My YOShInOn reader downloads somewhere between 3,000 to 30,000 items in a cycle [2] and chooses 300 top-scoring items out of 20 clusters. When I complete those, it runs another cycle. It has extra screens that show articles that it thinks would get >10 votes or a comment/vote ratio > 0.5 on HN as well as screens to show top-scoring articles from particular sites and feeds (arXiv, lobsters, ...)
Articles in the primary feed are shown to me one at a time, I thumbs up or I thumbs down. The RoC for the classifier is about 0.78, I read TikTok gets 0.84 so I'm pretty happy.
The problems with it: (1) It depends on arangodb for which the license doesn't allow me to commercialize it and I wouldn't feel OK with open sourcing it. Right now I'm writing a python-arango replacement which will get it and my image sorter running on postgres out of a single code base. (2) the batch organization doesn't work well for certain topics like sports where articles have a shelf life.
[1] doesn't have to be "creepy blond girls want to follow you" or all outrage all the time, an algorithmic feed can apply any heuristic that you like.
[2] depending on how fast I am reading, quality gets better when I am reading slow. The system blends in a certain percentage of randomly chosen results to maintain calibration -- I've been thinking about making it run at a target quality level where it blends in more randoms if it thinks it is showing me too many good results.
Newsblur offers something like this, but my feeds aren't a firehose, so I've never investigated.
It's a real simple algorithm. You control everything about these post organization rules and can update and change them as you wish. I thought it worked rather well back in my firehose reading days.
My favourite thing about my feed reader is that I can be sure I see every item from my subscriptions. Sure I don't need this for every subscription. In fact I may subscribe to a few more feeds if I did have an algorithmic filter that I could enable for a subset of my subscriptions. But I am quite happy without an algorithmic feed. In fact I think my current strategy of using sites like Hacker News for the "algorithmic" portion of my feed may be a better solution overall because it means that I am exposed to a wide range of things. Right now most of my feeds are self-curated where I appreciate the reliability of seeing every item. But I supplement with an algorithm for a wider perspective and discovery.
After all this isn’t about consuming all possible information. It’s about finding some stuff to read when you have downtime, that’s it. No need to overthink it.
That's an unfair statement to RSS users, hacker News trades and people generally.