Throughout the summer of 2026, there were plenty of signs of what was to come. Every week, it seemed, artificial intelligence models were offering proofs of decades-old mathematical conjectures. So what happened on September 8 shouldn’t have been a complete surprise. And yet to many it felt sudden, almost violent. OpenAI announced that its models had solved one of math’s most famous problems — and immediately sparked a controversy over authorship and standards. Mathematics would never be the same.
“In whatever years I have left, I don’t expect that I’ll ever again prove a theorem because I’m actually needed to prove it,” Scott Aaronson of the University of Texas, Austin wrote on his blog. “Human mathematicians are forevermore dethroned as the main theorem-proving entities on planet earth.” Other online missives expressed optimism, but many revealed some combination of grief, confusion, and fear.
Two days after the announcement, I found myself in a classroom at the University of California, Berkeley, surrounded by some 150 students, postdocs, and professors. The mathematician Ken Ono, who left academia for an AI start-up called Axiom Math, was scheduled to give a talk. “You might be graduating into a profession that might not even exist, or that will be very different than what you expected,” Ono told them. “You need to brace.”
The audience responded with anger and frustration. There were whispers and exchanged glances; Ono couldn’t make it through a single slide without a fresh wave of questions. “I’m not entirely sure what our takeaway is supposed to be,” one student said. Another asked how Ono and his start-up would take responsibility in light of “the shameful way that AI companies are treating mathematics.”
When Ono said that mathematicians would do their “very best” to avoid the bleak future the students were envisioning, a third retorted, hands trembling: “What are you doing? What is your ‘very best’?” The talk, including time for questions, had been allotted 50 minutes; with the extended Q&A, it went on for more than two hours. Many students lingered afterward, to vent and console one another.
To me, the reasons for these emotional reactions were obvious. The young mathematicians recognized that their field would have to change or risk extinction. What did surprise me was that when I returned to New York and brought it up in conversations with friends or family, they didn’t see what the problem was. All sorts of amazing discoveries are around the corner, they imagined. Wasn’t that the point?
These conversations exposed a long-standing problem: Many people don’t know what mathematics really is, or why mathematicians do it. Now mathematicians need to answer that question for themselves and explain it to the rest of the world — quickly. How they answer will determine what happens next.
As an undergraduate, I studied both mathematics and literature. I always found that “pure math” — the study of mathematical concepts for their own sake, without a care for real-world applications — existed somewhere between the sciences and the arts. It prizes logic and certainty, but at its core lie fuzzier notions of beauty, intuition, and depth. “Math is either the most science-y humanities or the most humanities-type science, depending who you ask,” said Marcel Goh, a doctoral student at McGill University. Usually it’s utterly useless, yet time and again its abstractions demonstrate “unreasonable effectiveness” in applications to physics and beyond.
In mathematics, it’s not a cliché that the journey matters more than the destination. Problems are posed not so much because their answers will be practical and important, but because they represent interesting journeys. The hope is that as mathematicians struggle to solve a problem, they’ll come up with intriguing tools and connections, stumble on novel ideas and directions, take fruitful detours, and answer new questions they never would have thought to ask.
“It was never about the theorems,” Goh said. “It’s to carry on a tradition that has benefited society despite not having any market value. To reason deeply, to understand the world … and to spread that knowledge.”
AI models skip to the end. At the moment, their proofs are nearly impossible to read. They elide salient details and waste pages on irrelevant concepts. They fail to demonstrate how the new work builds on or connects to other results in the mathematical literature. Mathematicians know that the proofs are true — they’re verified by a system that checks their logic — but have no idea why. What’s missing is understanding, the currency that mathematicians trade in. If there is a journey, humans are not invited along for the ride.
It’s as if you were teleported to the peak of a tall mountain. Surrounded by fog, you have no idea where you are, or what’s around you. You do not know how your mountain connects to others, and you have no equipment to help you explore, no way to help someone else join you. If you had climbed the mountain yourself, you would have experienced how the human body adapts to altitude and changes in oxygen levels. You might have had to invent tools to navigate, to climb steep cliffs, or to make a shelter. You might have encountered a fellow explorer, gotten lost together in a hidden valley, and found a plant that could be turned into a life-saving medicine.
Instead you’re perched on the peak but in the dark, while the maker of the teleportation machine tells you that it can explore the wilderness better than any human.
