Quantum error correction requires resource states — ancilla qubits prepared in specific configurations — to implement logical gates. Preparing perfect resource states is expensive; postselecting on quality is cheaper but traditionally uses fixed-size checks that don't scale with the code distance. The overhead per logical gate stays high.
Note (2026-11-12): checked against the paper's full text. The method and the headline match: postselecting resource states whose size grows with the code distance, using a soft decoder output called the partial gap, with a fourfold reduction in the overhead per logical gate at the same logical error probability. Two points are overstated in the essay. The paper cites earlier work that already used soft decoder output (the logical or complementary gap) for postselection, so 'the information existed but wasn't being used' is not its claim; what is new is applying a gap-based metric to sub-circuits whose size scales with the code distance, with the partial gap estimating the logical gap after the resource is consumed. And the fourfold figure is for implementing gates by teleportation through a cluster state, in a context where idle errors are not significant, analysed at a fixed rejection rate; the paper says realizing it needs hardware with idle coherence times much longer than the time to prepare, measure and postselect a resource state.
The fix is to postselect at the right scale (arXiv:2603.08697). Instead of checking fixed-size resource states, directly postselect sub-circuits whose size scales with the code distance, using decoder soft information to predict quality. The key metric is the “partial gap” — a quantity that predicts how much the logical gap changes after consuming a resource. Checking this before use yields a 4x reduction in overhead per logical gate at equivalent logical error probability.
The structural insight: the information about resource quality already exists in the decoder's soft output — it just wasn't being used. Standard postselection discards resources that fail a binary check. Partial-gap postselection uses the continuous quality signal that the decoder already computes. The overhead reduction comes not from better resource preparation but from better resource selection. The waste was in the decision, not the production.