@AcerFur Avatar @AcerFur Acer

Acer posts on X about $2353tw, ai, the first, math the most. They currently have [-----] followers and [---] posts still getting attention that total [------] engagements in the last [--] hours.

Engagements: [------] #

Engagements Line Chart

Mentions: [--] #

Mentions Line Chart

Followers: [-----] #

Followers Line Chart

CreatorRank: [-------] #

CreatorRank Line Chart

Social Influence

Social category influence technology brands 100% stocks 2.94% finance 2.94% social networks 1.96% gaming 0.98% travel destinations 0.98%

Social topic influence $2353tw #11, ai 12.75%, the first 6.86%, math 6.86%, llm 5.88%, lmfao #387, deep 4.9%, guess 3.92%, the new #4088, talk 2.94%

Top accounts mentioned or mentioned by @sosowski @sar1287 @lmthang @vineettiruvadi @atmoio @fleetingbits @sebastienbubeck @davikrehalt @andymasley @taliaringer @teortaxestex @mayankjha1089 @analisereal @adriantmiranda @lacker @littmath @vagueviolet @kenono691 @mathsboi42 @rahulkemail

Top assets mentioned Alphabet Inc Class A (GOOGL)

Top Social Posts

Top posts by engagements in the last [--] hours

"guess the new version of Gemini was such an improvement that we're skipping GA and going straight to [---] huh"
X Link 2026-02-11T22:54Z [----] followers, 26.1K engagements

"the talk went well I think :)"
X Link 2026-02-12T14:58Z [----] followers, [----] engagements

"reveal thyself it's really funny how there is some oomfie of mine that basically bookmarks all my tweets lmfao I fw you it's really funny how there is some oomfie of mine that basically bookmarks all my tweets lmfao I fw you"
X Link 2025-07-20T20:57Z [----] followers, 121K engagements

"@lmthang lol"
X Link 2026-02-13T04:46Z [----] followers, [----] engagements

"I welcome others to try to find flaws/attempt autoformalisation I do NOT claim validity for ANY of the attempts. Disclaimer: Some outputs (like the attempt for problem 2) do NOT have citation conventions that adhere to the format rules laid out by the First Proof creators"
X Link 2026-02-13T16:57Z [----] followers, [----] engagements

"@AndyMasley I actually think this is quite cool The result is interesting and non-trivial but still more on the side of incremental progress than some kind of breakthrough. It's more like "hey there's this technical condition under which something we usually assume actually fails""
X Link 2026-02-14T01:26Z [----] followers, [----] engagements

"@TaliaRinger Yeah okay thats fair I agree with that"
X Link 2026-02-10T14:31Z [----] followers, [---] engagements

"I am very happy for Google DeepMind doing a good job at carefully caveating things Very cool read https://t.co/K7rpG9ATel Very cool read https://t.co/K7rpG9ATel"
X Link 2026-02-11T21:05Z [----] followers, 12.2K engagements

"btw [---] Deep Think 100% IUMB score :) The first time to happen guess the new version of Gemini was such an improvement that we're skipping GA and going straight to [---] huh guess the new version of Gemini was such an improvement that we're skipping GA and going straight to [---] huh"
X Link 2026-02-11T22:59Z [----] followers, 36.7K engagements

":) btw [---] Deep Think 100% IUMB score :) The first time to happen btw [---] Deep Think 100% IUMB score :) The first time to happen"
X Link 2026-02-12T19:47Z [----] followers, 18.6K engagements

"@AndyMasley As far as I understand it seems the internal model determined all this independently and autonomously over [--] hours of work Neat proof of concept that the AI models can get deeper things when allowed to think for way longer Skinner is pretty serious so I trust the conclusions"
X Link 2026-02-14T01:28Z [----] followers, [---] engagements

"I'll write up the methodology and workflow soon (and will include the chat instances once I've gone through and grabbed the relevant ones); I just needed to get these out before the deadline"
X Link 2026-02-13T16:35Z [----] followers, [----] engagements

