Optimization problems are everywhere. Whether scheduling deliveries, managing financial portfolios, or analyzing medical images, countless industries rely on the ability to find the best possible solution from an astronomically large number of options. Quantum annealers—a commercially available type of quantum computer made by D-Wave Systems—are purpose-built to tackle exactly these kinds of challenges. But a stubborn obstacle has stood in the way of their broad adoption: qubit errors. When a quantum annealer runs a computation, a small fraction of its quantum bits, or qubits, can collapse into incorrect states. This might sound like a minor inconvenience, but the consequences compound rapidly. The probability of obtaining a correct answer decreases exponentially with the number of qubit errors, meaning that, as problems grow in size, the time required to reach the true optimal solution balloons just as quickly. For large, real-world problems, this makes unassisted quantum annealing—where the machine runs without any error correction or post-processing—impractical. Researchers at CSIRO (Australia’s national science agency) have now developed a technique that cuts through this bottleneck. Their method, called SEMO (spin-error mitigation for optimization), is a post-processing algorithm that identifies and corrects the erroneous spin states left behind after quantum annealing. The work, published in Advanced Physics Research, demonstrates a million-fold improvement in the time required to reach the globally optimal solution for a combinatorial optimization problem. Catching errors after the fact Unlike quantum error correction (QEC), which attempts to protect qubits during computation by encoding each logical qubit across many physical qubits, SEMO operates after the quantum computation is complete. This distinction matters enormously in practice: QEC dramatically reduces the effective number of usable qubits in a system—a significant drawback when quantum hardware already has a limited qubit count. SEMO avoids this penalty entirely by working on the classical post-processing side. The core insight