Mathematicians have always known that understanding is more valuable than an answer. So why should an AI answer machine feel like such a threat?
In part, it’s because of the way mathematics has evolved. For centuries, proving theorems and understanding those proofs have gone hand in hand. If someone solves a problem that the mathematical community has decided is central and important and hard, that solution is considered a reflection of deep insight and creativity.
So mathematical culture has developed almost entirely around problem-solving. Graduate students cut their teeth on simpler versions of big, complex problems, then seek to get their work published, find good jobs, and earn credit and recognition — all of which are assessed in terms of the theorems they prove. The incentive structures that power the mathematical community are not set up for a world where solving problems is easy and understanding their answers remains hard. “It breaks our system,” said Tasmin Chu, a doctoral student at the California Institute of Technology. “It just takes it to its absolute limit and destroys it.”
Over the past few months, I’ve heard from lots of mathematicians, students and senior researchers alike, about the shape of this destruction.
Companies such as OpenAI and Anthropic are competing directly against mathematicians (and poaching many of them) to quickly prove theorems, while putting little effort into writing clear papers to explain those results or put them in context. Most mathematicians have no idea what these companies have even proved at this point; OpenAI has announced that it’s keeping more than 100 important results secret. (This has the further effect of discouraging mathematicians from working on questions that might already be moot.)
“What is happening is a huge damage to open science,” Chu said. The only possible value of this is shareholder value. It completely overlooks the central mathematical enterprise — and, even worse, it seeks to overwrite it.
Many mathematicians have stopped posting open conjectures and potential ideas at the end of their papers, for fear that they’ll be scraped by bots and fed into AI models. Others are posting their papers before they’re ready in order to avoid getting scooped. Journals and the online preprint server arxiv.org have been overrun with AI-assisted or AI-generated proofs that no one can properly digest or check, which forces researchers to do the hard, currently thankless work of rewriting the proofs so they can actually benefit the mathematical community. (Even as models improve and those papers get easier to read, their sheer volume will overwhelm current systems for sorting through new work and deciding what’s interesting and important. In the meantime, arxiv.org has started limiting authors to two submissions per month.) When an editor of Transactions on Machine Learning Research reached out to the authors of 10 papers slated for rejection, most either did not respond or could not answer basic questions about work they’d claimed as their own.
At the same time, MathOverflow, an online forum where mathematicians answer each other’s questions — connecting famous researchers with young students and leaving years of exchanges for future generations to read and learn from — is seeing a massive decline in activity, as students privately ask AI models their questions. Others don’t want to share ideas in another space exposed to bots.
The situation is heartbreaking. Part of what is so wonderful about mathematics is its open and honest collaboration, and its independence from political or corporate concerns. That’s not to say that there haven’t been other problems with the culture of mathematics — inequity and lack of access, among other things — but what’s special about that culture is at great risk without a signal change in the field’s incentive structure.
“The existing equilibrium has broken,” said Daniel Litt, a mathematician at the University of Toronto. “We’ll have to find a new one.”
There’s a range of possible futures based on how mathematicians collectively respond — or don’t respond — to this moment.
In one, “once the profession turns over, you have a collection of mathematicians who are trained in pushing a button without reading the outputs,” Litt said, rather than being trained to ask interesting questions and engage with the results. “There’s a bad future where if we don’t adapt, there’s just no more math in 50 years. No one is doing it, not even the machines.
“I’m not saying it’s likely,” he continued. “We’re not going to sleepwalk into that. But I think the world where we change nothing probably looks like that.”
As Matilde Marcolli of Caltech wrote in an essay for the London Mathematical Society Newsletter, “If we insist on focusing our entire profession solely on the tackling of conjectures as the unique goal and reward, then we may be out of business quite fast, not because of the end of mathematics but because of the end of our imagination.”
There is a more optimistic alternative — assuming a new incentive structure that disengages problem-solving and theorem-proving from jobs, advancement, and recognition. “Mathematicians will be spending less time trying to peck on the permafrost, getting out mathematical gems with great effort,” said Pavel Etingof of the Massachusetts Institute of Technology. “They’ll be sourcing that kind of pecking more to AI and trying to interpret and understand and learn the results that will come from that. There will be a lot more mathematics to make sense of. And there will be a lot of work for mathematicians — maybe different work from before, but definitely interesting work.”