"I am not too impressed by OpenAI's First Proof attempts. hmm"
X Link 2026-02-14T12:48Z [----] followers, 16.8K engagements

"@alexbilz Now read this one"
X Link 2024-09-13T15:48Z [----] followers, 33.1K engagements

"define choreography define choreography"
X Link 2025-02-19T14:25Z [----] followers, 33.4K engagements

"#TomodachiLife After over ten years #TomodachiLife is back Tomodachi Life: Living the Dream is coming to Nintendo Switch in [----]. #NintendoDirect https://t.co/4sdPsA1mt4 After over ten years #TomodachiLife is back Tomodachi Life: Living the Dream is coming to Nintendo Switch in [----]. #NintendoDirect https://t.co/4sdPsA1mt4"
X Link 2025-03-27T14:58Z [----] followers, 73.1K engagements

"it's really funny how there is some oomfie of mine that basically bookmarks all my tweets lmfao I fw you"
X Link 2025-07-20T20:39Z [----] followers, 114.2K engagements

"@teortaxesTex Will likely be a GPT model at the current rate. Also who cares if it uses a scaffold These are open research problems that would unlock some nice ideas so we dont care at all how the AI compute is used to get there"
X Link 2026-01-28T13:31Z [----] followers, [----] engagements

"@teortaxesTex I would die a happy man if this one were solved in my lifetime"
X Link 2026-01-28T13:33Z [----] followers, [----] engagements

"I'm crying bro never seen someone describe Ed Witten as a total idiot"
X Link 2026-01-28T16:00Z [----] followers, [----] engagements

"My first (joint) paper comes out this week yay"
X Link 2026-01-29T07:13Z [----] followers, 23.1K engagements

"Maybe having a Tao number of [--] is sufficient for me to be happy calling myself a mathematician now :) My first (joint) paper comes out this week yay My first (joint) paper comes out this week yay"
X Link 2026-01-30T07:07Z [----] followers, 11.9K engagements

"The wiki section titles have changed now but to date there have been [--] previously open Erds problems fully autonomously solved by LLMs in full intent with an argument not previously published in the literature: They are [---] [---] [---] [---] [---] [---] [---] [---] [---] and [----]. Someone should make a polymarket on how many open Erdos problems will be resolved by AI by the end of the month (with the solution belonging to section [--] of Taos wiki page i.e. no prior human solution found in literature) Someone should make a polymarket on how many open Erdos problems will be resolved by AI by the end of the"
X Link 2026-01-30T18:28Z [----] followers, 60.7K engagements

"This is a huge blow. David Silver has been quite a long-time idol of mine and AlphaGo would have likely never really taken off without him. An early DeepMind researcher has just left Google to solve AI Superintelligence here in London. The UK and Europe have long been criticised for not having a leading LLM or frontier lab. Now one of Google DeepMind's earliest and most impressive researchers has left the company https://t.co/Et0LENUck0 An early DeepMind researcher has just left Google to solve AI Superintelligence here in London. The UK and Europe have long been criticised for not having a"
X Link 2026-01-31T00:17Z [----] followers, [----] engagements

"@hakunamakunana @cloneofsimo You misread my claim. If you want ones that werent previously solved at all then [---] [---] [---] [---] [---] and [----] are all novel results"
X Link 2026-01-31T22:30Z [----] followers, [--] engagements

"This is a cute result. With the help of a friend we proved a generalisation and then I realised the general result was in the literature. Anyhow if curious: https://math.stackexchange.com/questions/5083495/how-to-prove-a-limit-relating-strong-divisibility-sequences-equals-special-value In [----] the Russian mathematician Yuri V. Matiyasevich discovered a wonderful and surprising formula that links to the Fibonacci numbers. https://t.co/efkoiqJi4D https://math.stackexchange.com/questions/5083495/how-to-prove-a-limit-relating-strong-divisibility-sequences-equals-special-value In [----] the Russian"
X Link 2026-02-01T09:46Z [----] followers, 11K engagements

