That doesn't make a lot of sense given the novel hardware that runs in a lot of chips these days. TPUs, neural accellerators, perceptron branch prediction, flash based neural networks.

Thise are not abstractions over existing posix APIs or von neumann architectures, they are novel ways to do computation from a different branch of math.

Heck we don't even know if some of these are not breaking the church turing thesis by computing with true real numbers and not the computable subset that turing machines can handle.