Hmm. Best one I know is @Nusa. 🤔
Nusa, do you think a recommendation engine like that would be useful for publications? Like, click your ten favorite publications and we suggest ten more? That sort of thing?
i still think that this recommendation engine idea i have will be a killer feature for nostr user retention though... and it can be a service that apps call in the first week or so to get recommendations to populate a recommendation list that discreetly is available to users to click and open
i get it why my boss thinks this is a proper value add for social networks, when i think through my own onboarding experience
it's really not hard to build this out either, although it needs a hefty database that spiders the network and filters it for a compact representation of the data (replies, reposts, quotes, follows, mutes)
well, the idea is in the air, anyway
nostr:npub1syjmjy0dp62dhccq3g97fr87tngvpvzey08llyt6ul58m2zqpzps9wf6wl has been implementing my idea of a service that does reverse proxy for you to expose your personal relay to the network, enabling full decentralized outbox, maybe someone else will build this...
we really need some competent front end devs in this team
Discussion
of course nerds are always great recommenders lol, especially the collector type
ohhh nostr:npub1636uujeewag8zv8593lcvdrwlymgqre6uax4anuq3y5qehqey05sl8qpl4 is a front end dev? hmmmm i could think of a few small tasks that could be nice to have done
I’ve been thinking about this, yes. Have a couple of my own ideas percolating. I also have a demo selective crawler set up to pick from and have some plans to give people previews and selections, essentially saving them from endless test articles.
it's an interesting idea, like nostr:npub1l5sga6xg72phsz5422ykujprejwud075ggrr3z2hwyrfgr7eylqstegx9z has been talking about with improving upon web of trust to count engagement data as well, and then you can add on top full text indexes and machine learning text graph analysis
and then also stuff like noticing idle follows, or follows that a user never engages with
nostr:npub176p7sup477k5738qhxx0hk2n0cty2k5je5uvalzvkvwmw4tltmeqw7vgup has been doing some work in this direction also
I liked the idea of essentially running a spam filter backward. Instead of marking things 💩, you 🌟 things, and it adjusts your filter.
Kind of like nostr:npub1utx00neqgqln72j22kej3ux7803c2k986henvvha4thuwfkper4s7r50e8 algo, but with the addition of examples, to the set of rules.