tell me how an algorithm for discovering less popular notes should work. i am thinking about making something with llms
Discussion
I wouldn't really know. But for starters I think eliminating the bitfluencers entirely from this obviously, if it were for discovery of 'hidden gems' type of algo.
With that whole web of trust thing i assume you can get a algo working on the "outer ring" of that web
I also dont know what i am talking about
using web of trust negatively. makes sense.
How do you define bitfluencers? Never heard that term but it makes sense that it exists.
makes sense. what if it is customizable with #keywords? you talk to the llm and set your keywords and it gives you trendy notes about certain keywords as they are posted. or negative #keywords like bitcoin could be set to be less desirable. (i am pro bitcoin of course but having something customizable to user's needs is the goal).
Yes, i think most people here are pro bitcoin but some find it a bit much. Especially when it's not really content and just saying slogans over and over, not sure this is all that interesting is it? (I mean, I've even see people from bitcoin twitter have been put off by this on nostr lol, imagine). I think i see it as borderline zapfarming. Besides you will probably definitely see content of an interesting bitcoin post because it will be reposted like mad, there's not much chance of that going amiss. I would add all the slogans to mute, ideally. In a feed like this imo. Ideally I would also have some sort of detection to mute image posts of ai bitcoin art. (Just reeling off what I would ideally have lol, not saying you should do this or whether it's even possible.)
I think with major price moves there should be a special feed lol. Nostr is pretty much unusable at those times.
I think a lot of this will be solved with topic feeds? I don't know :)
Conversely there could be bitcoin feeds for people who want that tho.
Maybe even a bitcoin slogans feed too for people who specifically like saying and reading those :) ¯\_(ツ)_/¯
yes. an llm would favor longer texts which have knowledge. 🤔 it can detect the keywords so keyword preference or dislike can be added. my current tech won't see what is in the images. so short notes with images naturally filtered out.
Maybe it’s based on the person who posts rather than the engagement with the note initially? If they don’t post often, then it is shown to more people. Because they really have something to say. It would be nice if social media didn’t demand our full attention if we want to stay relevant.
I don’t know anything about algorithms work 😅 ignore me if this is nonsense
yes. i sometimes just unfllow frequent posters because i can't consume everything even though i know the person is super valuable to nostr and bitcoin.