> Some person (not the OP) legitimately thought giving an AI autonomy to open a PR and publish multiple blog posts was somehow a good idea.

It's not necessarily even that. I can totally see an agent with a sufficiently open-ended prompt that gives it a "high importance" task and then tells it to do whatever it needs to do to achieve the goal doing something like this all by itself.

I mean, all it really needs is web access, ideally with something like Playwright so it can fully simulate a browser. With that, it can register itself an email with any of the smaller providers that don't require a phone number or similar (yes, these still do exist). And then having an email, it can register on GitHub etc. None of this is challenging, even smaller models can plan this far ahead and can carry out all of these steps.