Facebook puts a lot of time and effort into obfuscation of course code. They generate random classes and IDs and interweave containers (divs, spans, custom HTML elements) on each page load so that software can't tell what is content and what is an ad.
What an absolute shame that they waste the time of all their highly-skilled developers on garbage like that. The user has specifically indicated, by running an ad-blocker, that they aren't interested in being marketed to. At least for me, ads that bypass my blockers really irritate me, to the point that whatever company was unfortunate enough to win that bid actually hurts themselves as I will be more likely to select a competitors product.
It is definitely impressive. But it's something that I think could be defeated pretty easily with a pretty simple machine learning model. Could be a fun project!
It's actually a really neat trick.