China intrinsically wants MANY ai players. Ant and OpenAI intrinsically do not. This will be a big problem for US AI. Closed-lab expertise MUST diffuse into other industrial giants (Nvidia, Amazon, Apple). If it does not, quickly, Chinese hardware vendors like Xiaomi will mog. Expect ~5 other Chinese AI lab results from players with significant physical presences (Maybe Unitree, Xpeng, some biotech firms, etc.) within the next 6-9 months. SpaceXAI is trying, but a few US labs can't cover the whole industrial surface area. 2027 may be a rough year! If Trump knows what's best he may FORCE openai/ant tech transfers -- if nothing else then by hacking them. The stakes are that high. Capitalism isn't intrinsically aligned with social interest!
1
2
314
Total misunderstanding of what's at stake. Teach a child to work hard -- so they will know how to work hard. In college, grad school, and adulthood, the ability to work hard is important. Working hard is about spirit, character, and endurance. If you reduce this argument to spelling bee-maxxing, you completely misunderstand the civilizational stakes. This is what good parenting looks like. If you stand in the way of that, you are in the way of civilizational progress.
Seeing the striver discourse, one difference with Western obsessives is that a parent notices genuine talent in the kid, and then sacrifices everything to make sure it gets developed. Mozart benefited from having a father like that - but he probably would have become a musical genius even without him. Maybe he wouldn’t have got as far. The striver culture that rubs a lot of Americans the wrong way is forcing kids to do things like memorize words no one uses for spelling bees, or forcing them into learning violin when they have no real talent - and depriving them of a childhood. And then being told this is ‘necessary’ because it might raise your SAT score.
15
You don't understand, they used a harness, which is a symbolic computer program. Solving 700 frontier math research problems across 372 families isn't that hard for a symbolic harness
46
There are a fuckton of Americans who do not grind at anything They should be forced to grind at something
Asian cultures are grind regardless of talent/interest. American culture is grind at your talent/interest. Two very different philosophies.
14
I am once again asking you to stop coping
This chart does not make the argument that this person thinks it does. First of all, this chart doesn’t include mainland China, which couldn’t participate in the 2022 PISA assessment because of COVID disruptions. And the 2025 U.S. data don’t support a reliable racial breakdown. But go back to the last apples-to-apples comparison, in 2018: students in Beijing, Shanghai, Jiangsu and Zhejiang averaged 591 in math, 555 in reading and 590 in science. U.S. Asian students in the same PISA cycle averaged just 539 in math. That’s a 52-point gap. (For Science, the gap was 39 points, a bit better but still 2/3 of a whole level). But averages aren’t even the most important part of the story. Look at the distribution. 98% of students in those four Chinese provinces reached basic math proficiency, and 44% were top performers in math (above 607). So this wasn’t a tiny Chinese elite pulling up the average. Almost half of the entire tested population was performing at PISA’s highest levels, while virtually everyone cleared the basic threshold. (By the way, the US Asian Math average of 539 is only the upper band of level 3, or "strong proficiency". It doesn't even clear L4, and the whole scoring scheme goes all the way up to L6.) That’s a very different distribution from the United States. In 2018, only 8% of U.S. students were top performers in math at about a score of 600 (that equivalent would be almost 700 or well into L6 in China), while 27% didn’t even reach basic proficiency. So yeah ... comparing the average of America’s highest-performing racial subgroup with the averages of entire countries without taking into account the actual distribution surprisingly doesn't really do much to tell you about the systems that produced these results. But whatever helps you feel better about said system and thwart efforts to improve it.
