4 ms·

I share the concern but as a web developer, it's pretty clear that keeping people watching a tight, edge-cacheable pool of videos is far cheaper than actual org

I share the concern but as a web developer, it's pretty clear that keeping people watching a tight, edge-cacheable pool of videos is far cheaper than actual organic usage.

ericmcer 4y ago
Wow I can't believe I had not thought about cache-ability as a factor. That makes a ton of sense.
bavell 4y ago
This is my hunch too. By heavily curating all content, a tiny portion will be served to a significant fraction of users, keeping costs lower by reducing the long tail as much as possible.
dvlsg 4y ago
Suggesting videos that users don't end up watching is even cheaper. Not a great user experience though.
oliwarner 4y ago
This started me wondering how Shorts/Reels/etc play into that paradigm. They're a major part these sites these days. But they're shorter, smaller videos with tighter compression. 1min ~= 5MB, and it's easier to mix people through a pool of these (even allowing for content skews) with a little 200TB edge cluster able to hold 76 years of contiguous programming. Probably more efficient than full-format videos.
tobinfekkes 4y ago
This is a great point. And additionally, when the storage and compute are cloud-based, and they get charged on a per-compute basis, you can start to see why every platform's search has gotten worse: they're incentivized to restrict the amount of searching that happens, because each search requires more compute, and probably compute of things that aren't cached already? So you limit the amount of search and recommend the cache-able content, and cha-ching. This is why searching for transactions in Stripe is a disaster, or finding email history in SendGrid is a pain, or any other search function on a sufficiently large data set. Remember the compute costs: that's why it's purposefully built to be hard to find what you're looking for.

4 ms·

I share the concern but as a web developer, it's pretty clear that keeping people watching a tight, edge-cacheable pool of videos is far cheaper than actual org

by oliwarner 4y ago

I share the concern but as a web developer, it's pretty clear that keeping people watching a tight, edge-cacheable pool of videos is far cheaper than actual organic usage.

ericmcer 4y ago
Wow I can't believe I had not thought about cache-ability as a factor. That makes a ton of sense.
bavell 4y ago
This is my hunch too. By heavily curating all content, a tiny portion will be served to a significant fraction of users, keeping costs lower by reducing the long tail as much as possible.
dvlsg 4y ago
Suggesting videos that users don't end up watching is even cheaper. Not a great user experience though.
oliwarner 4y ago
This started me wondering how Shorts/Reels/etc play into that paradigm. They're a major part these sites these days. But they're shorter, smaller videos with tighter compression. 1min ~= 5MB, and it's easier to mix people through a pool of these (even allowing for content skews) with a little 200TB edge cluster able to hold 76 years of contiguous programming. Probably more efficient than full-format videos.
tobinfekkes 4y ago
This is a great point. And additionally, when the storage and compute are cloud-based, and they get charged on a per-compute basis, you can start to see why every platform's search has gotten worse: they're incentivized to restrict the amount of searching that happens, because each search requires more compute, and probably compute of things that aren't cached already? So you limit the amount of search and recommend the cache-able content, and cha-ching. This is why searching for transactions in Stripe is a disaster, or finding email history in SendGrid is a pain, or any other search function on a sufficiently large data set. Remember the compute costs: that's why it's purposefully built to be hard to find what you're looking for.