Fields Medalist Deng Yu: If AI Solves All Math, I'll Write a Yuri Novel
In a social media post, 37-year-old Yu Deng, a mathematics professor at the University of Chicago and 2026 Fields Medal winner, stated that if artificial intelligence could solve all mathematical problems, he would go home to write his own yuri novel. He later clarified to English media that he is not currently writing a yuri novel, and that statement was entirely a joke. He does not believe AI will develop to the point of solving all the mathematical problems mentioned in his post. He also noted that AI has caused anxiety and impatience in the mathematics community, and that the situation will gradually become clearer after the current wave of advancements over the past couple of years, with mathematics slowly adapting and a new ecosystem forming.
Yu Deng was awarded the Fields Medal on July 23 at the International Congress of Mathematicians in Philadelphia for his rigorous derivation of the Boltzmann equation from hard ball dynamics, derivation of wave dynamics equations from nonlinear dispersive systems, and development of probabilistic methods in nonlinear dynamics. The citation described this work as one of the century's major advancements in deriving the fundamental laws of physics from first principles. He completed his undergraduate studies at Peking University and MIT, and earned his PhD at Princeton. A nearly 200-page paper co-authored with Zaher Hani and Xiao Ma was published in March 2025.
He has publicly expressed his interest in Japanese manga, particularly yuri themes, since his time at MIT, and related materials can be seen in his office in the award introduction video; he was reading a romance novel when he learned of his award. Another Fields Medal winner, Jacob Tsimerman, has chosen to leave mathematics to work on safety research at OpenAI, citing the observation that open problems are being solved by models. The scientific community is also discussing whether AI has touched upon million-dollar open problems.
The statements themselves do not change proofs or prizes, but they have moved the question of whether models will outsource theorem production from seminars to the public arena. The buyers are model companies vying for mathematical talent and computational narratives, while the sellers are mathematicians who still consider papers and problem lists as their professional assets. Funding and attention are flowing towards proof assistants, formal verification, and safety positions. Beneficiaries are laboratories that can treat open problems as benchmarks; those under pressure are young researchers who have staked their entire professional identity on "problems not yet exhausted by machines."
Source: Public Information
ABAB AI Insight
Yu Deng's research is already providing a strict bridge for the question of "how micro reversibility leads to macro irreversibility". Now, with a joke, he has brought the same question to a professional level: if machines clear the problem list, what irreversible choices remain for humans? The rules of the Fields Medal are for those under 40, awarded every four years, recognizing those who can still define the next round of questions, not those who have already completed theorems. By framing his fallback as writing a yuri novel, he acknowledges that aesthetic judgment cannot yet be replaced by the same set of automated solvers.
The capital path is more straightforward in the case of another winner. Jacob Tsimerman went to OpenAI after winning, following the path of "entering the model-making institution before problems are consumed by models." Mathematical talent is flowing from university lectures to safety and alignment teams, with salaries and computational power funded by laboratories, and the priority of papers yielding to internal evaluations. Yu Deng has chosen to remain at the University of Chicago to continue working on equations while publicly joking that he does not believe "all problems can be solved." Both paths are competing for the same group of minds under 40.
Comparing to Maryna Viazovska's sphere packing in 2022 and Maryam Mirzakhani's aspiration to be a writer in 2014: it is not new for mathematicians to consider literature as a backup plan; what is new is that this backup is being referred to as the "condition for machine victory." The industry phase is characterized by tools penetrating proof assistants, but not yet taking over the definition of problems. A narrow version of Hilbert's sixth problem has just been written by a human team in nearly 200 pages, indicating that the hardest derivations still rely on long-term combinatorial structures, rather than being generated in a single dialogue.
Structurally, this is a boundary test for technological substitution. If substitution only occurs in formalizable derivation steps, mathematicians will become problem selectors and validators; if even the problem list is exhausted by models, pricing power will shift from journal reviews to benchmark evaluations. The mechanism is: once open problems can be treated as competition scores, capital will continue to pile models; the parts that cannot be treated as scores—whether the problems are good or worth pursuing—remain on the human side. By naming the latter part as a novel, the joke indicates that judgment has not yet been incorporated into the loss function.
ABAB News · Cognitive Laws
- Machines first seize verifiable problems, leaving humans with unverifiable judgments.
- The fallback framed as a novel indicates that aesthetics have not yet been written into evaluations.
- Awards for those under 40 are purchasing the next round of questioning rights, not a warehouse of solutions.