I don't at all get what the hype is about from this article, the concept of "you can combine two things and get a thing" isn't novel or radical.

I get a feeling that the kicker is about associativity, which is to say "when combining a bunch of things in a row, you can combine them any way as long as you only combine the adjacent ones". The kicker being that if you have that, you can parallelize.

So, seeing associativity allows one to identify parallelizable problems.