PSI Uses Anthropic's Fable 5.2 to Crack 18-Year-Old Quantum Problem

Physical Superintelligence, a startup lab known as PSI, used early access to Anthropic's Fable 5.2 model to solve an eighteen-year-old open problem in quantum computing — the Aaronson-Ambainis conjecture. David Turturean, a recent MIT graduate and member of PSI's technical staff, ran the solve, and the proof landed with the field's own co-poser of the problem vouching for it.

Scott Aaronson, the University of Texas researcher who posed the conjecture with Andris Ambainis, wrote that he doesn't quite understand the proof but believes it's correct. His reason: it was autoformalized in Lean, a coding language for math, and explained by a team of top researchers with whom PSI shared the solution.

"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," Aaronson wrote. He described himself as feeling satisfied, awestruck, disappointed and fatigued by the news.

What does the result actually change?

The conjecture concerns whether quantum computers can solve certain problems faster than classical ones. The answer PSI produced is a negative result: it closes off one avenue quantum computers might have used to crack difficult problems at high speed, though not all of them.

"It's a negative result that allows people to know, okay, we're no longer looking there," Turturean told TechCrunch. For quantum complexity researchers, that's a real redirection of effort — one search direction is now formally dead.

Aaronson noted he was surprised the problem was 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?" he wrote.

That gap is exactly the argument for PSI's approach. The company's model isn't raw frontier capability — it's specialized harnesses and scaffolding around models.

"An effective kind of research system should have lots of things related to harnesses, lots related to tools, memory systems, simulation infrastructure," PSI CEO Matt Pines told TechCrunch. "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."

How does PSI make money today?

For now, PSI's revenue comes from applying its tools to data center optimization — simulating heat, power, water and sound, what Pines calls environments "where the performance of that site as well as its impact on the community are a direct function of its physics."

The long play is converting theoretical physics breakthroughs into actual technologies. Pines said investor conversations six to nine months ago required some stretching: "'no, no, no, trust me, pretty soon, theoretical physics is going to be mostly the domain of machines.' And it's not there yet, but it's on its way."

The physics part is the hard part. AI has claimed high-profile solutions in math and computer science, where results can be expressed in validated code. Less abstracted disciplines like the life sciences and physics have proven tougher. Last week, the startup Discovered Materials launched a $1 million grand challenge for researchers using AI to understand problems like superconductivity and heat dissipation — a signal that money is now chasing exactly this bottleneck.

"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. PSI is focused on taking computational results and validating them with experiments 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."

Who verifies AI-generated science?

As "neolabs for science" become a pathway for potentially groundbreaking research, the participants are still working out the standards. Because PSI had to hold its result until Anthropic's model was nearly ready for release, the team had time to share it with relevant researchers before publication.

"As an academic, we wanted to play our role in doing things the right way," said Jim Halverson, a professor at Northeastern University who advises PSI. He'll attend a summit at MIT next week where leading physicists will try to hash out a response to the new 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."

Turturean, who deferred a master's at MIT's computer science and AI lab to join PSI, said the delay gave him his first chance to sit with an AI-elicited solution and reshape it for human readers — because the academic community will expect him to be able to explain it.

"I'm totally accepting the idea that this might be the last paper I ever put out," Turturean said. "In a few months, like, why would you even put out a paper?" Reading the proof through, he said, made him realize: "oh my God, the weight of what AI can do."

For practitioners, the PSI result sets the template the next wave will follow: specialized harnesses over general models, Lean-style formal verification as the trust mechanism, and expert review before release. The open question — which next week's MIT summit will confront — is whether the institutions of science can keep pace with the machines now feeding them.