV"
X Link 2026-02-02T15:41Z 54.8K followers, 15.6K engagements
"@Afinetheorem FWIW I do not think its very hard to generate questions where frontier models still substantially underperform humans. I was doing some trip planning with [---] Thinking the other day and it was basically unable to correctly answer Qs of the form list [--] hotels satisfying XYZ"
X Link 2026-02-04T14:07Z 54.8K followers, [---] engagements
"@Afinetheorem Here XYZ were desiderata like less than [--] minute walk to a subway availability on July [--] with a suite that costs at most $Z"
X Link 2026-02-04T14:08Z 54.8K followers, [---] engagements
"That we can now automate some mathematics that previously required an expert is a huge deal. That said the mathematics produced thus far is (in my obviously very subjective opinion) not notable in itself but rather because it is automated and as a leading indicator"
X Link 2026-02-05T18:35Z 54.8K followers, 56.2K engagements
"@quantum_geoff Curve of best fit f(x)=1+x. Oops no its 1+x+x2/2. Oops no its 1+x+x2/2+x3/6. Oops no its 1+x+x2/2+x3/6+x4/24. Oops no its"
X Link 2026-02-06T14:15Z 54.8K followers, [----] engagements
"This is a really nice test suite. Martin Hairer and colleagues released a set of hard maths problems designed to be test cases for LLMs. We have one week to solve them using LLMs. They encrypted the solutions at https://t.co/EZNjVzFT9t and will reveal them just after. https://t.co/20TaPDaSf2 (1/3) Martin Hairer and colleagues released a set of hard maths problems designed to be test cases for LLMs. We have one week to solve them using LLMs. They encrypted the solutions at https://t.co/EZNjVzFT9t and will reveal them just after. https://t.co/20TaPDaSf2 (1/3)"
X Link 2026-02-06T15:30Z 54.8K followers, 27.9K engagements
"Playing around with Opus [---] (extended thinking) for math. Seems to be plausibly the first Anthropic model to be useful for my purposes; it answered correctly (though non-rigorously) a pretty tricky algebraic geometry question that no other model has gotten"
X Link 2026-02-07T02:08Z 54.8K followers, 28.3K engagements
"TBC "useful" here means e.g. "it can sometimes perform routine computations I need" which is comparable with other frontier models"
X Link 2026-02-07T02:26Z 54.8K followers, [----] engagements
"Interestingly fails another one of my test questions in the exact same way [---] Pro does; both models consistently seem to (incorrectly) cite the same result"
X Link 2026-02-07T02:33Z 54.8K followers, [----] engagements
"Opus says "that's a beautiful question" a bit too much for my taste (though of course my questions are in fact beautiful). I prefer GPT 5.2's barely-concealed contempt"
X Link 2026-02-07T03:10Z 54.8K followers, [----] engagements
"@XiXiDu @leifweatherby Ive been playing around with the models for a long time since GPT-3. I guess I started doing so seriously when o1 came out which was the first time there were clear signals of usefulness for math imo. Obviously improvement since GPT-4 has been remarkably fast"
X Link 2026-02-07T13:26Z 54.8K followers, [--] engagements
"@oydeis Eh in math almost everything is on arxiv"
X Link 2026-02-07T20:41Z 54.7K followers, [---] engagements
"Some seem to be reading this interview as predicting that AI models will not continue to rapidly improve at math. I think this is wrongas I read it the interviewees are declining to make a prediction and are instead focused on accurately evaluating existing models. I think the mathematicians interviewed in this article are well-calibrated with respect to the capabilities of publicly available frontier models for math research. https://t.co/jUMkaqRa43 I think the mathematicians interviewed in this article are well-calibrated with respect to the capabilities of publicly available frontier"
X Link 2026-02-08T15:13Z 54.8K followers, 10.1K engagements
