Simple "for you" algorithm at https://beta.iris.to/ . It first checks who have reacted to the same posts as you, and then shows posts that those people have reacted to.

It's building the feed using IndexedDB in a web worker. When you open the page, it loads some events from relays into IDB, but the more events you have there the better the results. It rebuilds the feed every minute, or you can hit refresh.

I find the feed more interesting than plain timeline, but it should also include recent posts by users whose posts you react to, and some randoms. Ordering by score instead of time would be interesting to try.

nostr:npub1dcl4zejwr8sg9h6jzl75fy4mj6g8gpdqkfczseca6lef0d5gvzxqvux5ey was experimenting with logistic regression algorithms already last year. That might produce better results. https://github.com/adamritter/light-nostr-client/blob/master/src/lib/logistic_regression.ts

Some of you don't like algorithms, but don't worry; the CEO of Nostr has promised to not make them mandatory.

Reply to this note

Please Login to reply.

Discussion

The dawn of the algorithms. 😎

Turn ON/OFF algos like with a switch, or like we change listr directly on the main feed will be cool.

Like to see experiments about algos, I don't see them bad in itself, they are a tool, and as any other tool, everything depends on how they are used, or not.

Not using something is freedom itself.. thanks to the NOSTR Ceo that didn't put a mandatory on them, and thanks to you to work on Nostr itself 😜🫂🍀💜🤙

I like algos, if I can control them

What do you think of something like what's described here, some sort of way to find lively discussion clusters?

nostr:note1nl2kr8088vzcmlmwmcmn0je0mz7s8vpmqrk7aaphcgvzxnjmat0qek9y7y

This is great, we need more algorithms that serve the user instead of ad revenue.

Nicely done. Giving still the unsupportes kinds though my algo is json, json, meme, json, json, pic so nvm 🤣. It's beautiful 👀

Hey nostr:npub1g53mukxnjkcmr94fhryzkqutdz2ukq4ks0gvy5af25rgmwsl4ngq43drvk where do you maintain issues for Iris? Is it on github, or kieran’s hosted git.v0l.io ?

A simpler algorithm would be to see what your friends and their friends are liking. Is this already included? If so, maybe their input can be ranked higher?