Vole released two new Windows 10 builds to Insiders and what appears to be happening is the outfit is implementing a new cloud-based logic for the Windows Update system, in order to avoid unexpected updates when you really need to get work done on your computer.

The update uses  a predictive model that will improve over time in order to better understand when you are going to use your device. This way, Windows can make sure it’s not disrupting your work and install the update when you are actually expecting it to. It will consider contextual things like if you were currently using your device before restarting or try to predict when you move from your device to grab a coffee, etc.

Vole says the company found the new model to produce “promising results” internally. The reliability of the system is still up in the air, of course — but since it’s a deep learning system, it is going to improve over time as Microsoft trains the model with more data. This way, the system will not only be more reliable for everyone over time, but it should also be more personalized to your own habits and device usage. Still, Windows will have to install updates eventually, so it may still seem “unexpected” to some. Either way, the feature should ship with Redstone 5, which is expected to arrive for the public later this year, sometime around October.

Annoying updates have long been a beef with Windows 10 users.  Microsoft has implemented features like Active Hours to help avoid this kind of behaviour in recent years, though most consumers aren’t really aware of the feature’s existence. It sounds like a more intelligent system might be more useful.