The important this is that, as mentioned in the article, you need to have a way to converge states in a way that obeys certain properties.

Yeah, as I’m reading this more thoroughly I see that there wouldn’t be many states to express, and the merging itself isn’t something you’d express in states.

I initially thought more of the inner workings could be managed this way, but it seems better implemented as it is in the article.

You should check out the P language and consider the implications of CRDTs on the `machine` constructs :)

This turned out to be super interesting. Thanks!