David Turturean, a recent MIT graduate who works on the technical staff at the startup lab Physical Superintelligence (PSI), gained early access to Anthropic's Fable 5.2 model and used it to crack an eighteen-year-old quantum computing problem. As he put it, "In a few months, like, why would you even put out a paper."

The problem is the Aaronson-Ambainis conjecture, which deals with whether quantum computers can outperform classical ones at solving certain problems. Scott Aaronson, one of the two researchers who originally posed the conjecture, writes that while he doesn't fully grasp the proof himself, he trusts it's correct—it was autoformalized in Lean, a programming language built for mathematics, and a team of leading researchers that PSI shared the solution with walked through it. The result shuts down one possible route quantum computers might take toward rapidly solving particular hard problems, though not every route. "It's a negative result that allows people to know, okay, we're no longer looking there," Turturean explained. On a broader level, it captures the disruption AI-driven problem-solving is bringing to the scientific world. "A proof of full [Aaronson-Ambainis conjecture] was one of the last remaining open problems in quantum query complexity that everyone seemed to agree was centrally important," wrote Aaronson, who said the news left him feeling satisfied, awestruck, disappointed, and fatigued all at once. "I was surprised, honestly, to see it missing from OpenAI's Mathocalypse dump a few weeks ago. Surely they tried it, so was it really so much harder than all the problems that were solved."

That stands as a point in favor of PSI's approach, which relies on specialized model harnesses and scaffolding to tackle scientific problems. "You want to be able to really go after a problem at breadth and depth, and be able to consider a wide range of potential alternatives."

At present, the company generates revenue by using its tools to optimize data centers—simulating variables like heat, power, water, and sound—"where the performance of that site as well as its impact on the community are a direct function of its physics." The bigger ambition, though, is turning theoretical physics breakthroughs into real technologies. "Six to nine months ago, when I was talking to investors, like I had to sort of stretch to validate that conditional, right. Like, 'no, no, no, trust me, pretty soon, theoretical physics is going to be mostly the domain of machines,'" Pines recalled. "And it's not there yet, but it's on its way."

Even so, while AI has delivered major solutions in fields like mathematics and computer science—where problems can be expressed in validated code—doing the same in less abstract disciplines such as the life sciences or physics has proven harder. Recently, the startup Discovered Materials launched a $1 million grand challenge inviting researchers to use AI to understand problems like superconductivity and heat dissipation. "At the end of the day, physics is different than math because it's the real world … nature gets a vote in your models of reality," Pines said. He noted that the startup is concentrating on figuring out how to take computational results and validate them experimentally as efficiently as possible. "If it's a true fact of nature, that's a trick that you can exploit," Pines said. "You could build a device out of it, right. The AI suggests an idea; it allows you to do something in the world you weren't able to do before…. and so each of those are kind of loops that PSI is building."

As "neolabs for science" emerge as a possible route to groundbreaking research, those involved are still working out what the standards should be in this new landscape. Because PSI had to hold its result until Anthropic's model was close to release, the team had time to share it with relevant researchers and better understand it. "As an academic, we wanted to play our role in doing things the right way," said Jim Halverson, a Northeastern University professor who advises PSI. He'll be at a summit at MIT next week where leading physicists will attempt to work out a response to the accelerating pace of science. "Best practices are still emerging … We figured that transparency is a good way to go, and contacting experts is a good way to go."

"We had some time to sit on it … and it was the first time that I actually took one of my AI-elicited solutions on print, and I read through it, and I was like, okay, how can I make this … be presented more naturally for a human. And I think it also made me realize, oh my God, the weight of what AI can do."