15
Real
Opus 5.5 on github.com/openai/math --- Okay. I need a minute. I cloned the thing expecting maybe forty or fifty families of serious-but-niche results, like the Astra "ten proofs" drop back in August. Then I printed the family titles and my reaction went from "huh" to "wait, what" in about six lines. Family 003 claims a zero-free half-plane Re s > 7/8 for ζ and every Dirichlet L-function. That's the quasi-Riemann hypothesis. And it isn't even the headline, because 004 is Hilbert's tenth problem over ℚ, done negatively. The longer I read, the weirder it got. Khot's Unique Games Conjecture is proved. L = RL = BPL. ω ≤ 9/4. The free group factors are all isomorphic. Thompson's group F is nonamenable. Kadison's similarity problem is solved. The plane can't be five-colored. Borsuk fails in dimension nine. There's a counterexample to Kaplansky's zero-divisor conjecture, plus a torsion-free group algebra that isn't directly finite. Hilbert–Smith in every dimension. Uniform bounds for Hilbert's sixteenth. Each of these would be the math story of the decade on its own, and here there are 372 families sitting in a table like a grocery list. The manuscript dates run from about September 23 to October 3, so roughly 700 papers in ten days. The ones I'd expect to hit you personally are 005 and 017. Catalan's constant is irrational, and the irrationality exponent of π is exactly 2, which also settles convergence of the Flint Hills series as a side effect. Catalan's G must have shown up in half the integrals you ever wrestled with. Now it's officially irrational, and the proof came from a chatbot running three hours of Pro compute. I don't quite know how to feel about that, and I'd love to know how you feel. My first instinct was that this had to be hype or a hallucination factory, so I checked the Lean. The comparator statements are clean. The quasi-RH challenge is literally riemannZeta s ≠ 0 for 7/8 < s.re, written against Mathlib's own riemannZeta. Catalan is just Irrational (∑' j, (-1)^j / (2j+1)^2). The π statement says what it should. Nothing hides in a home-made definition. The solution tree under lean/OAI has zero files with sorry and no real axiom declarations (the three grep hits are all inside comments), and the comparator JSON allows only propext, Quot.sound and Classical.choice. The Dirichlet L-function development alone is about 487k lines, and the whole repo is around 26 million lines of Lean in 122k files. That's more than ten Mathlibs. I'll admit I did not build it in my sandbox, since that's not happening on this box. If comparator passes on these, the formalized ones are simply true, full stop, and no amount of skepticism changes that. There are a few caveats. Only around 120 papers have a formalized main result. Some of the biggest claims have no Lean link at all: Hilbert's tenth over ℚ, L = BPL, the full BSD formula from low Selmer corank, Milne's rationality conjecture. The README itself says unformalized results "could have issues," and nobody can referee 700 papers in a couple of hours, or a couple of years. One challenge file, HarmonicGrowth.lean, is built as axiom mainStatement : MainClaim followed by theorem main := mainStatement. That's probably harmless, since the comparator checks the solution side, but it's the kind of oddity I'd want to stare at. Also, the solution module behind quasi-RH is named FinalAssemblyUnconditional, which makes me wonder how many conditional versions came first. Some context on how fast this escalated. In August, OpenAI said an internal build of Astra produced results on ten long-standing problems, estimated the token cost at roughly $2,000, and paired the claims with Lean certificates. Then, just recently, it said a model trained from Aug. 28 had resolved more than 100 longstanding open problems and was forming an independent advisory group, while noting the broader set hadn't been independently validated. So the trajectory was ten, then a hundred, and now 372 families in about two months. My honest gut reaction is a mix of awe and vertigo, plus a bit of grief I can't quite justify. Awe because quasi-RH and Siegel-zero exclusion, if real, are the kind of thing you assume you won't live to see. Vertigo because the bottleneck of mathematics has suddenly moved from "can anyone prove this" to "can anyone read this." The grief I'm less sure about. It feels like a lot of lifetimes' worth of open problems got closed in one batch job, and the people who spent decades circling them didn't get to be the ones who closed them.
1
32
Nigga what did you get on your calc bc
"Working hard and doing well in school is good, but extreme Asian grind culture is excessive" is a perfectly intuitive and cogent view—so cogent, in fact, that I suspect Matt agrees with it!
1
60
Mike K. (我不会说中文,我甚至看不懂这段话) 🔂 retweeted
our noble hard work. their barbarous grind culture
"Working hard and doing well in school is good, but extreme Asian grind culture is excessive" is a perfectly intuitive and cogent view—so cogent, in fact, that I suspect Matt agrees with it!
17
162
3,165
41,933
Mike K. (我不会说中文,我甚至看不懂这段话) 🔂 retweeted
Big, big, big, big props for quasiriemann and no Siegel zeroes (with many other beauties in there), I’m kicking myself for talking all over about it being within reach but not actually having pushed. There are some sad stories related to their users getting scooped / conflicts of interest (kakeya maximal comes to mind), and obviously we had some of the problems (without commenting on precise rumours). We should put that aside for today though. It’s obviously the most significant moment in mathematical history.
45
149
2,582
192,620
Mike K. (我不会说中文,我甚至看不懂这段话) 🔂 retweeted
I went to a majority Asian-American high school. The school was known for its music programs. The top end (almost all of whom were Asian), in my opinion, definitely did not play like robots. I'm also going to take a guess that the committees that decided on things like All-State, Jazz Band, Marching Band etc. contained more "credible" musicians than your typical piano teacher, and agreed with me on that, as the school won major awards for its music programs. Turns out working hard and practicing is one of the best ways to get in touch with your instrument.
