New Fields Medalist Announces Joining OpenAI on Same Day

On July 23, 2026, at the opening ceremony of the International Congress of Mathematicians in Philadelphia, 38-year-old Jacob Tsimerman, along with three other mathematicians, received the Fields Medal. At the subsequent press conference, when asked about his future plans, he directly stated: 'I am about to join OpenAI to work on AI safety.' This announcement instantly turned the award ceremony into a transfer press conference.

Tsimerman's resume is impeccable: an IMO gold medal with a perfect score in 2004, entering the University of Toronto at age 16, completing his undergraduate degree in two years, and becoming the youngest full professor in the history of the university's mathematics department. He won the Fields Medal for proving the André-Oort conjecture and the Griffiths conjecture, becoming the first mathematician to win the Fields Medal while working at a Canadian university.

However, this mathematical genius's attitude toward AI is full of contradictions. In 2025, he co-authored a paper with Andrew Critch titled 'A Taxonomy of Human Extinction Scenarios Involving Artificial Intelligence,' systematically outlining five pathways by which AI could lead to human extinction. He has stated in interviews that there is a 'considerable probability' of AI causing human extinction and has supported a pause on frontier AI development. Yet now he has joined the world's most cutting-edge AI company.

In an interview with AMS Notices, Tsimerman explained his shift: he supports a pause on AI but does not believe companies and countries in global competition will actually stop. 'Since we can't stop it, rather than standing outside and resisting, it's better to get in, understand it, and help shape it, otherwise we might be left behind.' His logic is: the higher the risk, the more safety conclusions need the level of certainty of mathematical proofs, and OpenAI is precisely where mathematics and AI safety intersect.

He has already been using AI extensively for mathematical research. In 2025 alone, he uploaded five mathematics papers to arXiv. His daily workflow involves using ChatGPT and Claude to explore problem boundaries, find references, and have the models prove lemmas. He says: 'The proofs it gives are usually wrong, but it usually pulls out the right direction and methods, helping me get 80% of the way there, and I finish the rest myself.' AI has already doubled his research efficiency.

In May 2026, OpenAI's reasoning model found a counterexample to the 80-year-old unit distance conjecture. Tsimerman was one of nine mathematicians invited to verify it. He commented that if this result had been achieved by a human, he would support its acceptance by a top journal. This experience reinforced his judgment: AI has begun to participate in mathematical discovery. He subsequently publicly predicted that AI might become better than human mathematicians at proving theorems within two years.

Just three days before Tsimerman announced his move to OpenAI, his student Levent Alpöge posted on X a counterexample to the Jacobian conjecture, stating that the key construction came from Anthropic's Claude Fable 5. This conjecture, proposed in 1939, had remained unsolved for 87 years. The counterexample was only 216 characters long and could be manually verified by any mathematician. Fields Medalist Timothy Gowers commented that this was the first time he had seen an LLM solve a 'long-famous' problem outside its own field.

Alpöge graduated from Harvard University, and his undergraduate thesis advisor was Tsimerman. Now, the teacher is at OpenAI, and the student is at Anthropic—master and apprentice, each at a different company. In his only X post, Tsimerman predicted that AI would take over problem-solving, and his student used AI to fulfill that prophecy ahead of time.

Tsimerman believes that the profession of mathematician may cease to exist. He wrote in his post: 'The mathematical profession as we know it... I don't think it will continue to exist in its current form.' He has stopped recruiting PhD students, reasoning that he does not want to prepare students for a 'mathematical profession that may no longer exist.'

Tsimerman's move is not an isolated case. In December 2025, 57-year-old analytic number theory authority Ken Ono resigned from his tenured chaired professorship at the University of Virginia to join Axiom Math, an AI mathematics company founded by his student. Jeremy Avigad of Carnegie Mellon noted: 'The key to general intelligence is combining the insights of machine learning with the precision of mathematics.'

Behind this migration are hard reasons: traditional academic labs have annual computing budgets of only a few hundred thousand to a few million dollars, while a single large-scale experiment in frontier AGI research costs tens of millions of dollars. Universities can only conduct small-scale theoretical reasoning, while OpenAI and Anthropic possess tens of thousands of high-end GPUs and proprietary datasets. Additionally, tenured professors must handle teaching, administrative duties, etc., while industrial labs have extremely low administrative pressure and can directly assemble interdisciplinary teams.

However, the brain drain also brings concerns: the departure of tenured professors leads to a shortage of graduate advisors, and young people may develop the expectation that 'the best path is to go directly to an AI company,' thinning the basic talent pipeline. University research is published openly, while a large amount of core research in private labs is not disclosed, potentially allowing frontier theory to be monopolized by a few companies, leading to a closed academic ecosystem.

In his AMS Notices interview, Tsimerman described a thought experiment: suppose there is an AI oracle machine that can instantly determine the truth or falsehood of any mathematical proposition. The daily work of a theory builder would become: propose a definition, immediately ask a thousand questions, answers keep coming back, and an entire research direction could be initiated and concluded within days. He says: 'This is very, very different from modern mathematics.'

Despite his pessimistic view of AI, Tsimerman still believes mathematicians can play a unique role in AI safety. He says: 'This (AI) is still mainly an empirical and engineering science. If we can better understand how these things work, perhaps we can better guide them, understand them, and control them.' He is compiling a resource list specifically for mathematicians who want to transition into AI safety.

Tsimerman's choice has sparked broader discussion: the mathematical community is roughly divided into three camps—Terence Tao sees AI as a powerful assistant, Fields Medalist Akshay Venkatesh warns that mathematicians may lose direct understanding of mathematics, and Tsimerman believes the entire profession is at risk. Regardless of the viewpoint, an undeniable fact is that the boundary between mathematics and AI is dissolving, and mathematicians are standing at a crossroads of decision.