@anzu you are going to like it, I suppose (if you are not the author, yourself :-)

@Shamar the only instruction you need is NAND, that alone can implement all other logic gates. See also: https://www.nand2tetris.org/

Also there are Field Programmable Gate Arrays (FPGAs), where you have a grid of memory but each memory cell can also act as a gate or a path between cells, so you can basically “draw” your circuitry into the memory like you would draw pixels, and the FPGA becomes that circuit. But when programming FPGAs you do have to worry a bit about how you layout your circuits. I don’t personally understand why, but I think it has so that you don’t introduce crosstalk between the memory cells. So FPGA programming is pretty complicated.

Personally, I think as long as the circuitry implements instructions that are simple enough to understand, I am fine with the complexity introduced by modern superscalar computer architectures. The idea of permacomputing really is that we should assume that the supply chains we have now will eventually collapse, and in the future we will be stuck with only the set of all computer chips that have been manufactured, so we will have to treat these chips as incredibly valuable resources, and try to make each one last for many generations, possibly hundreds of years.