jaimemat retweeted
Anthropic 的数学家宣布证明了 167 年未解的黎曼猜想
但他拿出来的不是论文
是 14 个代码压缩包
做这件事的人叫莱文特·阿尔珀盖
正牌的哈佛与普林斯顿数论学者
但这次他不是一个人完成的
和他一起算的是 Claude
这件事最离奇的地方
还不是他宣称攻克了数学界皇冠上的明珠
而是这篇证明根本没办法给人看
为什么不写常规论文?
因为写不出来
阿尔珀盖算了一笔账
要是把推导细节全翻译成人类能读的自然语言
这篇论文足足有 25 万页
两层楼高的打印纸
全世界所有的数学家不吃不喝看一辈子
都未必能审完
所以他干脆省掉了人类同行评审这一步
直接把证明写成了计算机代码
拆成 14 个压缩包挂在个人网站上
代码用的是专门检查数学逻辑的 Lean 语言
按照阿尔珀盖的说法
在最后的冲刺阶段
Claude 的推导速度已经远远甩开了他
连广义黎曼猜想也一并推导了出来
换句话说
这篇证明是人类开了个头
AI 在后面接力狂飙
推导出人类肉眼不可能看完的逻辑链
最后封装成机器代码
扔给另一台电脑去跑编译器
阿尔珀盖并不是网上凭空冒出来的民科
今年 7 月
他就用 Claude 找出了三维雅可比猜想的反例
掀翻了这个从 1939 年立到现在的硬骨头
两个月前
他又和纽约大学的团队合作
用 AI 攻下了纳维-斯托克斯方程的部分解
如今他把枪口直接对准了黎曼猜想
现在整个数学界没人看草稿纸
几百台服务器都在疯狂跑这 14 个代码包
大家在盯的只有一件事
这套几百万行的代码里有没有偷换公理
有没有逻辑断点
编译器能不能一路绿灯跑通
如果跑通了
人类数学史的规矩就彻底改写了
以后决定一个大猜想到底成不成立
不再是老教授们闭门审稿三年
而是电脑直接对电脑交卷
jaimemat retweeted
🚨有人在直播验证Claude证明黎曼猜想,目前进度99%!
有人专门搭了一个网站,正在公开直播验证Anthropic数学家Levent Alpöge和Claude提交的黎曼猜想Lean形式化证明。
这哥们真的下血本,直接用了128核CPU、244GB内存的服务器,下载了全部14个压缩包,总计228.2MB,里面包含175,187个Lean源文件。
目前已经运行了47分钟,完成26,088/26,329个构建任务,暂时没有任何构建失败!
网站每两秒刷新一次,整个编译过程、日志、源码校验和全部公开。
目前已经确认,Main.lean中确实声明了riemannHypothesis : RiemannHypothesis,使用的是mathlib中已有的黎曼猜想定义。
不过目前还在编译阶段,后面还有最终定理和公理依赖审计。
如果最终真的通过完整验证……
那我们可能真的要见证历史了。
这实在太疯狂了。
地址:http://106.14.11.164/
jaimemat retweeted
The most famous unsolved problem in mathematics have just fallen, and Claude helped bring it down.
For 167 years, the Riemann hypothesis beat every mathematician who tried it. Today, an Anthropic mathematician, Alpöge says he and Claude, have proved it.
He says Claude left him “in awe,” and that one of its ideas will make fellow mathematicians “fall out of their chairs.”
Alpöge says that since his student days he had secretly planned to take on this problem:
he took two very different ideas that worked on a simpler “toy” version of the problem and steered the work toward the real thing.
From there, Claude took over the advanced areas he admits he hasn’t mastered, and proposed one of the key steps, which he says will make fellow mathematicians “fall out of their chairs.”
Claude put together, the work took about six weeks, and the two of them formalized the whole argument as machine-checkable code.
Using Codex, we found a solution to Hermite's 1848 problem—178 years later!
Can every real number be encoded by a sequence that eventually repeats exactly when it is a cubic irrational?