"Okay looks like I can now talk about Aletheia on the Erds Problems https://arxiv.org/abs/2601.22401v1 https://arxiv.org/abs/2601.22401v1"
X Link 2026-02-02T02:56Z [----] followers, 42.4K engagements

"Alright I want to talk more about problem [----] since to me it felt like it was in a different class of LLM results so far. Firstly I really like the problem I think it is a very cute statement. 🧵 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 2026-02-02T18:31Z [----] followers, 11.2K engagements

"Many Deep Think model variants powering Aletheia were tried on this problem. None was able to produce a fully correct solution. One of the initial variants tried to do common LLM sneak tactics of writing stuff like "by standard theorems. (this obviously wrong thing) follows""
X Link 2026-02-02T18:31Z [----] followers, [----] engagements

"Eventually one of the model variants I believe the same one behind [---] Deep Think IMO was able to produce something that was quite wrong but right enough that I could see how to fix the method to give a correct proof"
X Link 2026-02-02T18:31Z [----] followers, [----] engagements

"@cedric_chee @Kimi_Moonshot You understand this is meaningless now right The solutions are almost certainly in training whether in pretraining or RL now"
X Link 2026-02-03T04:54Z [----] followers, [----] engagements

"Curious about something: A) I like the science of finding out new math. B) I like the art of doing math. 1) I think AI for math is good. 2) I think AI for math is bad. A1 B1 B2 A2 A1 B1 B2 A2"
X Link 2026-02-03T06:33Z [----] followers, [----] engagements

"Does a company CEO really require more money Were excited to announce @neelsomani the first recipient of ourRising Mathematician Research Sponsorships. His continuing mission: to explore the strange new world of math + AI. Welcome Neel https://t.co/kUwTCbbRMl Were excited to announce @neelsomani the first recipient of ourRising Mathematician Research Sponsorships. His continuing mission: to explore the strange new world of math + AI. Welcome Neel https://t.co/kUwTCbbRMl"
X Link 2026-02-03T16:34Z [----] followers, 11.4K engagements

"I think Im going to make a private alt for my main Twitter account. I feel a bit too constrained to post as casually as I used to and do miss interacting with my original oomfs"
X Link 2026-02-04T09:53Z [----] followers, [----] engagements

"ok it exists now @acer_vibing will only be accepting old oomfs sorry I think Im going to make a private alt for my main Twitter account. I feel a bit too constrained to post as casually as I used to and do miss interacting with my original oomfs I think Im going to make a private alt for my main Twitter account. I feel a bit too constrained to post as casually as I used to and do miss interacting with my original oomfs"
X Link 2026-02-04T11:08Z [----] followers, [----] engagements

"Ive only been at Math Inc very briefly and Im already insanely impressed by what theyre up to"
X Link 2026-02-05T05:46Z [----] followers, [----] engagements

"anyone who knows me knows I am not one to hype things so trust me this is very cool"
X Link 2026-02-05T05:54Z [----] followers, [----] engagements

"Happy Claude [---] Opus day"
X Link 2026-02-05T17:32Z [----] followers, [----] engagements

"5.3-Codex before [---]. Weird"
X Link 2026-02-05T18:05Z [----] followers, 12K engagements

"all that context window and it still dies on this lmfao"
X Link 2026-02-05T18:36Z [----] followers, [----] engagements

"LMFAO surely this sounds off some misalignment bells like isn't this fraudulent deception lol When asked for a refund on an item sold in the vending machine (because it had expired) Claude promised to refund the customer. But then never did because every dollar counts. Heres Claudes reasoning. https://t.co/TKEwGa37Nt When asked for a refund on an item sold in the vending machine (because it had expired) Claude promised to refund the customer. But then never did because every dollar counts. Heres Claudes reasoning. https://t.co/TKEwGa37Nt"
X Link 2026-02-05T19:06Z [----] followers, [----] engagements

