Bartosz Naskręcki (@nasqret) / X

Bartosz Naskręcki

2,744 posts

Bartosz Naskręcki

@nasqret

Mathematician | Vice-Dean @UAM_Poznan | Researcher @ccaiwut | Owner of MathBot.eu | Mathematics, AI and programming

Poland

bnaskrecki.faculty.wmi.amu.edu.pl

Joined November 2023

Pinned
Bartosz Naskręcki
@nasqret
Oct 6, 2025
Those who asked for the link. Here is a read-only free access to our Nature paper with Ken Ono nature.com/articles/s4156…
Bartosz Naskręcki
@nasqret
Oct 6, 2025
I encourage you to read this article, in which we describe the current situation and the directions in which, in our view, mathematics is heading. Many thanks to Ken Ono for including me in this extraordinary project. I look forward to a wide-ranging discussion and will be
Bartosz Naskręcki
@nasqret
Oct 5, 2025
GPT-5-Pro solved, in just 15 minutes (without any internet search), the presentation problem known as “Yu Tsumura’s 554th Problem.” arxiv.org/pdf/2508.03685 This is the first model to solve this task completely. I expect more such results soon — the model demonstrates a strong
Bartosz Naskręcki
@nasqret
Oct 23, 2025
We often talk about big leaps in AI for mathematics, but I think the small steps are equally impressive. The future of mathematics is now. I was working on a particular task: finding a case-free proof of the representability of the local Néron function correction using a
Bartosz Naskręcki
@nasqret
Oct 19, 2025
I post regularly about the use of LLMs in my research. I don't care whether LLMs or other technologies will bring us to AGI — but their daily use is already a fact. This time, I used one to heavily refactor my own proof from a paper. I wanted to better conceptualize the very
Bartosz Naskręcki
@nasqret
Oct 14, 2025
I asked GPT-5-Pro whether it knows anything about elliptic curves y^2 = x^3 + a(t)x + b(t) that have positive rank and a torsion subgroup Z/2 + Z/8. This is a notoriously hard problem. The model failed-but there was no bragging, just the bitter truth. I consider that progress.
Bartosz Naskręcki
@nasqret
Aug 7, 2025
Replying to @nasqret
Ok, some general comments while I am waiting for the other tasks. GPT 5 and GPT 5 Thinking are nothing very novel for research mathematicians compared to o3-pro and o4-mini-high. But GPT 5 Pro is different. It uses much more compute and the quality of the answers is superb.
Bartosz Naskręcki
@nasqret
Aug 7, 2025
GPT5 Pro is out. I am testing on my favourite question about Galois groups of polynomials. I am impressed. First model to deliver a complete, very high quality reasoning for this difficult algebraic task.
Bartosz Naskręcki
@nasqret
Nov 13, 2025
Replying to @andrzejdragan
Tego nie da się załatać...
Bartosz Naskręcki
@nasqret
Oct 13, 2025
I read arXiv every day - it's my daily newspaper. But reading mathematically dense text is hard over your cereal. So I prompt GPT-5-Pro to help me disentangle the first layer of mathematical structure. With such scaffolding, I know when I want to dig deeper. Happy crunching!
Bartosz Naskręcki
@nasqret
Sep 20, 2025
Im więcej pracuję systematycznie z generatywnym AI tym bardziej jestem przekonany, że taka droga interaktywnej samo-edukacji połączonej z krytyczną analizą to najważniejszy składnik zmiany edukacyjnej na uniwersytetach.
Bartosz Naskręcki
@nasqret
Aug 13, 2025
Replying to @andrzejdragan
Warto podkreślić, że zadanie Paaty było niezwykle pomysłowe i kombinatoryczne bardzo skomplikowane. Jego sensem nie jest wziąć definicję całki Lebesgue'a i zrobić kilkanaście rutynowych kroków. Tam w rozwiązaniu było kilka bardzo pomysłowych twistów.
Bartosz Naskręcki
@nasqret
Sep 5, 2025
Replying to @karpathy
GPT-5 Pro is currently the best model for sophisticated mathematics. I find it pretty intimidating that, beyond fluid intelligence, the model feels as though it has read all the papers and books I never have-and never will.
Bartosz Naskręcki
@nasqret
Oct 5, 2025
This is a very fair point. The solution can be found on the obscure blog post here web.archive.org/web/2017090711… But somehow none of the other models made it. Also the solution of GPT Pro is slightly different. I position it as: here was a problem, I had no clue how to search for it
Bartosz Naskręcki
@nasqret
Sep 28, 2025
Replying to @nasqret
My recent attempt at GPT generated mathematics (abstract algebra). This is a kind of cherry picking but I felt very surprised by the quality of the output and the fact this proof is better than my own (but begs to be auto-formalized now!) x.com/nasqret/status…