Why haven't anyone randomly generate a bunch of code of various kinds, use LLM to create some summary of it and dump them to github with a restrictive license? The patent office isn't here to enforce anything.

It'd be like copyright trolling the Library of Babel. The set of useful programs would be totally eclipsed by incoherent gibberish (even if there were a means to ensure that the randomly generated code were syntactically correct). In other words, the signal to noise ratio would be microscopic and running this scheme in finite time would effectively result in zero valuable code being successfully squatted.

I'm curious as to where this is going. What happens next?