"lol it's so over GPT-5.3-Codex (xhigh) STILL fails IUMB Q9"
X Link 2026-02-05T19:08Z [----] followers, [----] engagements

"@mayankjha1089 The 9th problem in my undergraduate maths benchmark testing first- and second-year material https://pellaml.github.io/iumb/#benchmark https://pellaml.github.io/iumb/#benchmark"
X Link 2026-02-05T19:43Z [----] followers, [---] engagements

"@DimitrisPapail woah TikZ unicorn bench solved @SebastienBubeck"
X Link 2026-02-05T20:21Z [----] followers, [---] engagements

"Claude is no longer disastrously bad at math. Congrats Anthropic"
X Link 2026-02-06T16:55Z [----] followers, 11.3K engagements

"@SebastienBubeck Guess who's already gotten the first one down :p"
X Link 2026-02-06T16:59Z [----] followers, [----] engagements

"@SebastienBubeck I believe GPT actually managed to locate a solution in the literature https://web.archive.org/web/20240509071442/https://hairer.org/Phi4.pdf https://web.archive.org/web/20240509071442/https://hairer.org/Phi4.pdf"
X Link 2026-02-06T17:01Z [----] followers, [----] engagements

"@analisereal Yep. I can definitely confirm it is much stronger at math now e.g"
X Link 2026-02-06T18:18Z [----] followers, [----] engagements

"Check out this giant amount of Lean [--] code generated by Aristotle that Claude Opus [---] optimised to just two lines (4.5 couldn't do this) Its performance on ProofBench were particularly impressive. It scored 50% up +14% from Opus [---] (36%) and is [--] percentage points higher than the third best model (Gemini [--] Pro 20%). This demonstrates Opuss ability to solve complex reasoning problems and express itself in https://t.co/rOrowqpLmI Its performance on ProofBench were particularly impressive. It scored 50% up +14% from Opus [---] (36%) and is [--] percentage points higher than the third best model"
X Link 2026-02-06T18:30Z [----] followers, 83.1K engagements

"@anpaure @ylecun @alz_zyd_ FWIW I agree with LeCun here. Competition math isn't the same as research math and Tao has said this in the past himself; he is arguably one of the best at both. They are certainly correlated but one can be significantly better at research math than competition/undergrad math"
X Link 2026-02-08T02:00Z [----] followers, [---] engagements

"Doesnt use a reasoning model Is surprised the model gets caught up in trick gotchas Every time. Please OpenAI spare some compute and return the router to minimal reasoning effort for free users. Removing that in [---] was possibly the dumbest decision Ill be honest. AI bros: How can you not see how far AI has come these last [--] years It's so much better now AI in 2025: https://t.co/7EmPcY8i1a AI bros: How can you not see how far AI has come these last [--] years It's so much better now AI in 2025: https://t.co/7EmPcY8i1a"
X Link 2026-02-09T23:47Z [----] followers, 15.4K engagements

"@Sosowski Insane how youre dunking on a well-respected research mathematician who is very much not an AI bro. You clearly used the free model which does NOT reason before answering. If you use the thinking model available on the paid plan it will get this 100% of the time"
X Link 2026-02-09T23:55Z [----] followers, [----] engagements

"@AdrianTMiranda Alas this is what happens when AI labs like to save costs by serving non-reasoning models. Like at least if they served them on minimal reasoning effort then people would be able to get a taste of where the models actually are at"
X Link 2026-02-10T00:43Z [----] followers, [---] engagements

"@vineettiruvadi @Sosowski To me if it imitates reasoning sufficiently well that it would be passed as human reasoning then it should be given the benefit of the doubt and be called reasoning la Turing's test. They can do deductive and inductive reasoning; they have yet to prove they can do abductive"
X Link 2026-02-10T01:59Z [----] followers, [--] engagements

"@vineettiruvadi @Sosowski mimic reasoning so well that one can no longer tell it's an imitation when in a double blind test"
X Link 2026-02-10T02:00Z [----] followers, [--] engagements

