It seems like CPU peripherals are gaining more and more intelligence. I was reading about PCIe's potential successor CXL and the amount of features that allow peripherals to do their own thing with a minimum of CPU intervention seems to be increasing.

So at some point I bet you'll be basically uploading BPF like code to peripherals directly in order to accomplish really fast I/O. Don't know how that's going to look at all from a post-POSIX API perspective.