Quantum error correction is a crucial component for achieving scalable and fault-tolerant quantum computing, with topological codes offering a promising approach due to their hardware-efficient architectures. However, the Tanner graphs of these codes, such as the toric code, contain numerous short cycles that significantly degrade the performance of traditional belief-propagation decoding methods. To address this limitation, researchers have proposed a novel decoding technique called neural belief-matching decoding, which aims to improve the accuracy and efficiency of quantum error correction for topological codes1. This new approach has the potential to enhance the reliability of quantum computing systems by mitigating the errors that occur during quantum computations. The development of more effective decoding methods is essential for overcoming the challenges associated with quantum error correction, so what matters most to practitioners is that this new technique could lead to more robust and reliable quantum computing architectures.