Thread

I think you’re wrong. Pretty sure there is a way to replicate the TikTok algorithm on nostr.

Hmmm, I have some ideas, but the way you framed this makes me hesitate to share 🤣

People generally underestimate the use of the search function

I'm with you on this one. I mean, I'm very happy for folks to play with DVMs, algorithmic relays, etc. All great experiments. But I really don't want it imposed on me. (If any client starts enforcing algorithmic feeds, I'll stop using it, and if it becomes the dominant way of doing things on Nostr, I'm out.) I also don’t get why every problem on Nostr is dismissed as a "discoverability" problem that requires an algorithmic solution. Classic XY problem... Folks here can hate on the Fediverse content and architecture as much as they want (I myself dislike the fact that identity and content ownership are bound to specific instances. This is why I'm bullish on Nostr), but I’ve been making ActivityPub work with linear feeds, lists, and hashtags for ages. And yes, I can find basically all the content I want there: from niche tech stuff to content specifically for Brazilian expats living in small British/Irish towns. It honestly doesn't get much nichier than that. I'm doing this using a small personal instance that isn't even pulling a fraction of Mastodon’s content in a pool of active users much larger than Nostr. Still, the niche content gets to my feeds. What we're lacking here on Nostr is a culture of boosting/resharing, content lists, and hashtags, so that content "lives" longer and spreads without needing endorsement from a Nostr influencer account. That, and a good onboarding experience so that the combination of spam bots, manosphere/homophobic/anti-immigration/hardline MAGA content, or even generally well-intended but overly enthusiastic Maxis and Nostr devs (guilty) don’t drive users away in the first 10 minutes before they even get a chance to check out catstr, artstr, gaming, travelling, music, retrotech, or whatever else they enjoy.

I want discovery to feel like that feeling of walking past shops or cafes you've never been to before. Some places might look interesting, by the signage or the color of the building, maybe the name. Most of the places are okay, but some of them become a new favorite lunch spot or the place you always find the perfect gift, the place friends hang out and chat.

Nostr wont “win” by replicating old ideas. The pathway has to be through communities, relays, and interoperability that complement the user preference. Something new which is emerging is not like the old way of doing things. We are pioneers of time and space ✨

The internet is built on top of distributed independent servers with specific owners. Google managed to provide a de-facto "global feed" for the internet for a given keyword. Why shouldn't some nostr feed algorithm providers reach a similar status? I understand that the problem becomes harder when you take into account each user's social graph, but I don't think that this makes it unsolvable

Creio que o algoritmo deva ser tratado nos relays, não nos clientes.

Nostr is not a social media app! As stated, "Nostr is a simple, open protocol that enables global, decentralized, and censorship-resistant social media". Nostr must focus on the set of rules and specifications that determine how data is transmitted between peers, and nothing else. Decentralized means that anyone can do their own thing, in their own way, as long as they follow the protocol/specifications to communicate with each other while being tied to an identity. Relays must be used to receive and deliver data that is validated based on the protocol, and nothing else. The issues raised are software-related and have nothing to do with Nostr. All enhanced features are client-based and could be connected to the client/provider's own server for those features. There is already enough centralization for "the ones that came first". Their place in the sun is deserved, but still centralized. We don't need to include in the protocol what users or service providers require for their enhanced feature needs. Again, Nostr is NOT a social media APP! Are we going to have a protocol based on lobby requirements or a protocol that "enables global, decentralized, and censorship-resistant social media"? Please, folks, stop asking for things that don't belong in the protocol! Ask the client/service providers for those features instead. What's gonna be? Thanks, have a GM, GD and GN.

I hope the goal is not to be X or TikTok. I hope it’s something else entirely. You build your own lens on the world. Feels like discovery here won’t come from one giant algorithm but from many small ones, maybe even ones I make or choose. Still wrapping my head around it, but I think that’s the point. This place isn’t finished — it’s possibility.

The focus should be on building decentralized discovery methods that leverage Nostr’s strengths. Relays, user-curated lists, zaps as a signal, and trust-based networks could all play a role. The challenge isn’t replicating Bluesky or X, it’s inventing something better that fits Nostr’s unique architecture.

That's why I started building a Nostr client that doesn't need such algorithms. Are these recommendation feed algorithms even necessary?

I think all these recommendation feed algorithms are not needed at all. You're basically just giving the algorithm the right to shove whatever it wants into your head. I stopped using algorithm recommendations and social networks built around them a long time ago, preferring to choose the content I consume myself. By the way, this approach helps you get rid of your addiction to social networks and clear your brain of garbage, because now you don't have an endless feed of incoming information that your brain loves so much. I like the way it's done in Telegram. I choose exactly who to subscribe to and who to read. I only learn about other personalities on Telegram and their publications from those I'm subscribed to, or from other sources outside of Telegram. And that's why I started developing a unique Nostr client that doesn't require any recommendation feeds.

what the fuck is this myth of global feed and whether or not it exists. whether or not such thing exists, it would provide in such massive amount of data that no real user would ever download.

global feed is feed of whatever relays you are connected

I don't understand @fiatjaf's argument here, at least, I don't understand the *strength* of it ("NEVER") ? First, on "global feed", it's a bit like (a more extreme version of) "the mempool" - in a decentralized system there is no single synchronized global state, but .. there can be pockets of smaller synchronized state, at the relay level .. I guess even at the client level? If you have instant access to 10k or 100k messages your client can present them how it wants. I admit to not knowing much of anything about relays, so tell me if I'm wrong. If the answer involves "it's more expensive", remember it can be paid for; I already pay for relays, it's completely normal. Also, it has always been very obvious to me that the "social media" application of nostr would obviously benefit from algorithms for feeds. (though such a thing wouldn't magically make nostr suddenly popular).

From Doubt to Financial Success Initially, I was cautious about online investments, but a recommendation from a friend led me to Susan platform on WhatsApp, which sparked my curiosity. I decided to invest $5000, and to my surprise, I saw immediate returns! This success completely transformed my financial outlook. Today, I'm proud to be earning over $100,000. If you're ready to start your investment journey with confidence, reach out to ⚡ Susan ⚡ on WhatsApp or Email. Begin your path to financial growth today! Contact Susan: WhatsApp: +1 (472) 218-4301 Email:susandemirors@gmail.com

communities + natural language filtering and sorting of them. full agency. View quoted note →

@PDJ This has been my biggest desire from Nostr so I guess I've been thinking about it wrong

Soon enough, local, private, low-power AI will be a basic service available on smart phones. Maybe even on legacy phones due to optimizations. That AI can learn what the user prefers to see, and the users feed can be optimized from there. My primary complaint about nostr is missing important posts, ones that I don't see until I scroll back for almost an hour. If the AI used simple cues like which posts I boost, zap and like, it could dig up those posts I haven't seen yet from my existing follow list and present them first (like a ICYMI tab). This would be far better than croud-sourcing the feed from total likes of posts in my feed. This could be a good start at the "hands free" curated feed everyone seems to like. How do you get important posts from people you don't already follow? You don't need to because one of your follows likely will reference or repost it. That is how news works.