For eight months, a physics problem has sat on a public tracker with no classical computer able to fully reproduce it. That problem, a simulation of disordered quantum matter run on IBM hardware, is now being presented by IBM and Algorithmiq as evidence that quantum machines can beat classical ones at something real.
The work models how information travels through a material with uneven local properties, a stand-in for catalysts and battery electrolytes. Researchers designed it so the couplings could be tuned on an IBM Quantum Heron processor, letting them steer where signals flow, localize, or interfere. The result has held against open challenge since it first appeared on the Quantum Advantage Tracker in late 2025.
Verification is the novel part. Classical simulators disagreed with each other on the same quantities, so the team built a trust framework around noise: they injected controlled errors, recalibrated gates, and repeated runs on multiple machines. Because the answers stayed consistent under deliberately altered conditions, they argue the output can be trusted even without a classical check.
Algorithmiq is publishing monoprop, its strongest classical method for molecular ground states, so outside groups can attack future advantage claims. CEO Sabrina Maniscalco calls the result the firm’s strongest published evidence yet, while IBM Research director Jay Gambetta sees proof that quantum systems can both outperform classical methods and be believed. IBM announced similar results the same day with the University of Chicago and Qedma.
