[hawtkey]

I think the main target is ML applications that depend on GPU access and being able to run a WM is just a side-benefit.