Oddbean

Is anyone working on algorithms for feeds on here? It just seems so obvious that this is one of the two things that will determine whether nostr *as social media* (sure it can be other things) is going to achieve any scale ... but I never see anyone talking about developing this? The other is scaling, but that's a very vague thought, I have no idea what the issues or solutions around running a relay are (except, paying for service with LN is great).

This is the response I'm wondering about, i.e. is this perspective the reason we don't hear much about it? Algos can be user chosen so that argument has always seemed dumb. I used to use mastodon and the developer had a similar reason for not having algos. I think it's incoherent and stupid. Time ordering is an algorithm, just an extremely simple one. It's totally different in a FOSS scenario.

Agreed 100%. Nostr devs may be afraid of a negative feedback but I think if you query most users they would be in favor of it. And if I had a choice of selecting/building my algo based on a certain criteria I will spend more time on Nostr and think less of what other social networks are doing.

I prefer not having an algo. At least, not here on Amethyst. I enjoyed Twitter most when I could just see the real-time feed of the folks I followed. I suppose maybe having alternative feeds, that use an algo instead of going full global might be reasonable. But I would want that as an optional secondary feed. Otherwise, I just want my timeline.

People have an odd idea of what “the algo” is. You are using an algorithm right now, it’s simply that you prefer it in chronological order. But you don’t view your feed without a set of rules for ordering posts. There is always an “algo.” The question is whether you want it to prioritize people you tend to interact with more (your friends), or whether you want to lower the frequency of reposts you’ve already seen, or whether you want to order by engagement from your other follows, etc. I agree with OP, this is something that needs to be worked on and it’s odd that everyone has such baggage with something so fundamental merely because there’s a negative association with how Twitter did it. 🤷🏻‍♂️

Understood. Though the general idea is that someone else dictating what I see rather than what I want to see. These are needed in fiat social because it is how the make a profit. Feeding you content you don't want to see. I'd be open to personalized controls, where the algo is based on self-selected options. But still, the Twitter chronological feed was far better than they're algo feed.

100%, but that’s because the algorithm was designed to work against us, not for us. It isn’t about the filtering itself, it’s about the fact that the social platforms used it to trap us rather than find us good and relevant content.

To guys point… the difference is simply choice. As long as users are free to choose their algos on Nostr, across any client (which will require a NIP standard that Nostr does NOT currently have), Nostr users will always be in control of their content feed. Freedom of choice makes all the difference. https://yakihonne.com/article/naddr1qqgx2e3cxqenqef3x56kvd3nxymxzq3qmanlnflyzyjhgh970t8mmngrdytcp3jrmaa66u846ggg7t20cgqqxpqqqp65w7jk7e3

Yeah I wasn’t meaning have it centrally “imposed,” but rather have options for weighting certain posts or people. Or even an ongoing thing based on a simple upvote/downvote mechanism maybe? Regardless, I mean user controlled algorithms and experimenting with different options.

my 2 cents - NO algorithms (ever) - use block and use follow for users, keywords, etc - improve search - most importantly take responsibility for your own feed! the problem with mainstream Internet (or anything) is that it panders to people who want no-thinking easy mode, censorship and don't care who controls what they see. people like this uninformed, ignorant (they'd actually rather not know), and stupid. do I want to see that on nostr? NOPE!

Yes. Personal, opt-in recommendations from webs of connections you control and score. No global consensus, no trust you didn't award, no mandated agreement between parties who don't wish to voluntarily agree. Fully auditable, portable, subjective and sovereign.

Can you briefly describe 1 simple algorithm that isn't just notes ordered by creation date? Maybe if we start with that one concrete thing we can try to implement it and see how hard it is to do it in a nostric way. Every time someone brings up this topic they talk about it in generic abstract terms, so it's hard to make the first step.

here is one such algorithm https://github.com/pablof7z/dvm-references/blob/master/apps/dvms/src/dvms/you-might-have-missed.ts fetch events you signed fetch your follows compute activity windows (i.e. clustered time windows in which you were active) fetch reposts from your follows remove events where a repost occurred during a time in which you were active return the list of stuff your follows reposted while you were away and this is what the output can look like https://vendata.io/jobs/nevent1qqsr8693plxux3dufslf6t45qs5vpgk0wxq0dv4229n7yg6syg57zdqpp4mhxue69uhkummn9ekx7mqzyrafsj7hmweg9ur7zmn6apajdg48hxuskujx53rhrux0ttjcqx84yw53h7e

Thank you for the description, it has really cleared things up for me. Some of these DVM results are always impressive when I see or think about them, but I may be too dumb because I never grasp the entire thing in a concrete way.

I have been thinking one neat way for a DVM to expose a feed to a user, which they will be more likely to consume is to 'run the algo' and then broadcast the events that match into an algo specific relay. This would allow normal clients like amethyst to show a feed of events from this relay's global feed vs. users having to go to a webpage and/or modify their follow list. I think the algo based stuff would be far more useful on a per event basis than it is trying to lump a diverse human's post history into a category. Then the user can use their trusted client to add follows when they see an event they like and use nostr more 'naturally'.

for example: create a drag and drop interface that allows users to drag options/keywords/time/dates/blocked tags/whatever into some sort of order or priority including blocking that allows them to customize their feed(s). maybe even allow them to have multiple tabs with different feeds that they create.

I have a simple algorithm for the Hot feed in Nostur: - get your follows - fetch posts from the last X hours that they liked or reposted - sort posts by most liked/reposted, only count likes/reposts from your follows, don't count duplicates

What are your thoughts on scaling? According to @PABLOF7z's estimates an average Nostr client following 850 users would only have to connect to 56 relays in order to fetch posts from all of them. @cloud fodder can give us some estimates on how much it costs to host one average Nostr user in a relay on https://relay.tools/

There is a certain amount of sharing resources that seems to make running relays very cost efficient. For relay.tools there are 3 tiers of sharing. The server is shared by multiple relays, each relay is shared by multiple users. One popular pattern for sharing right now is someone runs a relay and adds their follow list (or any number of lists) to it's access control and/or enables payments to keep spam out. This is good for the outbox model because clients will be more likely to hit one of these relays where most of their friends notes are without going overboard on connections. Relays only get used if the relay operator promotes their use, otherwise they mostly just sit there idle (but usually do start receiving traffic from blastr, and scraped by nostr search engines when nostr watch discovers them). As nostr moves to more of an outbox model and/or nip42 auth it will become clearer the cost per user, as it is now it's fairly nebulous. That being said strfry is very efficient and can easily handle 10k simultaneous connections on a low tier VPS.