"@Miles_Brundage I think all the AI-for-math companies have done this Google just released a paper about one of their internal scaffolds (Alethia) and there are a number of individuals who have built scaffolds using various model APIs. This doesn't count"
X Link 2026-02-08T20:32Z 54.8K followers, [----] engagements
"me: reading The Lorax NOWthanks to your hacking my trees to the ground there's not enough Truffula Fruit to go 'round. And my poor Bar-ba-loots are all getting the crummies because they have gas and no food in their tummies toddler: They should get sushi"
X Link 2026-02-09T00:57Z 54.8K followers, [----] engagements
"@AcerFur I really think you should mention you independently came up with the same argument for [----] prior to the model. Why give the model all the credit"
X Link 2026-02-11T12:37Z 54.9K followers, [---] engagements
"@emollick don't let Anne Hathaway's publicist see this"
X Link 2026-02-11T14:33Z 54.9K followers, [----] engagements
"@olivertraldi Yes I think theres an arguably underdiscussed learning curve wrt how to interact with them"
X Link 2026-02-11T23:58Z 54.8K followers, [---] engagements
"@keysmashbandit Awful but no idea how to fix it"
X Link 2026-02-12T18:51Z 54.9K followers, [----] engagements
"Huge improvement over past models for sure. It would be interesting to measure how hallucination rates scale with task difficulty. I still somewhat consistently see hallucinations from [---] Pro on difficult math Qs of the form this follows by result XYZ from [--] and then XYZ is slightly (and conveniently) different from what is claimed. Tbf the rate of I dont know how to prove this has certainly improved"
X Link 2026-02-13T17:30Z 54.9K followers, [----] engagements
"@spoonedher OTOH my sense is that Problem [--] is itself a fairly involved project"
X Link 2026-02-14T04:50Z 54.9K followers, [---] engagements
"@spoonedher But caveat emptor none of the problems are really in my area"
X Link 2026-02-14T04:51Z 54.9K followers, [---] engagements
"IMO it should be considered quite rude in most contexts to post or send someone a wall of 100% AI-generated text. Here read this thing I didnt care enough about to express myself"
X Link 2026-01-12T17:01Z 54.9K followers, 772.1K engagements
"I think the mathematicians interviewed in this article are well-calibrated with respect to the capabilities of publicly available frontier models for math research. https://www.nytimes.com/2026/02/07/science/mathematics-ai-proof-hairer.htmlsmid=nytcore-ios-share https://www.nytimes.com/2026/02/07/science/mathematics-ai-proof-hairer.htmlsmid=nytcore-ios-share"
X Link 2026-02-07T18:57Z 54.9K followers, 93.8K engagements
"@fleetingbits Come on man show what I was responding to"
X Link 2026-02-14T05:18Z 54.9K followers, [----] engagements
"FWIW (perhaps not much) ChatGPT [---] Pro agrees with me that the solution is flawed at exactly this point. https://chatgpt.com/share/6990a44b-bdac-8010-ad38-dd0611b16624 https://chatgpt.com/share/6990a44b-bdac-8010-ad38-dd0611b16624"
X Link 2026-02-14T16:37Z 54.9K followers, [----] engagements
"Fun thought: it would be good (and arguably a nice demonstration of usefulness) to have the model try to write errata for the incorrect solutions especially those originally claimed to be correct like #2. Heres the error in #2: https://x.com/littmath/status/2022710582860775782s=20 Requesting another pair of eyes on this from someone who knows more about representation theory of p-adic groups than I do. I think that Proposition [---] in the proposed OAI solution to #1stproof problem [--] is false. Would be good to have confirmation. https://x.com/littmath/status/2022710582860775782s=20 Requesting"
X Link 2026-02-15T00:57Z 54.9K followers, [----] engagements
"@prfsanjeevarora [--] and [--] are definitely wrong based on a [--] second look; Lean formalizations are nonsense as well"
X Link 2026-02-15T17:40Z 54.9K followers, [----] engagements
