This is a bold claim - is your friend sure they're not "just" sharing some system calls, but still running both kernels?

I said they’re experimenting with Windows simply being a desktop environment on top of Linux. Only the Linux kernel would be involved. NTOSKRNL wouldn’t be present. I suspect if they’re trying to get native Windows apps to run, some call translation layer would have to exist (similar to what WINE provides).

Perhaps it is a bold claim, I haven’t seen it first hand. And it certainly has me hoping it goes much further than experimentation. I’d love to have a Windows environment on top of ext4 or zfs and the Linux kernel.

It would be an interesting proposition - yet the burden of legacy code is too much. Essentially we are talking about a WINE-ey situation here, where your true honest humble int WINAPI WinMain program's messages have to be translated to Linux kernal calls and vice versa. It would be a huge hassle for everyone involved plus the corporate crowd would absolutely detest it (unless every corporate bell and whistle was there and I very much doubt it).

Considering MS has the actual source to the Win API, they could very well complete what WINE does on a linux kernel.