You’re making a lot of assumptions there; does media production (hence consumption) look the same on a non-central-algorithm landscape?
I believe you can’t judge the system from within the system; you can only build parallels and let new dynamics (if any) emerge, but I doubt the modern relationship with information is inherit to the human condition.
We shall see.
Also you have to separate YouTube the media storage platform and YouTube the content aggregator. These are 2 functions YouTube integrates today.
A decentralized solution may handle the latter function with a mix of creative and traditional media storage.
Yeah, the storage part is 100000x easier than the distribution part; no question.
Anyone over 21 would have to be kidding themselves if they think that company X’s dominance means it’s going to be there forever. Makes me laugh every time I see someone express this sentiment because it’s a current market (loss)leader.
Im just waiting for Meta to bite the dust and see who inevitably replaces them.
As for the storage, I am quite fond of the idea of using IPFS; peer-driven content delivery.
The aggregator, on the other hand, hm... tricky. Locally ran algorythmns that ingest incomming new posts from, for instance, a Nostr stream?
Yeah I was playing around with that.
Could be effective by transposing a persons profile as their “NostrTube” channel.
Thought about writing a proof of concept that pulled in notes with video media links and also a place for a custom Kind that had metadata associated.
I'd be in support of that! Maybe use the "extra" field in kind:1 in combination with a tag in the "tags" field to make these both discoverable and visible to normal clients. Or just use a tag that links to a special-kind note insted where all metadata can be found, so that clients that support it, can show it as an embed.
Meanwhile, just querying a relay for `["REQ", ["kind", 999]]` would return only those "video notes".
I think there's something in this =)
NIP-89 does this a million times better if I may say so myself 😂
I completely forgot about this XD. Just re-read it, and i feel defeated. :) But yes, yes it does.
Don’t. This is what’s great about the nostr eco system and nostr:npub1l2vyh47mk2p0qlsku7hg0vn29faehy9hy34ygaclpn66ukqp3afqutajft more than most has experience writing so many different apps.
I take it as corrective nudging in not reinventing the wheel
Ofc, these discussions are useful
Why would you want/need an agggregator? That’s the whole point of nostr 😅
Maybe not using correct terminology and phrasing
Nostr front end that pulls in just videos specific posts.
We have kind 1 stuff. Maybe we have a front end for just videos.
I’m all for specific apps where maybe I use Damus for what I’m using Damus for now but I can go cut through posts to find videos from creators on another front end.
I’m probably using the wrong terms…
No errors seen, all good man. :)
Discovering new things related to something I just watched or combing through a specific topic - just two things that I can immediately think of.
An aggregator is great when you dig yourself into a new topic. For instance, say you want to learn more about Speedrunning. Youtube's algo can aggregate a lot of speedrun videos for you and you can chose where to start digging the newfound rabbit hole. x)
This is, unironically, a great feature and why I think an aggregator has it's merit.
Topic-based relays.
You don’t need top-down order when you can have emergent order.
You can be fully in control of how you design your relays sets that give you the view you want/need to see.
That's completely fair. I have only see the same model for years - even aside from Youtube with a german-ran platform named Clipfish years ago. The rough idea was the same as what Youtube is.
Honestly, I do have to think on the "non-central algorythmn" part. Yes, a lot of what drives consumption on Youtube is based off that algorythmn. But it is based on knowing what actually is on the platform. So I wonder how an algorythmn would work in a network of dozens of individual "tiny platforms" (i.e. peertube instances or whatnot) and then generate recommendations based off that. Hm...