Our six-page paper constructs one algorithm for all real numbers.
jaimemat retweeted
纽约大学库朗研究所刚成立了"数学与 AI 中心"(MAIC),官方使命里有句话值得逐字读:让前沿 AI 成为科学研究的对象,让数学界能拆解它、改进它、复现它——最终理解它。
拆解。不是"使用",是拆解。数学家把 AI 放上了解剖台。
联合主任 Joan Bruna 的原话:能做数学推理的前沿模型,目前是"难以理解的黑箱"。
另一位主任 Vlad Vicol 说:"我们需要一次重置,可能还只是许多次中的第一次。"
具体计划:联合 UT Austin 等高校组建学术联盟,造数学科学的开放基础模型。目前是路线图,模型还没发布。
这一周数学界吵的全是焦虑:期刊还剩什么、审稿谁来干、博士生还敢不敢招。
库朗的回答是却是另一种
数学家不只想拿到 AI 的答案了,他们想把给答案的机器拆开。
jaimemat retweeted
China is replacing plastic with fast-growing bamboo, creating eco-friendly bags and tableware that can biodegrade in just 4–6 months
jaimemat retweeted
Dent the surface any way you like. As long as the charge is inside, the flux out of it doesn't change.
Here's why. Follow a thin cone out from the charge. Where the surface is further away or tilted, the patch is bigger, but the field through it is weaker by exactly the same amount.
Watch the flux stay put while the surface area wobbles all over the place. Then the charge slips outside, every cone goes in and back out, and the flux drops to zero.
jaimemat retweeted
太惨了。美国物理学家 Edward Witten,今天来香港科技大学,分享黑洞热力学,结果期间习惯性地打开 Claude,被提示你所在的地区不可用 🤣
希望教授的账号,别后续被封了。
jaimemat retweeted
大家好,很高兴分享我们面向数学探索与证明的自进化智能体系统 MathRSI 🎉 目前已经一次性解决了六个开放数学问题~
不同数学问题既有各自独特的领域结构与难点,又在探索与证明上存在底层共性;而且研究越往深处推进,剩下的硬骨头越难,静态不变的流程很快就会遇到瓶颈。为此,我们构建了MathRSI, 在持续攻克前沿数学问题的过程中,把有效归约、反例与失败教训不断沉淀并反哺系统本身,让研究框架随问题难度一起演进(目前已从 v0 迭代至 v6)。
🧩 三层协同架构
• MathRSI Base(提炼共性):通用五阶段研究闭环(准备 → 路线迭代 → 证明 → 验证 → 交付),内置最难分支预检、上下界协同演化与跨模型审阅。
• Domain Plugins(深耕领域):按数学分支沉淀专用归约、失败图谱与独立核验程序,按需挂载并持续深化各领域的独有方法。
• Workspace Memory(长程记忆):单题独立维护问题规格、研究 Wiki、中间引理与断点状态,跨轮次持续读写复用,支撑深水区长程推进。
jaimemat retweeted
索邦大学数学家 Piet Lammers 刚拿到 140 万欧元的欧洲研究经费,却做了个反常的决定:暂时不再招博士生。😭
理由就一句话:“我不知道他们到底还有没有前途。”
他自己研究的问题,用他的话说,几乎全被解决了。他现有的博士生,本来盼着花几年啃一个数学难题,现在的工作变成了验证 OpenAI 给出的解答。
注意,他没遣散任何学生,钱也照拿。他只是不敢再往这条路上引新人。
另一边,阿姆斯特丹的数学家 Zuiddam 判断完全相反:AI 让数学更容易拿去用,未来可能需要更多数学家,不是更少。
两边看的其实不是同一件事——
Lammers 愁的是“怎么练出新人”:以前博士生靠啃难题出师,现在题都被 AI 做了,老路走不通了。
Zuiddam 看的是“外面有多需要人”:AI 把数学门槛拉低,用得上数学的地方变多,懂行的人反而更缺。
所以我认为AI 未必会让数学家变少,但它正在把过去那套“闭门啃几年题”的剧本彻底改写。
还有就是经费能批下来,可接下来人怎么培养,谁知道啊。
jaimemat retweeted
Researchers at Caltech and NVIDIA used AI to discover a possible singularity in the 3D Euler equations without forcing.
Now there are also rumors that the Navier-Stokes problem has been solved. What a crazy day!
We have found stable singularity on 3D Euler!
anima-ai.org/2026/09/07/stab…
Our starting point is a physics-informed neural network (PINN) to come up with an approximate answer, and then to argue stability around that to complete the proof.
