What isn't clear from this snippet, that I think is incredibly cool, is that all of the expressions within a function (here, called processes), actually run in parallel.

In this snippet:

   count(N) :- N =< 10 | even(N), N2 is N + 1, count(N2).

count(N) is defined defined as a process that takes an argument named N, and if N is less than or equal to 10, the process changes state to a random process on the right side, and is also forked to become the other two processes.

Where it gets weird, is that the resulting three processes have data dependencies between them, so if the third process were executed first, count(N2), the dependency on N2 wouldn't be satisfied, and so that process would suspend, and a new process be chosen for execution.

It's easy to look at this line of code and think that the three expressions will execute in the order they're written, but any interleaving of them is possible, with that sort of non-determinism being built into the language by design.

I'm currently reading the Strand book, and it more or less describes the language as being Prolog, but without unification and backtracking: instead treating dataflow as being the basis for the computational model.