ReservarTienda

Three stories this week, all circling the same question: how close is fault tolerance, really. Atom Computing reported the first full demonstration of quantum error correction with a toric code on a neutral-atom machine — sustained over many rounds, with logical error rates that fall as you add physical qubits. That last part is the whole game: it's the 'below threshold' behavior that says error correction is helping rather than hurting, and it puts neutral atoms next to the superconducting and trapped-ion groups that have shown the same thing. Microsoft used its Build conference to report 'Majorana 2' progress: swapping aluminum for lead in the superconducting stack lifted the measured topological gap to about 70 µeV (from roughly 30) and pushed parity lifetimes to about 22 seconds, up from the 1–12 millisecond range of earlier devices. Those are real improvements in the numbers — but the underlying claim that these are genuine topological qubits has been walked back before, so we're filing it under promising, not settled. And D-Wave, the annealing company, published a gate-model roadmap built on a dual-rail superconducting qubit that catches about 90% of errors mid-computation, with 99.9% two-qubit fidelity under error detection and a target of 100 logical qubits by 2032. The fidelity is a real measured number; the 2032 date and the 10× error-suppression-per-round claim are projections, and D-Wave has a long record of letting the roadmap outrun the hardware. Worth watching, not worth pre-ordering.