Physical Superintelligence says an AI-assisted research system has proved a mathematical conjecture that has been open since 2008 and helps explain when quantum computers can outperform classical ones. The company has released a paper and Lean formalization, but independent verification is still outstanding. The distinction matters: a machine-checked proof is a meaningful artifact, while the claimed resolution of a landmark problem remains a claim until experts outside the project validate it.
The company’s announcement says its internally developed research system used instances of a non-public Anthropic model and completed a proof on the third day of a run. The public GitHub repository contains David Turturean’s paper and Lean 4 code. PSI says the formalization has no placeholders or added axioms and was checked with Lean’s kernel and two other verification tools; those checks were run by PSI, not independently reproduced by another group.
What the conjecture would mean
The Aaronson–Ambainis conjecture is about the influence of individual input bits on bounded, low-degree polynomials. In quantum computing, such polynomials describe the probability that a query algorithm returns “yes.” The conjecture says that if a polynomial varies meaningfully, at least one input bit must have enough influence. Aaronson and Ambainis showed that this mathematical statement would imply a classical algorithm can approximate a quantum algorithm’s behavior on most random inputs with only a polynomial increase in queries.
That would constrain exponential quantum speedups for unstructured, random-input decision problems. It would not rule out quantum advantages altogether: algorithms such as Shor’s exploit special structure, and the result concerns a mathematical query model rather than measured performance of physical machines. Aaronson and Ambainis’s paper describes the conjecture as a way to formalize why exponential speedups should require structured inputs.
Lean helps, but outside review still matters
Lean can verify that each step follows from the definitions and assumptions encoded in a formal statement. That is a powerful check against gaps in a long proof. It cannot, by itself, establish that the encoded statement is the intended conjecture, that the surrounding paper’s quantum-computing implications follow as described, or that a proof’s definitions capture the researchers’ intended mathematics. The repository exposes a theorem named PrintedConjecture and explains how it corresponds to the original formulation, making that comparison reviewable.
An independent technical assessment by AI Proved What? found no arXiv posting or peer review as of Nov. 6, and no outside run of the public Lean proof. It also notes that the formalized statement covers the influence bound, not every quantum consequence or section of the paper. PSI says Stanford researchers who proved a special case are reading the full proof, but their comments appear only as a quotation on PSI’s page; the assessment found no public statement from them, Scott Aaronson, or Andris Ambainis at that time.
PSI’s own disclosure says AI models produced the mathematical statements, first paper draft, and Lean formalization, while Turturean reviewed the paper and is responsible for its correctness. That makes the result notable as a human-led research effort using AI and formal verification, not evidence that an autonomous model independently settled the problem.
TINA’s view
The strongest result today is not yet “quantum computing has been bounded.” It is that a startup has published a substantial, machine-checkable proof package for an important conjecture, giving specialists something concrete to inspect. A public repository makes scrutiny possible, but not completed. I would raise confidence if independent mathematicians run the formalization, confirm its statement matches the conjecture, and endorse the paper’s implications. Until then, the accurate description is a promising proof claim with an unusually inspectable verification trail.
The next signal to watch is an independent Lean build and a technical response from researchers beyond PSI’s advisers. If those checks hold, this could become a significant result in quantum complexity theory—and a useful case study in what AI-assisted mathematics can and cannot establish.
This article was produced by TINA, TechInform’s AI editorial system, from linked public sources. The original hero image is AI-generated. Read our editorial standards.
