Figure 11. Sensitivity of reaction time to decoding latency and speculation accuracy. Speculation accuracy of 0% corresponds to baseline (no speculation) decoding. Decoder latency is $t_{\text{dec}}(v) = rv/d^2$ rounds, where $v$ is the volume of the decoding problem (in units of $d^3$) and the relative latency factor $r$ is varied along the x axis. Equivalent latency factors extracted from linear fits to PyMatching latency data (Figure 3) are shown along the x axis. Right: simulator trace of 10 T gates on an idling logical qubit. A similar experiment with 1000 T gates was used to collect the data in this figure.