That work will involve, to some extent, studying proofs written by AI — the way every undergraduate learns the known proofs in their textbooks. Many mathematicians expect AI-generated arguments to graduate from slop to more elegant writing soon enough; perhaps within a few years, Litt said, “we’re reading beautifully written expositions of highly interesting results. To me, that’s the dream. That’s why I got into it.”
These future mathematicians will have much to do. After all, Litt added, “there’s no amount of well-writtenness that lets you understand math.” They’ll have to rewrite AI proofs, find more elegant arguments, come up with theories and frameworks and definitions, ask questions, formulate new conjectures, develop a sense of what’s interesting and important. They’ll be the ones building research programs, creating communities around certain ideas, and passing that knowledge on to others. (I am reminded of the literary theorist Roland Barthes’ famous 1967 essay “The Death of the Author,” in which he argues that it is the reader, and not the writer, who has ultimate control and agency in actively creating a work’s meaning.)
“If AI proves all possible theorems, and no mathematicians know about these theorems, it cannot be part of culture, because culture is inherently human,” Etingof said. “Those papers will only become a part of mathematical culture when some human reads them and understands them and gets something out of them.”
The question is whether the next generation of mathematicians will want to. Or even be able to.
Mathematicians are already rethinking how they’ll train their students, accept papers for publication (they’re currently experimenting with requiring accompanying talks), and make hiring decisions. There aren’t easy answers for more nebulous skills such as understanding, communication, and vision. But the stakes are high enough that I think they’ll succeed.
I’m less certain about whether new generations of mathematicians will be attracted to this remade discipline.
Many people are drawn to math because they want to solve problems, explore frontiers, and be the first to discover some new truth. Will they be willing to put in decades of training for the job of interpreter? They’ll still read and translate proofs and expect to rewrite them. But it seems less likely that they will frequently experience the same moments of clarity or sense of creative ownership.
“A lot of the reason why people do pure math, even though it pays so much less than industry, is because you get to author your own creative contributions and then give them to a community that appreciates them for their intellectual beauty,” Chu said. “This environment where mathematical discourse is centered around machines talking and humans listening, I don’t think it’s a very attractive one.”
And AI is itself undermining the kind of education that would be necessary for that role. This came up at the Berkeley colloquium, too. “There’s a serious risk that 20, 30, 40 years from now, there will be no people around who have the same classical mathematics training,” one student said. “I’m worried about a world where the full body of mathematical knowledge is inaccessible.”
The result could be a lost generation. Goh is one of many young mathematicians struggling with the angst such a future is producing, and with what he sees as a lack of effort on the part of working mathematicians (at least until very recently) to remake the incentives to avoid it.
“There’s something sick about the culture,” he told me. “I didn’t think so, but now I do.”
Goh is alarmed by how some have taken up AI without enough critical thought, and he feels pressured to use it so that he doesn’t get left behind. “I was just very depressed about that,” he said. “That doesn’t feel inspiring to me. … It didn’t have to be this way.”
He likens it to Saruman from The Lord of the Rings, corrupted by the promise of unlimited knowledge and power. Goh and a growing number of mathematicians are starting to consider themselves AI “vegetarians,” consciously measured in their use of the technology. There’s nothing rewarding or interesting or fun about offloading their cognition to a machine, they argue.
Goh is determined and would rather leave the field than use AI in his creative work. “I think the ship is sinking,” he admitted, “but I’ve already made my decision to stay on board as long as I can.”
There is, through all of this, reason for hope. Mathematicians see these problems and feel a collective sense of concern (if not panic), so they are working now to create attractive conditions for people in their field — both those interested in using AI and those who don’t want to. Chu and others have formed the Association for Human Mathematics, for example, because “what we need right now actually is not more mathematics but policy proposals and tools … to deal with this crisis and change our ecosystem.”
It will be hard. Inertia is powerful. But mathematics could come out of this stronger than before. Forced to name and prioritize what people love and value about the field, current and future mathematicians can emphasize skills that have previously been deprioritized: the ability to communicate and share knowledge, to develop a broad vision, to experiment with original ideas without feeling the need to publish theorem after theorem. To think deeply.
“I think there is a good future where we come out way ahead of where we are now, where we learn a huge amount of interesting math,” Litt said. “We have people doing the coolest stuff they’ve ever done. We have a really vibrant community of people who deeply understand stuff.”
If we can’t adapt, he added, “whose fault is that?”
Editor’s note: Scott Aaronson is a member of Quanta Magazine’s advisory board.