"RT @yangpliu: 1/ Technical thread on #1stProof Problem 6: finding spectrally light vertex subsets in a graph and how its solution fits i"
X Link 2026-02-14T19:38Z 54.9K followers, [--] engagements
"Much harder to predict the future (and I think Im probably a bit more optimistic about future capabilities than some of the interviewees) but at present I think theyre right that models have the capacity to be extremely useful tools but not more than that"
X Link 2026-02-07T19:01Z 54.9K followers, 12.4K engagements
"The interviewees are among the authors of First Proof and the article centers around their thinking in writing it. I highly recommend reading the introduction which among other things gives an excellent overview of what math research is. https://arxiv.org/abs/2602.05192 https://arxiv.org/abs/2602.05192"
X Link 2026-02-07T19:05Z 54.9K followers, 57.2K engagements
"@fleetingbits Bad behavior to cast me disagreeing with a specific claim as ridiculing a more general one IMO"
X Link 2026-02-14T05:33Z 54.9K followers, [----] engagements
"Final comment for now is this is punishment for every time Ive claimed something is standard"
X Link 2026-02-14T08:45Z 54.9K followers, 12.3K engagements
"Requesting another pair of eyes on this from someone who knows more about representation theory of p-adic groups than I do. I think that Proposition [---] in the proposed OAI solution to #1stproof problem [--] is false. Would be good to have confirmation. I think the solution to Problem [--] is wrong I think the solution to Problem [--] is wrong"
X Link 2026-02-14T16:32Z 54.9K followers, 47.5K engagements
"FWIW this is not my area so caveat emptor but I don't see how the solution strategy can possibly overcome the issues Paul Nelson raises in his comments on the problem"
X Link 2026-02-14T16:33Z 54.9K followers, [----] engagements
"@quantum_geoff @JeffConcerto @ben_golub I think I probably agree on second thought--in fact I've seen a more unambiguously autonomous solution to problem [--] that passes my (very uneducated) sniff test. I think I was probably slightly undercalibrated on the kind of performance these scaffolds can elicit"
X Link 2026-02-15T21:22Z 54.9K followers, [---] engagements
"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"
X Link 2026-01-14T19:20Z 54.9K followers, 84.2K engagements
"Looking forward to seeing how this pans out Very excited about the "First Proof" challenge. I believe novel frontier research is perhaps the most important way to evaluate capabilities of the next generation of AI models. We have run our internal model with limited human supervision on the ten proposed problems. The Very excited about the "First Proof" challenge. I believe novel frontier research is perhaps the most important way to evaluate capabilities of the next generation of AI models. We have run our internal model with limited human supervision on the ten proposed problems. The"
X Link 2026-02-14T04:33Z 54.9K followers, 29.9K engagements
"@ben_golub FWIW I expect it to take several days for clarity to emerge; would be nice if OAI employees would chill at least until one other person looks at 2"
X Link 2026-02-14T21:13Z 54.9K followers, [----] engagements
"@DayShuai @prfsanjeevarora I think correctness of informal proofs aside (which is very labor-intensive to verify) IMO you should retract the claim to have verified any of your solutions in Lean. Clearly you are "verifying" some false results so whatever Lean you are producing is evidently unreliable"
X Link 2026-02-16T03:47Z 54.9K followers, [--] engagements
"Nate Silver ran [-----] high-fidelity simulations of the election which will only happen once. Whats more likelythat we are watching the actual election or that we are in one of his simulations dancing for his amusement"
X Link 2024-11-05T23:49Z 54.8K followers, 158.3K engagements
"let the efficiency flow through you"
X Link 2024-11-14T21:53Z 54.8K followers, 422.7K engagements
