Unpopular opinion: I want algorithms!

So on Xitter, I "like" stuff, not only to give the other guy a cue but also to train the algorithm to give me more of this. But maybe that is the wrong approach, as the feed might get saturated easily with stuff I "like" while never seeing stuff I might *also* like but never got to see because of the deluge of likable stuff.

On Nostr, clearly we want to be in control of the algorithm but so far, that is very rudimentary. It's only these options:

* Posts from follows

* Posts and replies from follows

* Stuff that does not take me into account (a relay's feed, trending posts, popular people, ...)

Ultimately I want some LLM to read the fire hose and show me what I most likely find relevant but without some Sam Altman learning my preferences or pushing his agenda on me.

Until we figure that out, what can we do with likes and replies? How do we find for each user the posts that are pushing the comfort zone but aren't too far out there? #coffeechain and GM is all good but nobody wants exclusively those.