We already know that the Windows 10 April 2018 Update has been the most rapidly deployed of the major Windows 10 updates. Microsoft today said a little about how this speed was achieved—and made the update fully available to every Windows 10 system, representing the final stage of the rollout process.
The rollout of each major update is performed gradually. Microsoft uses information about successes, failures, and incompatibilities collected from the earliest systems to receive the update to determine whether it should be made more widely available. For this update, there was an extra factor in the mix: machine learning. The company built a machine learning model to identify which system characteristics meant that the update was likely to succeed. With this model, viable candidate systems could be more rapidly identified, in turn enabling the update to be more aggressively pushed to those systems. The result was fewer rolled back
fewer crashes, and less negative user feedback.
During the deployment of the update, incompatibilities were detected. As an example, Microsoft says that a version of Avast Behavior Shield caused a reboot issue. The immediate fix was to blacklist systems with the problematic software, meaning that within 24 hours of the problem being detected, at risk machines were no longer being updated. Avast subsequently fixed its software, and with the corrected software in place, the deployment could be resumed.