He is talking about Bearblog, the blogging platform that he hosts his website with. Bearblog has it's own discover page and ranking system and every once in a while a post on Bear about AI/tech/whatever will be posted on Hackernews and all of those new people will "upvote" the Bear post and send it to the top of the Bear discover page.
Easily half the other events on the calendar are kid-related. Of course it's going to infer that, absent other direction, the most likely overarching theme of the visible events is "child care".
Sure, but the LLM needs to prove that it can make inferences as well as or better than a human, in order to be useful. Aside from that, it's not human, so there's no need to be fair - it should do what we tell it, not decide on its own.