Controlling a quantum computer is a game of inference. Every measurement record tells the controller what the device is doing and what to play next. State preparation, feed-forward, error correction, calibration and stabilization all relate to learning, deciding, and acting, each within its own timeframe.

In this talk, we discuss the basic need for adaptive circuits, we show that these loops need different classical machines, and that we must have the right classical compute resources available within the needed time. We discuss how QM's Orchestration Platform allows for the needed seamless quantum-classical integration and how we continue to push the boundaries with our partners and community. We present results across qubit platforms and across classical compute resources and times, from calibrations performed in hundreds of nanoseconds, to AI-enabled autonomous calibrations.