You know, I enjoyed this, it's nice to get some random, interesting stuff to browse on occasion.

I have been waiting for a project like this since last year when someone on Twitter mentioned it. I can finally doomscroll without my brain rotting.

Update, it does not run properly in Firefox on iOS. After it has loaded to 100%, the site refreshes.

Is there a technical reason that you can't click on the filled heart icon again to undo your like?

Is it just not implemented, or would it be too much effort to undo the effect on your weights from liking the article?

> It is made as a demonstration of how even a basic non-ML algorithm with no data from other users can quickly learn what you engage with to suggest you more similar content.

Yeah, it got really sticky real fast. From the random (?) selection it starts off where I couldn't recognize anything but popular TV shows, it immediately over-indexed on that content and I had to fight for my life to see anything else in the feed that I would recognize and consider a good algorithmic pick for my interests.

Which is brilliant, because Instagram has the same issue for me - absolute metric tons of garbage and whenever there is a gem in that landfill of a feed that I interact with positively, it's nothing but more of that on my feed for weeks until I grow sick of that given thing. In conclusion, Instagram could have used this 30 line algorithm and I'd have the same exact experience when using it.

Algorithmic feeds are obviously problematic for turning several generations into lobotomized zombies, but they are also just not very good at nuance, so it is not even a case of something that's bad for you but it just feels so good. It's just something that's bad, but is able to penetrate the very low defenses in human psychology for resisting addiction and short-term gratification and there is no incentive to improve them for the sake of the user as long as they work for the advertisers.

Did you write your own summary parser for this? I wrote one in the past and found the wiki markup quirky to deal with. The wiki dumps do provide summaries but they seem to suffer similar issues.

Built something similar for research papers: https://www.producthunt.com/products/soch

Very small suggestion: Can you make the entries actual links/anchor tags so that it is possible to copy link, middle-click to open in a new tab, and so on?

How does it actually work? Can you add an "about" page that goes into the algo? Or can you add more info on the readme on github? I'd love to learn more.

Would be cool / horrific to do this, but then have AI influencer bots that teach you the things in influencer style click bait style videos.

You can do this manually, obviously. The key is the starting point. The design of thermonuclear weapons is always a good place to begin.