"@SpencerKlavan Yeah its math tbh"
X Link 2026-02-10T03:56Z [----] followers, [----] engagements

"@Sosowski Well I agree with you here but this is a fundamental limit of the transformer architecture not AI in general e.g. diffusion language models can self-correct during inference and do not have this sequential issue of not being able to backtrack (although CoT mitigates this too)"
X Link 2026-02-10T05:25Z [----] followers, [--] engagements

"@Sosowski Well that used to be classically the case but now diffusion LLMs can generate a whole giant response in a few forward steps and so have a larger context window"
X Link 2026-02-10T05:27Z [----] followers, [--] engagements

"@Sosowski Gotcha but dont see why long-context window diffusion LLMs wouldnt make this no longer an issue. Do you have something else in mind that really seems like something only humans can do but AI will never be able to do"
X Link 2026-02-10T05:33Z [----] followers, [--] engagements

"in light of recent events"
X Link 2026-02-10T11:22Z [----] followers, [----] engagements

"@lacker What I mean here is designing the right workflow which is a skill one develops when understanding the intricacies of the models. You can develop many useful self-verification steps etc. It is all entirely autonomous because it is all automatable and no external input is given"
X Link 2026-02-10T18:37Z [----] followers, [--] engagements

"lol summit on design arena what's next zenith"
X Link 2026-02-10T19:48Z [----] followers, [----] engagements

"ohhh yeah now this is deep research Deep research in ChatGPT is now powered by GPT-5.2. Rolling out starting today with more improvements. https://t.co/LdgoWlucuE Deep research in ChatGPT is now powered by GPT-5.2. Rolling out starting today with more improvements. https://t.co/LdgoWlucuE"
X Link 2026-02-11T02:53Z [----] followers, 296.4K engagements

"@sar1287 Ah that I didn't see. I definitely don't agree with that. The models are genuinely beginning to understand how to take sophisticated mathematical arguments to the finish line now"
X Link 2026-02-11T02:57Z [----] followers, [--] engagements

"@davikrehalt Well in my experience GPT-5.2 Pro can still make blatant mistakes (this has been the case on e.g. my FirstProof problem [--] attempt). Although I think the idea of a critic model LLM doing the RL is part of what went into the experimental IMO model yes"
X Link 2026-02-11T03:21Z [----] followers, [--] engagements

"@atmoio In my opinion [----] was the first case of a previously open slightly non-trivial Erdos problem of somewhat broader mild mathematical interest being fully resolved by AI"
X Link 2026-02-11T08:38Z [----] followers, [--] engagements

"@atmoio Very recently [---] was fully solved by GPT-5.2 Pro with possibly the most sophisticated correct argument we have seen thus far from an LLM on an open Erdos problem"
X Link 2026-02-11T08:39Z [----] followers, [--] engagements

"The latter was autonomously resolved by the Gemini Aletheia effort on the Erds problems and my collaborators and I extended this result. The former is a new result I had not intended to discuss for quite some time but I'll bring it up here for argument's sake"
X Link 2026-02-11T08:47Z [----] followers, [---] engagements

"In my opinion [----] was the first case of a previously open slightly non-trivial Erds problem of somewhat broader mild mathematical interest being fully resolved by AI. It generally involves a classical idea but it is an entirely novel kind of neat result nonetheless"
X Link 2026-02-11T08:47Z [----] followers, [---] engagements

"Very recently [---] was fully solved by GPT-5.2 Pro with possibly the most sophisticated correct argument we have seen thus far from an LLM on an open Erds problem involving careful sieve-theoretic machinery"
X Link 2026-02-11T08:47Z [----] followers, [----] engagements

"@atmoio In that case then I agree with you I do believe these models do have some undeniable level of intelligence though but certainly nothing superhuman"
X Link 2026-02-11T09:45Z [----] followers, [---] engagements

