Real-time quantum error decoding must process correction signals within microseconds, but current decoders run 10-100x too slowly

For quantum error correction to work on a running quantum computer, the decoder -- the classical system that reads error syndrome measurements and determines which corrections to apply -- must operate in real time, producing correction signals faster than new errors accumulate. For superconducting qubits with coherence times of 50-300 microseconds, this means the decoder must complete each round in single-digit microseconds. Current software decoders typically require tens to hundreds of microseconds per round, making them 10-100x too slow for real-time operation. Why it matters: Because decoders cannot keep up with error accumulation rates, error correction is currently only demonstrated in slow, post-processed experiments rather than in real-time on running computations, so the 'below threshold' error correction results reported by Google (Willow) and IBM are not yet usable for actual computation, so the transition from 'error correction works in principle' to 'error correction enables useful quantum algorithms' remains blocked, so the fault-tolerant quantum computing era that the entire industry roadmap depends on cannot begin until this decoding speed gap is closed. The structural root cause is that optimal decoding of quantum error correction codes (like the surface code) is computationally NP-hard, so practical decoders must use approximations. The best-known approximate decoder (Minimum Weight Perfect Matching) has superlinear time complexity that grows with code distance. Building a decoder that is both fast enough and accurate enough requires custom hardware (ASICs or FPGAs), but the error correction codes and qubit architectures are still changing rapidly, so investing in custom decoder hardware is risky when the target keeps moving. Riverlane is attempting to build a universal decoder hardware platform, but this is a bet on stabilizing the QEC code landscape.