It actually works really well for some viewing patterns. For myself, it's the best recommendation algorithm. When I'm listening to 80s rock, it gives me a bun

It actually works really well for some viewing patterns. For myself, it's the best recommendation algorithm. When I'm listening to 80s rock, it gives me a bunch of other 80s rock, notably including things I've enjoyed listening to before so I can just yoink that right into my playlist. Also throws in some 2000s dubstep and some 60's country that I like and a couple of things that I've never heard before in case I've started out on 80s rock but just am not feeling it tonight. When I'm watching thought-provoking mini-documentaries, it gives me more by the same creator along with some random related stuff. Some of that is stuff I've watched before but might want to re-watch or share with someone. Or to compare the thoughts of this guy vs that guy. Your usage pattern might be different. But for many of us, it works quite well. I would assume they're optimizing for the most common use cases. And yes, I liked that video, I do want to see it again.