Isn't that pretty much what all platforms these platforms do these days, with algorithmic feeds? What level of control do you plan to add on top of it? And more importantly - how do you plan to monetize a model that's pretty much dependent on dopamine hits, without making it _more_ addictive/a time-waste?

Well, I've been building algorithms which explain _why_ something was surfaced, not just being a black box. Everything is controllable from data sources, ranking/filtering weights and a feedback mechanism to tweak this.

As for monetisation, I've thought about this a lot - and I don't think the time wasting is needed to create a good product. Google Search certainly isn't fueled by dopamine hits, and from an ads standpoint, our users are saying exactly what they want to see every morning.

Feature request: When I find an article that the algorithm did Not suggest — but I like the article.

Then, it'd be nice to know how I could tweak the algorithm and its config values, so it would find such articles, thereafter.

And to dry-run the algorithm to verify that now it does indeed find that article.

(You don't have a website yet?)

Thanks for the suggestion. A Chrome Extension was my solution for this.

I have an MVP and real userbase which I spent time growing. Now I have a platform to experiment with the grander idea here.

That being said, I'm not willing to link to this yet because it's tied to my public persona and not yet my full time job.

Some platforms like Facebook and Twitter do, but HN and Reddit are (outwardly) based on votes, so you're seeing what the community thinks is important, not what you yourself do.

Mind you, there's a lot of people (myself included) who want to keep up to date on what others find interesting at the moment.

Reddit is arguably the original “algorithmic feed” - votes + the decaying algorithm make it way more “addictive” than if it was just a chronological series of posts, ordered by number of votes.

The popular/news sections of Reddit are also more heavily based on non-vote signals, like FB.

I am looking forward to more details on this.

I also think that allowing the user to track the changes made to this algorithm along with some metrics ( like how much time user spent on topics, time wasted etc) would provide a helpful feedback loop.