realistically how feasible is it going to be to have something be completely free of it?

It'll be feasible if we encourage a culture of not using it. It doesn't have to be the main mode of development — and all the big names can keep their slop generators — but as long as there remains a demand for slop-less software, there will be people willing to make it happen.

There's also the saying "don't let the perfect be the enemy of the good": besides the fact that I don't trust slop-coders to put in any more effort into architecture/security/performance than they do actually learning the languages and writing the code, I also hate that they are willing to fund giant data centers that deplete local reservoirs and cause blackouts for small communities. In this case, I don't care if it's "no ai at all" or "no ai as much as is practical," because both are still better than "full steam ahead."

They likely won't catch every bit of AI generated code, but I don't think that's the point. In my view the point is to take a stand and try. It also signals that the authors care about their code in a certain way.

There are all sorts of things that are impossible to perfectly enforce, but that doesn't mean we just give up. Insider trading is extremely difficult to prove, and I bet more people get away with it than don't, but I still think making it illegal is a good policy.