A quantum computer runs on qubits, which are extremely sensitive: any noise can corrupt the information. To fix that, researchers use special codes. One of them, the surface code, lays qubits out on a flat lattice, like tiles on a floor. When an error occurs, a defect shows up on the lattice, meaning a glitch in the parity check.
In the past, a big classical computer collected all the defects. It looked at the whole lattice and decided how to remove them. But the more qubits there are, the harder it is for that computer to keep up. Now scientists have come up with something different: let every node of the lattice know on its own where to move the defect. It is as if each floor tile pushed the trash into a corner by itself, without asking the manager.
This approach works even during a computation: when the computer loads data, couples qubits or reads out the result. And it takes few resources: a constant number of classical bits per node. A quantum computer like this could be built right on a chip, and it would correct errors by itself, with no data sent outside.
The decoder uses a principle first applied in the Bravyi-Haah broom algorithm: defects are swept toward the corner of a cluster, like trash to a wall.
The idea of local error correction without a central controller is reminiscent of the decentralized systems in Robert Ibatullin's novel "The Rose and the Worm," where computing networks self-repair after damage.