An Unreleased Anthropic Model Helped Prove a Quantum Computing Conjecture

Physical Superintelligence Inc., a physics research lab, said on Friday that it used an unreleased model from Anthropic to prove the Aaronson-Ambainis conjecture, a question in quantum computing that has stayed open since 2008. Business Insider reported the lab's announcement, which frames the result as a genuine mathematical advance rather than an engineering demo.

The conjecture is essentially an educated guess that had never been proven. It concerns when quantum computers can be significantly faster than ordinary computers, which makes it one of the most fundamental questions about the boundaries of quantum advantage.

According to PSI, David Turturean, a member of the lab's technical staff, developed a research system that included the Anthropic model, and that system produced the proof within three days. The lab described the outcome as “settling one of the most fundamental questions in quantum computing.”

“Quantum computing still has plenty of mysteries, but now there is one fewer,” the lab said in its announcement. The framing matters: the contribution is a narrower map of what is unknown, not a new piece of quantum hardware.

The substance of the result is equally notable. The proof, PSI said, shows that certain exponential advantages of quantum computing require special patterns or structure, meaning quantum computers have limits, which was long suspected but not previously established. Quantum machines would still be a major upgrade for some tasks, and existing hardware remains at an early stage of development.

Researchers who had previously worked on the problem described the proof as a “huge result” that resolved a decades-old question. Guy Blanc, Jordan Docter, Carmen Strassle and Li-Yang Tan of Stanford University said of the proof, according to PSI's announcement, that “everything is brought together in a rather magical way.”

The crucial technical detail is missing, however. It is not clear which unreleased Anthropic model PSI used, and Anthropic did not respond to a request for comment from Business Insider. That leaves outsiders unable to judge how much of the work the model actually did, or how the research system around it was assembled.

The result lands in a fast-moving stretch for AI-assisted mathematics. In September, OpenAI said it had solved a 90-year-old math problem, the Navier-Stokes equations. Terence Tao, a prominent mathematics professor at UCLA, said last month that math's “romantic era” was over, and MIT researchers who polled graduate students and postdocs found 76% said they were unlikely to stay in academic math after graduating. What to watch next is whether PSI publishes the proof and names the model, and whether the mathematics community can verify the result independently.