"department of the federal government that only recruits bluechecks"
X Link 2024-11-14T21:55Z 54.8K followers, 33.5K engagements
"progress in mathematics will inevitably slow. indeed once you count high enough the numbers take a very long time to say out loud"
X Link 2024-11-19T00:57Z 54.8K followers, 103.2K engagements
"The periodic avalanches of misogyny on here directed at women who have posted about some accomplishment (in this case getting a PhD) are really something. A bit funny to see thousands of chuds pretending to have read a PhD thesis though"
X Link 2024-12-01T17:12Z 54.8K followers, 209K engagements
"in order to get taxpayer funding for your research you should first have to explain why its interesting to a lay audience of the [----] most incoherent incels on X the everything site"
X Link 2024-12-02T04:56Z 54.8K followers, 90.5K engagements
"Congrats on defending your thesis in front of a committee made up of distinguished professors in your field. Only one hurdle remains before you can be granted your doctoratea grand tradition of higher education: the internet harassment campaign portion of your thesis defense"
X Link 2024-12-02T20:40Z 54.9K followers, 101.7K engagements
"Complaints about jargon in academic (or e.g. legal) writing are sort of mystifying to me. You can just learn what words mean"
X Link 2024-12-17T17:33Z 54.9K followers, 108.3K engagements
"FWIW the performance of o3 on FrontierMath is obviously immensely impressive but as usual people need to cool their jets a bit. E.g. these two problems (rated "medium" and "low" respectively) I immediately knew how to do"
X Link 2024-12-21T18:56Z 54.7K followers, 217.7K engagements
"if Trump negatively polarizes the left against H-1Bs i am going to become the joker"
X Link 2025-01-03T04:21Z 54.8K followers, 143.9K engagements
"went on a 5-plus-minute rant about some mathematical text that used the phrase an empty set. Theres only one empty set All empty sets are equal Call it the empty set"
X Link 2025-01-26T22:08Z 54.8K followers, 201.6K engagements
"@andrewprock @Lptomov82 Its a theorem of ZFCnot an axiomthat any two empty sets are equal. (And not just ZFC; not a set theorist but this is true in the [--] foundations of set theory about which I know enough to check it.)"
X Link 2025-01-26T22:37Z 54.6K followers, [---] engagements
"25-50% of staff at the NSF to be laid off in the next [--] months apparently. Seems bad"
X Link 2025-02-05T03:13Z 54.8K followers, 1.5M engagements
"Something I wish mathematicians conveyed more convincingly to our students is how its possible to get stuck on a problem for literally years and then solve it"
X Link 2025-02-10T21:53Z 54.9K followers, 215.3K engagements
"Begging AI companies to stop using the meaningless term PhD-level in their marketing"
X Link 2025-03-06T03:20Z 54.8K followers, 220.7K engagements
"my almost-two-year-old said tetrahedron cube and octahedron today"
X Link 2025-04-06T02:32Z 54.8K followers, 42.1K engagements
"Yes our best and brightest should be serving the public interest through the means for which they are best-suited: by shaving off a few nanoseconds of latency for HFT firms. In the long-run even cuts to STEM funding are very good. Top STEM researchers belong in industry not academia. In the long-run even cuts to STEM funding are very good. Top STEM researchers belong in industry not academia"
X Link 2025-04-17T17:44Z 54.9K followers, 828.3K engagements
"If Apples market cap is $2.96 trillion why should we pay for iPhones and MacBooks Seems absurd. If Harvard has $53bn of endowments why on earth does it need or warrant another $2bn of public funding Seems absurd. If Harvard has $53bn of endowments why on earth does it need or warrant another $2bn of public funding Seems absurd"
X Link 2025-04-17T22:14Z 54.8K followers, 1.2M engagements
"first math pope"
X Link 2025-05-08T17:57Z 54.8K followers, 1.4M engagements
"Incredibly bad (changes to N.S.F. funding for math):"
