Riverlane published benchmark numbers for its Deltaflow 2 error-correction stack: mean QEC latency of 16.32µs, a 4× improvement over Google Willow's 63µs benchmark and a 4× improvement in decoder throughput. The system handles up to 250 physical qubits and uses Riverlane's Local Clustering Decoder running on custom hardware — this is the layer that has to run fast enough that error correction doesn't become the bottleneck before the computation finishes. Separately, Los Alamos researchers showed that quantum learning models designed to avoid barren plateaus — a training problem where gradients vanish — turn out to be classically simulable, which deflates a substantial part of the quantum machine learning literature that assumed avoiding barren plateaus was sufficient for quantum advantage. And IQM opened its first US Quantum Technology Center in Maryland's Discovery District, partnering with the Capital of Quantum initiative to embed IQM hardware and engineering in one of the denser quantum ecosystems in the country.

• Riverlane Demonstrates Real-Time QEC Latency Performance Advancements

  Quantum Computing Report

• LANL: Some quantum learning models are classically simulable

  Los Alamos Reporter

• IQM Establishes First U.S. Quantum Technology Center in Maryland

  The Quantum Insider