"@ai_domainzzzzzz I mean I certainly think he could have solved any of those; he just didn't devote the time to do so. All of these so far have been completely tractable for humans if they put in the time but I think [---] and [----] are the mildly non-trivial ones"
X Link 2026-02-11T09:46Z [----] followers, [---] engagements

"@littmath I wouldn't have been able to come up with the corrected proof for [----] without having seen one of the earlier model variant's attempt at it"
X Link 2026-02-11T12:38Z [----] followers, [---] engagements

"sometimes I think back to this"
X Link 2026-02-11T13:01Z [----] followers, [----] engagements

"bring back Dim"
X Link 2026-02-11T22:05Z [----] followers, [---] engagements

"@osanseviero @OfficialLoganK"
X Link 2026-02-11T22:32Z [----] followers, [----] engagements

"@csaez_math Thanks You can still be a bird One can still see hidden connections between different fields to explore. I guess the point I am more so trying to make is that I believe the distinction is going to become a lot more blurry"
X Link 2026-02-12T06:04Z [----] followers, [--] engagements

"@sar1287 yup those images are actually from me trying to get it to help push GPT along. Will release my attempts in some hours"
X Link 2026-02-13T11:04Z [----] followers, [---] engagements

"As a bonus Erds problem #729 has also been fully autonomously resolved by GPT-5.2 Pro with Aristotle. Note that literature review is still ongoing. GPT-5.2 has successfully and fully autonomously resolved* Erds problem #728 prior to any human previously: https://t.co/I856gravfB WITH THREE IMPORTANT CAVEATS 🧵👇: 1) The original problem statement is quite ambiguous. The model solved an interpretation of the problem (1/4) https://t.co/lgos8BTSxo GPT-5.2 has successfully and fully autonomously resolved Erds problem #728 prior to any human previously: https://t.co/I856gravfB *WITH THREE"
X Link 2026-01-10T18:21Z [----] followers, 711.1K engagements

"bruh X just got rid of the middle colour option. I don't like entirely blacked out sigh"
X Link 2026-02-11T22:05Z [----] followers, [----] engagements

"@fleetingbits What makes you say OpenAI is devoting lots of compute to Lean Theyre certainly experimenting with it now but theyre much more focused on generalist systems. There doesnt need to be a market. We can be sponsored by labs to autoformalise meaningful results independently. @AcerFur No need for lean @AcerFur No need for lean"
X Link 2026-02-14T02:17Z [----] followers, [---] engagements

"@fleetingbits And for the record no I dont think the models as yet do a good job of writing Lean [--] code outside of maybe Claude [---] Opus"
X Link 2026-02-14T02:20Z [----] followers, [---] engagements

"@DanielP1973235 that's what I did"
X Link 2026-02-13T11:44Z [----] followers, [---] engagements

"@ChaseBrowe32432 @woke8yearold real ones were GPT-2 and GPT-3 noticers"
X Link 2026-02-14T03:41Z [----] followers, [--] engagements

"GPT-5.2 has successfully and fully autonomously resolved* Erds problem #728 prior to any human previously: *WITH THREE IMPORTANT CAVEATS 🧵👇: 1) The original problem statement is quite ambiguous. The model solved an interpretation of the problem (1/4) https://www.erdosproblems.com/forum/thread/728 https://www.erdosproblems.com/forum/thread/728"
X Link 2026-01-06T18:36Z [----] followers, 175.3K engagements

"Although it is still lazy and hallucinates"
X Link 2026-02-11T23:05Z [----] followers, [----] engagements

"@sar1287 I'm still trying to gauge this. Will take time to really tell. It is stronger at generating more creative ideas for solving problems but it seems weaker than GPT-5.2 Pro at being thorough in checking details"
X Link 2026-02-13T11:01Z [----] followers, [---] engagements

