How AI is breaking the rules of being smart, and open AI claims resolved 372 long standing mathematical problems 

Imagine you’re the founder of a frontier AI lab like OpenAI or Anthropic, and you’ve built a powerful AI model, but now your AI model has to solve the world’s toughest math problems.

Even if you train your AI model on math data from all over the internet, that is not a matter of high-level reasoning in your model. Meaning that hard brand-new math problem is not getting solved by your model. So what will you do? Where will you get such high-quality data? Think about it once.

OpenAI launched a program on 29th July 2026. The name of the program was ChatGPT for Academic Researchers, in which they said that by 2027, 1 lakh academic scientists and mathematicians can use our AI models for free. Understood?

In recent days Something extraordinary is happening in mathematics, and it’s forcing us to confront a deeper question: in an age when AI can solve problems that once defined entire careers, is deep learning still worth it?

In recent months, OpenAI has made a series of startling claims. On September 8, 2026, the company said nearly 10,000 of its AI agents had solved the Navier-Stokes equations — one of the seven Millennium Prize Problems — in roughly 88 hours. Weeks later, on October 6, it announced that an unreleased model had resolved 372 long-standing open mathematical problems. Among them were 16 problems that Fields Medalist Hugo Duminil-Copin had spent much of his career pursuing. The French mathematician later described crying so loudly that night that he woke his daughter.

These are not ordinary claims. The Millennium Prize Problems carry a $1 million reward each. Only one — the Poincaré Conjecture, solved by Grigori Perelman — has been fully accepted by the mathematical community, and Perelman famously refused the money. OpenAI’s announcements, however, remain claims rather than settled proofs. Independent researchers have already pointed to mismatches between the papers and formal verification in the Lean proof assistant. Still, the speed and scale of what is being attempted have shaken the field.

Terence Tao, widely regarded as one of the greatest living mathematicians, and 25 other Fields Medalists signed a declaration titled “Severe Misalignment of AI in Mathematics.” Their concern is not that AI should be banned from mathematics. It is that solving hard problems has become a competitive benchmark for AI companies, and that the sheer velocity of machine-generated results is ending what Tao calls the “Romantic Era” of mathematics — the age of solitary heroes who spent years on a single problem and transformed the field.

Steven Strogatz has expressed a similar emotional response. The fear is less about job loss and more about the sudden compression of a 40-year human tradition of slow, careful discovery into something machines can attempt at industrial scale.

Yet the discomfort of mathematicians, while understandable, may be missing the larger point. History offers a useful parallel. In 1986, mathematics teachers protested the introduction of calculators in classrooms, arguing that students would lose essential skills. Within a few years, calculators became standard tools. No serious person today claims that calculators destroyed mathematical understanding; they changed what “understanding” looked like.

AI is a far more powerful discontinuity. It is less like adding a calculator and more like moving an entire society from scooters to rockets overnight. The companies that once appeared exceptional can now be reverse-engineered in significant ways. The same logic that is unfolding in mathematics will soon appear in physics, chemistry, and every other domain. Change of this magnitude is not optional.

The real question, then, is not whether AI will keep advancing. It is what human knowledge is for in this new environment.

Deep learning has not become obsolete. It has become a multiplier. An intern with solid fundamentals and genuine curiosity, paired with AI, can now explore questions that previously required years of specialized training. An experienced expert who freezes in shock at the machine’s speed risks underusing the very tool that could amplify their hard-won insight tenfold or a hundredfold. The same AI that answers a beginner’s shallow question will answer an expert’s precise, well-posed one far more powerfully. The difference lies in the quality of the human asking.

Processor design offers a clear analogy. No one designs modern chips by hand anymore. Layout and physics calculations are handled by machines. Yet far more engineers work in the field today than ever before. Those who embraced the tools expanded the frontier; those who resisted were left behind.

The purpose of studying a subject deeply has shifted. It is no longer primarily about performing tasks that machines can now do faster. It is about understanding how the world works so that one can ask better questions, interpret machine-generated results, and build on top of them. Curiosity plus genuine knowledge, amplified by AI, is becoming the new competitive advantage. Breakthroughs may increasingly come not only from established veterans but from curious people who once would have been locked out of the conversation entirely.

This does not erase the real costs — job disruption, environmental impact, or the legitimate grief of those who dedicated decades to problems that machines are now racing through. Those costs deserve serious attention. But refusing to engage with the technology will not slow it down. Whether we love or hate AI, its pace will not be set by our feelings.

The most constructive response is neither denial nor blind enthusiasm. It is to treat deep learning as more essential, not less. Choose a domain that genuinely interests you. Master its fundamentals. Then use AI as a force multiplier rather than a substitute. The people who do this will not only adapt; they will help decide what comes next.

The romantic age of solitary mathematical heroes may be ending. A different, more distributed age of discovery is beginning. The choice is whether we participate in it with open eyes or watch from the sidelines.