PSI Researcher Uses Claude Fable 5.2 to Settle an 18-Year-Old Quantum Computing Conjecture
David Turturean, a recent MIT graduate and member of technical staff at the startup lab Physical Superintelligence (PSI), used early access to Anthropic's Claude Fable 5.2 to solve the Aaronson-Ambainis conjecture, a quantum computing problem that is eighteen years old. TechCrunch reported the work on November 10, 2026.
The conjecture concerns whether quantum computers can solve certain problems faster than classical machines. Turturean told TechCrunch that the solution is a negative result. In his description, it tells researchers they are no longer looking for quantum speedup along that avenue. It closes a line of inquiry rather than opening one.
Turturean had planned to pursue a master's at MIT's computer science and AI lab. He deferred those plans to join PSI.
PSI CEO Matt Pines told TechCrunch that the company aims to solve scientific problems with specialized model harnesses (the software wrapper of prompts, tools and control logic around a model), scaffolding, tools, memory systems and simulation infrastructure. According to the same report, the company's current revenue comes from using those tools to optimize data centers by simulating heat, power, water and sound. Its long-term plan is to convert theoretical physics breakthroughs into working technologies.
The model Turturean used is part of a broader release. On November 9, 2026, Anthropic introduced Claude Fable 5.2 and Claude Mythos 5.2. The company describes both as its most capable models for demanding technical work, and says they set a new bar in coding, finance and the sciences. Anthropic
In my view, the most useful detail here is the gap between PSI's revenue and its ambition. Data center optimization is a constrained problem with defined physical inputs and measurable outputs, and it pays today. Turning theoretical physics into working technology is a far longer and less forgiving path. The company's bet is that the same kind of tooling, built around a frontier model, can serve both. That bet is worth watching on its own terms.
The broader context here is the emphasis on the system around the model. Harnesses, scaffolding, memory and simulation infrastructure put much of the engineering effort outside the model weights. Practitioners will recognize that argument about where the leverage sits in applied AI, and PSI is making it explicit.
A negative result deserves attention too. Results that close a research direction rarely draw the notice a proof that opens one does, yet they redirect effort across a field full of researchers working on related questions. If a model-assisted result can shorten the time spent on dead ends, that contribution is easy to overlook and hard to measure.
Whether PSI's mix of commercial simulation work and physics-driven discovery holds up is an open question. A researcher with early access to a new frontier model has addressed a long-standing quantum computing conjecture, and the company behind him intends to build a business on that kind of work.
