Mathematician with wide interests. My job is at Barnard in NYC. Talk to me! Twitter is for fun 🦋
NYC
Joined June 2009
- Tweets1.3K
- Following866
- Followers2.7K
- Likes802
Pinned Tweet
I’ve made a visualization of the counterexample to the jacobian conjecture which involves high school math, three lines and two parabolas. In fact this diagram can be easily be turned into a proof that the construction gives a counterexample. Enjoy! jacobiancounterexample.net @__alpoge__
How can there simultaneously be a pull request to math lib from Kevin Buzzard on showing that the unit sphere is a Riemannian manifold [1] from Sept 26, and a formalization of Ricci flow [2] from Sept 28??? [1] github.com/leanprover-commun… [2] github.com/qinz1yang/differe…
Semon Rezchikov retweeted
Don't Cry - BUILD! 🦾
Leading China’s mathematicians joined forces in Beijing on Sunday to launch an alliance aimed at integrating artificial intelligence into mathematical research.
It will bring together eight leading institutions and aims to create a network of more than 1,000 scientists.
The alliance will focus on developing specialised AI models for mathematics, using the technology to generate conjectures and formalise proofs, and building systems whose reasoning could be more easily interpreted.
scmp.com/news/china/science/…
Actually this is a valid scientific point and I would applaud this. It’s not like priority matters in this situation. Why not release everything, even if it’s bad?
hi @openai release all your *failed* attempts so human mathematicians can build on them. or are you afraid of getting scooped
This really does start looking like they just train on all your data no matter what you say. The machines are smart no doubt but they are just stealing also for sure! Oh well.
Check out this loser!
>uploads a draft to Chat on Sept. 8th
>gets scooped by Chat on Oct. 6th
"The result released by OpenAI (#311) on October 6, 2026... identical to a preliminary version of the present paper that was uploaded to ChatGPT on September 8, 2026"
LMAO get rekt
So why the heck did the machine fail to resolve resolution of singularities in char p given that the answer is either a very complicated algorithm or a counterexample? Isn’t that supposed to be what they are very good at????
Bets: 1) the square of a circle in a torus bounding a half area embedded disk is displaceable 2) closed exact lagrangians have primitive homology class (b/ god is great) but Yasha’s insane regular lagrangian conj is false. Because the moral is that FLOER THEORY KNOWS THE TRUTH!
I really didn’t expect this but (speaking as a symplectic topologist) these OpenAI results put me in a great mood! We have always been good at cool conceptual ideas but been bad at fiddly examples. It seems that our cool conceptual ideas have been pretty close to the Truth ;)
Summary of OpenAI’s symplectic topology results (assuming correctness): FLOER THEORY INDOMITABLE, WE ARE GOOD AT OUR JOBS!
Amusingly a better counterexample to degenerate Arnold was posted to the arxiv earlier so oai got scooped. lol.
Anyway I conclude from this that we can probably figure out exactly how sharp floer theory is and it is *pretty sharp* and this is GREAT !!!
Oh my god, nearby Lagrangian is false, precisely by the approach that kragh had envisioned. This is awesome! All hail tom kragh.
The only person brave enough to disbelieve nearby Lagrangian. I’ll share a story about tom sometime. Much love
Honestly it’s cool that floer theory seems to give sharp results with regards to Hamiltonian dynamics!
Well I guess according to Tristan Buckmaster we know officially now that internal anthropic models and lots of compute were heavily used in the Navier Stokes collaboration. Which is obvious, but for whatever reason was studiously avoided being confirmed in Twitter? @__alpoge__
AI is like Excalibur — if you want it to help you, rather than to simply make you feel very weak, then you must be very strong ;)
People on this app don't appreciate how well purely human mathematics is doing. People prove major results literally every few weeks, with proofs that are insightful, beautiful, and solve many problems simultaneously. Currently reading this wonderful thing: arxiv.org/pdf/2608.26392
At the moment we seem very obviously in the centaur era of math — and it is cool. (And imo Navier Stokes really counts since there is a lot of human infrastructure in the setup.) I like these short arguments that someone should have found long ago!
SITUATION DETECTED: Holomorphic Dynamics. The wandering domain results got destroyed by the machine.
A paper of Ye proves a much stronger result just using hyperbolic metric considerations. The machine found the argument.
arxiv.org/abs/2609.23834
Semon Rezchikov retweeted
SITUATION DETECTED: Holomorphic Dynamics. The wandering domain results got destroyed by the machine.
A paper of Ye proves a much stronger result just using hyperbolic metric considerations. The machine found the argument.
arxiv.org/abs/2609.23834
Semon Rezchikov retweeted
Hide the proofs inside plastic eggs and have mathematicians hunt for them in a big outdoor event.
This advisory group created by IAS is advising OpenAI in how to release their math results
Mathematicians can give their opinion in this link
agmai.org
I share publicly my opinion below (more than happy to have debate / disagreements in the responses!)
