Agree, I think the problem with recommendation engines is a different one: they are not really focused on serving content that would be valuable to you, but one that will help increase metrics you probably don't care about or that are even potentially bad for you e.g. usage time

Yeah. It really helps getting users addicted.

Sad thing is, this is likely difference of degree more than that of a kind. That is, recommending you the content that's valuable and beneficial to you would still be net better for "engagement" and ad impressions, than random/no recommendations. It's just not as good as recommending lowest common denominator hyperoptimized addictive garbage.

For me it never worked. I don't understand people's fascination with YouTube, because the recommendations have never hooked me.

Instagram is much better at finding shorts, but I would prefer long form content like on YouTube, but decent recommendations.

Maybe it doesn't work so well when you are not a heavy user and the algorithm doesn't have a lot of data to work with. (I personally have spent way too much time there in the past, which makes the recommendations even better, and so on, leading to addiction.)

As far as I know, YouTube always assumes one account is used by one person. Since I use it mostly on the television, this isn't true. It stays logged into my wife's account, but we both use it. So it's constantly fighting between me going in and watching sports highlights versus her going in and watching ASMR hands in slime face massage simulation videos. The recommendations are schizophrenic and all over the place. Other streaming apps that were meant for television first allow you to assign multiple user profiles to a single account.