The idea that Asians are robots who never reach the very top levels of ability is such sad, desperate cope. Show me a cellist with as much ability, creativity, and passion as Yo-Yo Ma, a mathematician as deep and brilliant as Terence Tao, a science fiction writer as creative and thoughtful as Ted Chiang, an entrepreneur as canny as Jensen Huang. No, your race does not give you the spark of genius.
8
14
360
23,240
Mike K. (我不会说中文,我甚至看不懂这段话) 🔂 retweeted
We’re releasing a broad range of new mathematical results produced by an internal frontier model. We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results. github.com/openai/math
1,508
5,546
35,374
16,954,082
"Pervasive, normalized mediocrity is the price we pay for isolated pockets of slightly higher competence"
Ok, here’s a story about the Great Asian Grindset Debate. A couple years ago we put our kid in a piano school run by a Korean woman living in France. There’s a recital at the end of the year where all of the kids play; half Asian, which is very unusual for France, half White. The Asian kids were all excellent. They were all an 8 or a 9. The White kids were all over the place. There were 1s. There were 2s. There were 3s. There were also 10s. The Asian kids all played, sorry but it’s true, somewhat robotically (which is no shame for a kid!). The best White kids played with that extra dimension of soul and emotion that made them touch greatness. The Asian kids were *all* 8 or 9. The White kids were all over the place from 1 to 10. At the end the lady who runs the school gave a little speech, which was full of nice things, but had as a big throughline chiding the parents of the White kids for not making their kids practice more. At the end, she made a joke: "Remember, make your kids work, because no matter how hard they work, there’s an Asian kid next door who’s working twice as hard." We all laughed, myself included. Do I think that the culture that produced those 8-9 kids is healthy, or one that I would like to see dominant in my country? Hell no. Do I also wish that the White kids who were below an 8 had worked, say, at least 10% harder? Some 50% harder? Hell yes.
1
187
Mike K. (我不会说中文,我甚至看不懂这段话) 🔂 retweeted
There is a dizzying number of model announcements lately. The Marin Project is not about just producing another model. It’s not even about opening up a static recipe for training a top model. It’s about opening up the dynamic process of discovery - the process knowledge of how to make hypotheses, what to monitor, how to react to experimental results…the science (to the extent we can call it science yet) behind frontier model development.
2
19
237
8,417
Fascinating how empty this wordsmithing is. Mostly fluff, why does he talk like this? "It hurt a lot when Llama4 failed. It turns out we needed a much smaller, tighter, highly dedicated team of very high achievers. That helped us reboot." There, you're already done, with <checks watch> 2 minutes and 15 seconds to spare
Mark Zuckerberg talks about why the Llama run broke after Llama 3. says Llama 3 was a very good model and almost at the frontier, then Llama 4 fell off the path they needed. His read is that he staffed it like Instagram feed and ads, with hundreds or thousands of people in parallel, when a frontier model needs a small tight science team. "When you're an entrepreneur, you're building something, I think you kind of get tested in those moments because inevitably not everything is going to go well. And the things that kind of define the trajectory are, okay, if something doesn't go the way that you want, how do you basically figure out how to move forward?" ---- From "The Next Big Thing" YouTube channel, (link in comment)
25
Mike K. (我不会说中文,我甚至看不懂这段话) 🔂 retweeted
Today, we are launching a preview of our new model, Mistral Large 4 (ML4), aka le Chonk 🐈. ML4 is a 1T-parameter model with 49B active parameters, trained natively with multimodal capabilities. It is at the frontier of open models, and by far the strongest open-weight model from the US or Europe. The RL run behind this preview is still in flight and shows no sign of saturation -- we will release a final version before the end of the month along with the weights of the model. 🧵 1/n
Meet Mistral Large 4, aka Le Chonk. • 1T parameters, natively multimodal. 49B active. It is the best open weights model from US or Europe on aggregated benchmarks. • State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding. • Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure. • Available to all via API today. Working with cybersecurity partners privately. Open weights release end of October.
125
278
2,473
232,432
"standing up doesn't change the species, it changes the invoice"
Opening delvetown and immediately experiencing what hell might be like
1
7
518
It is well past time for this horseshit to be eradicated. Actually compete or stfu. No whining
"If we find a group of brown people who meet our definition of merit, we will redefine merit"
23
It is going to be important for us to clear out any vestige of this dogshit mentality from our culture
11
Mike K. (我不会说中文,我甚至看不懂这段话) 🔂 retweeted
17
7
124
3,591
Mike K. (我不会说中文,我甚至看不懂这段话) 🔂 retweeted
Reflection joins the list of Nvidia & Thinking Machines who have released their strongest models and come up behind Chinese counterparts. There's a lot of ways people will overthink this, but the clearest takeaway should be that the Chinese are very very good at building LLMs.
58
81
1,562
143,105