Youtube is very different than Netflix from a technical problem perspective. They serve free videos to anyone around the world that are uploaded by users.

It's closer to a live streaming problem than pre-encoded video like Netflix.

Having worked at Netflix I can say that the YouTube problem is much more complex.

I wonder what portion of Youtube's request traffic can be served with cache servers at the edge with a few hundred terabytes of storage. There's a very long tail but i would guess a significant portion of their traffic is the top ~10,000 videos at any given moment.

There was a Google organised hackathon on this topic. Given a set of resources, locations, and (estimated) popularity, Optimise for video load time by determining what should be moved to the cache when and where.

did anything from the hackathon turn out to be used?

I’m not sure as I don’t work at Google and didn’t see any follow up. Maybe some in the winning teams would know.

I do think there is some temporal logic already present in Google’s algorithm which wasn’t part of the challenge.

Not even close. YouTube has orders of magnitude more content and vastly more users. Google Global Cache was the inspiration for Open Connect.