A cheaper neural decoder matches the old accuracy, but its speed edge is simulated
Three Yonsei researchers swapped AlphaQubit’s attention for a lighter Mamba design and kept its accuracy. The speed edge at today’s code sizes is simulated.
Notes
These are my reads on specific events: a filing, a deal, a program change. Each one starts from the source documents and says what I think it means for investors. For the science behind them, start in Learn.
Event reads
Allnewest first
Three Yonsei researchers swapped AlphaQubit’s attention for a lighter Mamba design and kept its accuracy. The speed edge at today’s code sizes is simulated.
Xanadu and Rolls-Royce cut the estimated T gates for a 20 qubit fluid workflow from 2 × 10²⁰ to 9 × 10¹⁵ with cheaper data loading. Nothing ran on hardware.
A UIC team designed a quantum search for a Minecraft seed and tested a tiny version on IBM and Quantinuum. The full run needs 1.65 billion two qubit gates.