The challenge so far has been that PINNs have not been successful in discovering singularities on the current problem. A common failure mode is PINNs converging to a trivial solution. We take special care to nudge our PINN to interesting regions through a combination of constraints, and we believe this is a novel way of making PINNs work for such hard optimization problems.
We also carry out a detailed study of the transport field of our approximate profile, and show it has promising properties of local outgoing flow, essential for proving linear damping, which is an important ingredient of overall stability.
We believe that such physics-informed and physics-centric AI are critical ingredients across many areas of research involving physical systems, and LLMs lack such physical grounding.
We have been working on this problem for much of this year, and we just saw the announcement by Tristan on Euler with forcing. In contrast, we consider without forcing and use a PINN formulation. @Caltech
jaimemat retweeted
Terence Tao a few hours ago on Alpöge and Buckmaster’s new math breakthrough. TL;DR:
“a remarkable achievement”
“rhe arguments contain significant Al input, although the authors have been working extensively in the past few weeks to rework them into an acceptable form.”
“they had to release far earlier than planned due to external events”
“there does not seem to be anything in principle preventing the methods from extending all the way to Navier-Stokes”
“at this point, I would not be surprised if one could batter out such an extension by pouring an enormous amount of compute and Al assistance at such a task; but such an exercise does not particularly hold my interest”
jaimemat retweeted
A Physics paper that has 5154 authors ✍️
A record-breaking publication of a physics paper having the largest number of contributors ever- 5154. The article was published in Physical Review Letters and comprises a total of 33 pages. The actual research and references are only described in the first 9 pages. The remaining 24 pages are dedicated to listing the authors and their respective institutions.
The paper is titled,
“Combined Measurement of the Higgs Boson Mass in pp Collisions at √s=7 and 8 TeV with the ATLAS and CMS Experiments”
Terence Tao now saying a number of things I've been saying for some time: 🔽 (Good, people will listen a bit more to him. But unfortunately, this won't change attitudes one bit.)
jaimemat retweeted
From the Tao’s posts on the Navier-Stokes (which made many people speculate on a possible solution and I don’t get why since, actually, his point was crystal clear), I can’t agree more with these two.
Indeed, mathematics (which exceeds its ‘practical’ uses) is not a matter of desperation -even elegance transcends urgency. On the other hand, the immense effort of solving (trying to) such a beast comes with valuable byproduct results.
jaimemat retweeted
Claude Fable 5 has produced a hand-checkable counterexample to the Jacobian conjecture, an open problem dating to 1939.
The conjecture says that a polynomial map with a constant non-zero Jacobian determinant must have a polynomial inverse.
An 87-year-old problem was casually solved by Fable 5 on a sunday evening. Insane times to live in.
hello there the jacobian conjecture is false thanx to my close friend akhil for asking about it and my other close friend fable for working during the world cup final
((1+xy)^3 z + y^2 (1+xy) (4+3xy), y + 3 x (1+xy)^2 z + 3 x y^2 (4+3xy), 2 x - 3 x^2 y - x^3 z): \C^3\to \C^3, has jacobian determinant -2, and sends (0, 0, -1/4), (1, -3/2, 13/2), and (-1, 3/2, 13/2) to (-1/4, 0, 0)
jaimemat retweeted
Si la IA hace la tarea de tu hijo, el cerebro de tu hijo no participó en eso. Y ese uso de la IA es un problema. Un estudio de Stromberg, Lei y Wu (cepr.org/publications/dp2157…) siguió a 26,000 estudiantes de secundaria en China durante 30 meses. Muestran que el uso de IA elevó las notas de las tareas en 18% y redujo el tiempo que dedicaban a esas tareas en 30%. Al mismo tiempo, las notas en los exámenes a libro cerrado cayeron 20% luego de seis meses. Analizando un efecto de largo plazo, las notas en los exámenes de ingreso a la universidad cayeron hasta 24%. Como se ve en la figura, mas eficiencia para hacer las tareas, pero menos menos aprendizaje. Por otro lado, los estudiantes que usaron IA pero igualmente dedicaron el mismo tiempo que sus compañeros sin IA obtuvieron en los exámenes notas casi idénticas a las de quienes no usaron IA. Pero los que "externalizaron" su tarea completamente, terminando más rápido y posiblemente poniendo menos esfuerzo, que cualquier estudiante sin IA, solo sacaron notas altas en los ejercicios. La diferencia fue el esfuerzo cognitivo: si el cerebro hizo el trabajo, o simplemente observó cómo la IA lo hacía en su lugar.