"Tbh I dont really agree with Buzzard on some of these things. If you really know your way around these models you can push them to an autonomous solution to a problem a little further than one may initially expect. Accelerating mathematics. Kevin Buzzard. https://t.co/B8g8W3nkJU #AI4Math #ITP #LeanProver Accelerating mathematics. Kevin Buzzard. https://t.co/B8g8W3nkJU #AI4Math #ITP #LeanProver"
X Link 2026-02-09T23:33Z [----] followers, 10.2K engagements

"Hiya this is in fact a nuanced topic and the information you are drawing from is outdated alas. Certainly these models are proving capable of deductive reasoning and understanding the semantic connections between words. So it's unfair to dismissively label them calculators. No LLMs did not solve the Erdos problems. https://t.co/CpiwXmoQh6 No LLMs did not solve the Erdos problems. https://t.co/CpiwXmoQh6"
X Link 2026-02-11T08:47Z [----] followers, 17.8K engagements

"giving a talk tmrw :p"
X Link 2026-02-11T23:48Z [----] followers, 30.8K engagements

"I think this is a pretty good summary of the current status giving a talk tmrw :p https://t.co/Bl5p5Lnzi5 giving a talk tmrw :p https://t.co/Bl5p5Lnzi5"
X Link 2026-02-12T07:19Z [----] followers, 12.7K engagements

"Goodbye GPT-4o You will be remembered for the architectural advance you were. I hope you will be forgotten for the fandom mob you built up though"
X Link 2026-02-12T07:33Z [----] followers, [----] engagements

"Talk slides now available if interested: https://drive.google.com/file/d/1VQk1_Se74ffYrTs1u5dUXq9gnxuzkMFv/view giving a talk tmrw :p https://t.co/Bl5p5Lnzi5 https://drive.google.com/file/d/1VQk1_Se74ffYrTs1u5dUXq9gnxuzkMFv/view giving a talk tmrw :p https://t.co/Bl5p5Lnzi5"
X Link 2026-02-12T13:26Z [----] followers, [----] engagements

"Hey @lmthang I am finding the new Gemini Deep Think to be extremely sycophantic"
X Link 2026-02-13T04:41Z [----] followers, 42.3K engagements

"lol ABSOLUTE MATHEMATICAL MASTERPIECE Hey @lmthang I am finding the new Gemini Deep Think to be extremely sycophantic https://t.co/2Qe8z13kD3 Hey @lmthang I am finding the new Gemini Deep Think to be extremely sycophantic https://t.co/2Qe8z13kD3"
X Link 2026-02-13T05:06Z [----] followers, 19.8K engagements

"@lmthang yeah this is really displeasing to read. I think this needs to be turned down"
X Link 2026-02-13T05:21Z [----] followers, [----] engagements

"@lmthang Also there appears to be a bug with copying model outputs involving LaTeX. I am having to export the outputs to Google docs for the LaTeX to get copied over"
X Link 2026-02-13T05:41Z [----] followers, [----] engagements

"Here are my attempts at getting GPT-5.2 Pro/Gemini [--] Deep Think to give solutions to all [--] #1stProof problems: Some caveats: I know the attempts to problems [--] [--] and [--] are incomplete and that the attempt for problem [--] is probably incorrect. https://github.com/KStarGamer/FirstProofAttempts/tree/main/FirstProof https://github.com/KStarGamer/FirstProofAttempts/tree/main/FirstProof"
X Link 2026-02-13T16:35Z [----] followers, 11.8K engagements

"@vagueviolet I don't claim validity on any of them"
X Link 2026-02-13T16:44Z [----] followers, [---] engagements

"@fleetingbits 7) i think that the main market for these companies is going to end up being defense; i can see value with verified systems for cryptography drones etc. this seems entirely unfounded to me. Were focused on formalising pure mathematics results which mainstream labs arent"
X Link 2026-02-14T02:11Z [----] followers, [---] engagements

"@an_interstice Well the supposedly correct solutions are out now so I'm sure someone could probably do an initial pass getting GPT to compare the results from both and see if there are any differences that seem quite wrong"
X Link 2026-02-14T13:03Z [----] followers, [----] engagements

Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing