Very nice visualization!

I thought this was going to be like a Game of Life variation I came up with about 15 years ago where you stack multiple layers of 2D Cellular Automata, allow each layer to have their own rule set, do single steps of each and then define the interactions between the layer as applying Boolean operations before doing the next step (e.g. set layer 1 = 1 xor 2, layer 2 = 1 xnor 2).

I'm pretty sure that's effectively a subset of what you can encode with a multi-state CA (you can after all interpret each layer as a "bit", so e.g. with two layers each point can be in "four states", meaning any give combination of rulesets ber layer + masking operation should have an equivalent four-state CA, but I never bothered to figure out how one would map from one to the other, and then the hard drive where the code was stored crashed so I forgot about it.