A quantum error-correcting code only pays off if the logical error rate keeps falling as the code grows, and that holds only when the hardware is already below a critical error rate. Establishing it on a real device means measuring the ratio of logical error rates between successive code sizes, showing the ratio exceeds one, showing the encoded qubit outlives its best physical constituent, holding that performance steady over hours, and decoding fast enough to keep up with the machine. This is what turns error correction from a scaling argument into an engineering claim, and each of those five conditions can fail independently.
Google Quantum AI and Collaborators (Google Research and Google DeepMind, with academic co-authors at Massachusetts, UC Santa Barbara, Connecticut, Auburn, ETH Zurich, UC Riverside, Yale and MIT), ar…