OpenAI has announced that its latest AI model cracked a Millennium Prize Problem, sending a jolt through the mathematical community. The challenge carries a $1 million reward, but the bigger shock may be the speed, scale, and cost of the attempt: OpenAI used 10,000 AI agents, with an estimated effort of $15 million.
What happens when a problem that can occupy mathematicians for weeks, months, or years falls to AI in days? That question is now moving from speculation into the center of mathematical debate, as researchers confront a capability shift they did not expect to arrive so soon.
A result that changed expectations overnight
Prof Colva Roney-Dougal of the University of St Andrews said she once believed AI was unlikely to do anything remarkable soon. Three months later, the OpenAI announcement changed her view.
“I feel slightly shell-shocked,” Roney-Dougal said. “We’re all just waiting to see what happens. It’s coming so fast.”
Her reaction captures the pace of the upheaval. Mathematicians are shocked by how quickly AI capabilities have advanced, and the latest result has forced them to reconsider assumptions about which problems remain beyond the reach of machines.
Prof David Silvester of the University of Manchester described the field as “very unstable” given the pace of change. He said, “All the open maths problems could fall with enough resources.” For Silvester, the issue is not only what AI has solved, but what the $15 million effort suggests about the resources that can be directed at future challenges.
“This is irreversible. This is not going to change,” Silvester said.
Proof, review, and the cost of showing what AI can do
Mathematics depends on verification. Researchers check each other’s proofs, a process that played a role when mathematicians verified Fermat’s Last Theorem in 1993. An AI-generated result still has to face that standard, and Silvester suspects mathematicians may soon spend more time reviewing proofs produced by AI.
That could reshape the daily work of the field. Instead of spending years searching for a solution, mathematicians might review an argument generated in days, testing every step before accepting it. The work would not disappear, but its center of gravity could shift from discovering proofs to examining them.
Silvester also warned that the subject itself could feel different without the human process that has traditionally defined mathematical progress. “Without that, the subject seems completely different to me,” Prof Silvester said.
The scale of OpenAI’s effort has drawn a separate criticism. Prof James Robinson of the University of Warwick said big tech companies are burning fuel on problems that drive mathematics, using huge resources to demonstrate what their latest models can achieve.
“It’s frustrating to see these big tech companies burning this fuel up just so that they can show off about how great their latest model is,” Robinson said.
He called the behavior “immature playground boasting,” adding: “It seems like immature playground boasting writ large, underpinned by billions of dollars and the potential for significant environmental damage in an age when climate change is probably the biggest challenge we face.”
Universities face a new mathematics dilemma
The breakthrough is already raising concerns about work and education. Universities are having to tell students not to use AI for some problems, a sign that traditional assignments now face a direct challenge from systems able to attack advanced mathematical questions.
That creates a difficult balance. Mathematics trains people to work through hard problems, build arguments, and verify conclusions, yet AI may now solve some problems before students have had time to develop those skills. Universities must decide where AI belongs in learning and where its use removes the very challenge an exercise was designed to provide.
Mathematicians also face a broader question about the value of their expertise. If AI can solve problems that once demanded years of work, researchers may need to focus on judging proofs, selecting meaningful questions, and understanding what a solution reveals.
Prof Alexander Paseau of the University of Oxford believes mathematics will not lose much of its importance despite the upheaval. That view leaves room for a transformed subject rather than a diminished one: AI may change how mathematicians solve problems, while the importance of deciding which problems matter remains.
The $1 million Millennium Prize Problem has therefore become more than a mathematical target. It is a test of how far AI can reach, how much energy and money companies will spend to prove that reach, and how universities will prepare students for a discipline changing beneath them.
The next phase will depend on verification. Mathematicians must examine the AI-generated proof, and the field must decide how to absorb a result created through 10,000 AI agents and an estimated $15 million effort. One thing is already clear: the pace has changed, and mathematics is now waiting to see what falls next.
Based on
