Search engines seem to index HN in near real time. They must have custom scraping code to follow the incrementing post IDs.

Note that these topics often involve comments which you can predict very easily. Internet users are like that, agenda or no. Wasn’t it in the heyday of forums that you could recognize the most prolific/annoying members by their style and vocabulary? A model should have no problem pulling such things off.