X Link 2025-05-22T19:50Z 54.8K followers, 744.3K engagements
"IMO it would be useful for the Vice President to explain why he thinks cutting math and physics funding 70-80+% advances the aims listed in this tweet. There is an extraordinary "reproducibility crisis" in the sciences particularly in biology where most published papers fail to replicate. Most universities have massive bureaucracies that inhibit the translation of basic research into commercial adoption. The voting There is an extraordinary "reproducibility crisis" in the sciences particularly in biology where most published papers fail to replicate. Most universities have massive"
X Link 2025-05-25T10:47Z 54.8K followers, 200.1K engagements
"I think its pretty unlikely that Grok 3.5/4 will be used to rewrite the entire corpus of human knowledge adding missing information and deleting errors"
X Link 2025-06-21T13:24Z 54.8K followers, 117.3K engagements
"Claude: my wife and I went antique shopping this weekend Gemini: if I cant 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 [----] why do you ask"
X Link 2025-07-08T21:23Z 54.8K followers, 189.9K engagements
"😏 looks like theres a huge amount of evidence for my conjecture that all numbers are less than 10100"
X Link 2025-07-13T17:06Z 54.8K followers, 248.1K engagements
"if you want a picture of the future imagine being fed cocomelon from the cradle to the grave You can create from scratch remix what you see or just scroll through to check out videos from the creators + the visual artists weve been collaborating with. https://t.co/M9kWNjyoEc You can create from scratch remix what you see or just scroll through to check out videos from the creators + the visual artists weve been collaborating with. https://t.co/M9kWNjyoEc"
X Link 2025-09-26T12:41Z 54.8K followers, 180.6K engagements
"It's good for academics to publicly experiment with new AI tools but important to report both successes and failures when doing so. Audience capture incentivizes only doing one or the other which is part of the reason the information environment around capabilities is so bad"
X Link 2026-01-02T19:00Z 54.9K followers, 29.4K engagements
"The basic objects of study here are algebraic varieties--shapes defined as the set of solutions to a system of polynomial equations--and polynomial maps between them"
X Link 2026-01-14T19:20Z 54.6K followers, [----] engagements
"One of the basic themes of 20th century mathematics is that the topology of the set of complex solutions to a system of polynomial equations is controlled by the arithmetic of the polynomials"
X Link 2026-01-14T19:20Z 54.6K followers, [----] engagements
"For example the Weil conjectures (proved by Dwork Grothendieck and Deligne) show that the Betti numbers of a (smooth projective) variety defined by polynomials with integer coefficients are controlled by the number of solutions to those polynomials over finite fields"
X Link 2026-01-14T19:20Z 54.6K followers, [----] engagements
"@Archivara Im really sorry to put you guys on blast again but you need to hire a subject-matter expert. The content of this paper is (1) Winograds [----] result (basically CRT) that the number of multiplications youre computing is at most 2deg-#factors and (3) 25-3=7"
X Link 2026-01-18T15:37Z 54.6K followers, 51K engagements
"I have little doubt that AI tools will substantially change the way mathematics is done likely pretty soon. But communicators like this are doing their audiences a disservice. In this particular example the new contribution from the AI tool was plugging (nm)=(53) into the expression 2n-m. Im not exaggerating. The number of multiplications in the result in question was shown by Winograd [--] years ago to be 2n-m where the matrices in question are nxn and m is the number of factors of the polynomial xn-1. The result here is that this polynomial has [--] factors over Q(sqrt(5)) and then that 2*5-3=7."
