Researchers gave Claude specifications for eight quantum circuits drawn from chemistry, optimization, physics simulation, and number theory, along with an automated evaluator that scores each design. Working on its own, Claude found designs that improve on the best published circuits, in some cases by an order of magnitude or more.
Claude Mythos 5 worked in a closed loop: it proposed a circuit design, received feedback and a score from the evaluator program, and used that feedback to improve its next attempt. Across eight computational primitives, Claude discovered novel designs that used both fewer qubits and fewer gates than the best previously published algorithms.
In each case, the improvement resulted from uncovering new structural and mathematical insights about the problem. For example, when compiling a circuit for simulating an iron-sulfide molecule, Claude noticed that the molecule’s symmetries and repeating structure meant that some results could be calculated once and reused. In other cases, it found that existing circuits included costly components needed to handle only rare inputs, and it redesigned them without those components.
Designing a quantum circuit involves two constraints. First, a quantum circuit cannot make if-then decisions, so the circuit must include operations for all inputs. Second, a quantum circuit cannot discard information, because every step must be reversible. The only way to discard information is to uncompute it, at the cost of additional gates.
Claude discovered a new way to perform modular inversion, a step in elliptic-curve calculations, and made the costliest step of a leading quantum optimization algorithm about 12x cheaper.