La IA es una herramienta. Un bisturí en manos de un cirujano salva vidas. El mismo bisturí en las manos equivocadas hace daño. La misma IA que puede ser un tutor efectivo, desafiando al estudiante a pensar, explicar, luchar productivamente con el problema, se vuelve dañina si elimina el esfuerzo que el aprendizaje requiere. Esto no es grave si le pasa a pocos estudinates. Pero cerca del 80% de los estudiantes que usaban IA cayeron en lo que el estudio llama "tercerización de tareas" (homework outsourcing).
Esto conecta directamente con la clasificación que hacíamos con mis colegas Ezequiel Molina y Maria Barron, (blogs.worldbank.org/es/latin…) de tres grupos de estudiantes, los Empoderados por la IA, que la usan para pensar con más profundidad; los Dependientes de la IA, que la usan para evitar pensar; y los Excluidos de la IA, que no tienen acceso. Nos preocupaba que el grupo de los Dependientes pudiera crecer. Ya creció. Es, con mucho, el grupo más numeroso en este estudio. Completar una tarea no es lo mismo que aprender. El cerebro no construye conocimiento observando cómo trabaja la IA sino equivocándose, esforzándose y resolviendo las cosas por sí mismo. Ese esfuerzo cognitivo es esencial para el aprendizaje .
jaimemat retweeted
¿Ya podemos dejar de usar la cabeza y pedirle ideas a la IA?
En Stanford, un jurado leyó a ciegas ideas de investigación sin saber cuáles eran de humanos y cuáles de modelos: le puso mejor nota a las de la máquina.
Entonces hicieron lo que casi nadie hace: las ejecutaron. 43 investigadores, más de cien horas de trabajo cada uno, transformando esas ideas en algo real. Al volver a evaluar, el ranking se dio vuelta: menos efectividad y menos impacto que las nuestras.
Tiene un por qué bastante claro: para sonar original, alcanza con combinar cosas que nadie había pensado antes y en eso la tecnología es imbatible. Pero saber cuál de esas mezclas no se va a desarmar al primer empujón es otra cosa. Ese criterio humano es nuestro súperpoder.
En un mundo en el que se pueden generar miles de ideas, saber identificar cuáles valen la pena es la habilidad más valiosa.
Usá la IA para pensar CON vos y no POR vos.
jaimemat retweeted
这可能是人类写给 AI 看的最后一篇论文了。
最近刷到Stanford、CMU、Michigan 等 37 位作者联名的论文:《The Last Human-Written Paper》。
核心观点很狠:沿用几百年的论文,在 AI 时代可能已经过时了。
作者点出了两个被我们忽视已久的“隐形税”:
一个是叙事税。为了讲一个漂亮故事,我们把失败实验、死路、被推翻的假设都删掉了。AI 读到的是“通关攻略”,却看不到真正有价值的“踩坑记录”。
另一个是工程税。论文里的实现细节通常足够说服审稿人,但不够让 Agent 直接复现。很多关键 tricks 还藏在作者脑子、代码注释和 Slack 记录里。
所以作者提出 ARA,直接把论文改造成 Agent 能读取和执行的“研究包”:不只告诉你结论,还把怎么想到的、代码怎么跑、证据链在哪、哪些路走不通都打包进去。
我觉得这篇最有意思的地方是,它不是在讨论 AI 怎么帮人写论文,而是在问:
当 AI 也变成论文读者和执行者时,论文还应该长成今天这样吗?
未来科研输出的核心,可能不再是“写得多像一篇 paper”,而是能不能被 AI 理解、复现、追踪和继续扩展。
人类写论文写了几百年,接下来可能要开始写给 Agent 执行的研究包了。
arxiv.org/pdf/2604.24658








































