Is it really useful or is it just a meme?
I understand the basics of it, but I really fail to see any benefit of it over traditional procedural generation. Yea sure it's nice that you can just draw whatever and it will generate more of it but uglier, but that isn't useful in any capacity. Sure it's nice that you can directly add more constrains to it in seemingly easier way than engineering the whole procedural code to uphold these constrains. But the more constrains you add the more likely it is to cause contradiction and force you to restart an already very computation intense algorithm. And this can majorly fuck you up if you have an endless world which you want to expand in runtime and can't really start from scratch. To do anything interesting with it, you'd have to create so much constrains and design them in such a way that will always allow to completely avoid contradictions, that you might as well just drop WFC at this point and write normal generation code. Especially considering the fact that you need to divide your world in some form of tiles and these tiles only interact locally, except for constrains(which wouldn't be as trivial to write to eg get buildings that looks normal, symmetric instead of everything being organic and unpredictable).