AI research and quantum computing · Sources checked November 7, 2026
Physical Superintelligence (PSI) reported on November 6 that its research system, using a non-public Anthropic model, found a proof of the Aaronson–Ambainis conjecture. The announcement describes a mathematical result with implications for quantum computing, accompanied by material that other researchers can inspect.
For readers following AI progress, the useful question is what was checked and what conclusions the result supports.
What PSI reports
PSI says Lean, a formal proof assistant, accepted the proof. David Turturean developed the research system, led the work and reviewed the paper; PSI names him as responsible for its correctness. The company identifies the model only as non-public, so the announcement does not establish which publicly available Claude product could reproduce the work.
The quantum-computing scope
The original Aaronson–Ambainis paper studies quantum speedups in a query model, which counts accesses to input data. It connects the conjecture to simulating a quantum algorithm on most inputs with a classical algorithm using polynomially many queries.
That scope matters. A bound on queries in this model is not a measurement of how fast practical computers run. It does not establish that ordinary computers can efficiently replace every quantum algorithm. Readers should be wary of a headline turning this result into a verdict on all quantum computing.
What formal checking establishes
The public repository supplies the paper, Lean code and verification records, and describes the work as a preprint. Its authors report checks using multiple kernel implementations.
The repository also spells out the limits: readers must confirm that the formal statement matches the intended conjecture. The explanatory paper is not itself machine-checked, and some passages have no formal counterpart. The checking software, mathematical definitions and underlying computers remain part of the trusted foundation.
A useful habit for evaluating AI results
When an AI research announcement attracts attention, separate three questions:
- What is the exact claim? Keep the problem, assumptions and scope attached to the headline.
- What evidence can others inspect? Look for an actual artifact and a documented checking process.
- What would change the assessment? Follow expert scrutiny of the statement, proof and interpretation.
For practical AI work, apply the same discipline to code changes, calculations and reports: decide how the result will be checked before treating it as dependable. This article reports PSI’s claims and published materials; it does not independently validate the proof.