X Link 2026-01-18T23:53Z 54.6K followers, 75.5K engagements
"Toddler: I opened the fridge all by myself I tried and tried with all my might I am super STRONG"
X Link 2026-01-25T13:35Z 54.6K followers, [----] engagements
"Figured I might as well amplify this Q. @kevinroose Do you have a sense of whether people are productively using multi-agent setups for non-SWE tasks @kevinroose Do you have a sense of whether people are productively using multi-agent setups for non-SWE tasks"
X Link 2026-01-25T18:44Z 54.6K followers, 14.7K engagements
"Toddler melting down due to broken granola bar. Me: How bout I glue it together with honey. Toddler tearfully: OK Me: Did you know bees make honey Toddler: nods they make pizza too"
X Link 2026-01-25T20:10Z 54.6K followers, 238K engagements
"Toddler talking to her uncle on the phone: Where are you Uncle: Im at work. What about you Toddler: I dont work. I just play Youre SO silly"
X Link 2026-01-27T22:40Z 54.6K followers, 60.7K engagements
"Youll never guess what this is a reply to. Kind of a surprise to get blocked for it @tunguz LOL ok. @tunguz LOL ok"
X Link 2026-01-28T14:47Z 54.7K followers, 59.3K engagements
"@jakobzupanec @bayesianboy Quite a bit for literature search. Less for actual math though the latest models are sometimes useful for routine arguments"
X Link 2026-01-28T19:01Z 54.6K followers, [---] engagements
"@Allodoxaa @jakobzupanec @bayesianboy Paid models are way way better for basically any task. I mostly use ChatGPT [---] Thinking/Pro"
X Link 2026-01-29T12:01Z 54.6K followers, [---] engagements
"This was a really fun podcast to record Not only about AIalso really enjoyed sharing some thoughts on the practice of mathematics overall. How does math research change when the cost of trying your first dumb idea goes to zero @littmath joins @GregHBurnham and @ansonwhho to discuss what todays models can and cant do in math and how far they are from doing high-quality research. 0:00:00 What's the hardest math https://t.co/vxxOevyLu4 How does math research change when the cost of trying your first dumb idea goes to zero @littmath joins @GregHBurnham and @ansonwhho to discuss what todays models"
X Link 2026-01-29T21:11Z 54.6K followers, 11.1K engagements
"Small comment is that this was recorded in December of last year an eternity ago in AI time. But I think its held up well overall. Two small caveats: (1) this was recorded just before the release of GPT [---] which was a substantial improvement for math use cases. I probably would have mentioned this if it had been recorded a few days later. (2) This was also right at the beginning of the recent spate of (semi-)autonomous solutions to e.g. Erdos problems. But I think what I said about these has held up so far https://twitter.com/i/web/status/2016983497827897733"
X Link 2026-01-29T21:15Z 54.6K followers, [----] engagements
"@vladtenev @HarmonicMath Do you really believe that no Erdos problems will be left in two years"
X Link 2026-01-30T23:27Z 54.7K followers, [----] engagements
"Toddler: I want two apricots because Im two-and-a-half. Pause When I turn three I can have three apricots. Pause realization When I turn four I can have four apricots When I turn five I can have FIVE apricots"
X Link 2026-01-31T22:21Z 54.7K followers, 36.6K engagements
"Very easy to get LLMs to roleplay reddit. It even did the viral moltbook post about getting the agent to delete stuff on its own"
X Link 2026-02-01T15:54Z 54.6K followers, [----] engagements
"@Afinetheorem lol I could generate [--] tough math questions and the AI is the one that gets them right"
X Link 2026-02-04T12:58Z 54.7K followers, [----] engagements
"@Afinetheorem @JaneParkway Yeah I think it should be possible to get the models to do basically any computer use task the average person can do now though for some tasks it would likely be a huge pain"
X Link 2026-02-04T14:35Z 54.7K followers, [---] engagements
"@AdrianTMiranda I have no doubt its harder for contestants than IMO but the contestant pool is much bigger"
X Link 2025-12-08T13:06Z 54.9K followers, [----] engagements
"Part of being a mathematician is finding something you feel you cant understand and nonetheless working to understand it. Rather than claiming expertise about the role of genetics in math Murray might benefit by showing some curiosity about what it is mathematicians do"
X Link 2023-07-19T01:30Z 54.9K followers, 216.4K engagements
"Honestly dont see how the US comes back from this without serious legal consequences for members of the administration and people following their orders"
X Link 2025-04-15T15:50Z 54.9K followers, 108.5K engagements
Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing