Rewriting Yakread

I've been rewriting the Yakread codebase over the past many months. The first part of that involved a lot of experimenting with different ways to structure the code. I came out of that phase a couple months ago and progress on actually rewriting the app (instead of rewriting one part of it over and over) has picked up quite a bit. Once it's complete, Yakread will be faster and better and easier to continue developing.

After that's done, I'm going to shut down The Sample and migrate the users (readers, not publishers) over to Yakread. The two products are similar enough for a migration to make sense—in both cases you're signing up for an email that recommends stuff to read—and it'll be great to have everyone on a single app that I'm actually developing. During the past year, The Sample’s revenues finally dipped below its operational costs, so its days have been numbered.

I have of course been thinking about what future directions I'd like to take Yakread. A small update is that I'm planning to restructure the For You feed: instead of having everything (subscriptions, bookmarks, ads, discover recommendations) mixed together, I'll probably split it into three sections:

1. A single ad at the beginning
2. with subscriptions and bookmarks in the middle
3. followed by some discover recommendations at the end.

Beyond that, there are two distinct larger things I'd like to work on. I'm not sure yet which to do first. They both will probably need ongoing refinement, so whichever one I start with, I'll likely need to stick with it for a while.

Thing #1 is to revamp the discover recommendations. Currently the recommendations come solely from other Yakread users (via collaborative filtering, though you could also think of them as “what's trending” on Yakread, where “trending” means “one or two people liked this article” since there aren't that many active users). It'd be cool if Yakread could crawl the web and find things relevant to your interests without relying on other users. It would raise the quality ceiling for the recommendations since the number of current users would no longer be a constraint. There are probably some interesting things that could be done with LLMs. This would also fit well with Yakread being an open-source app: those who want to self-host it could still get valuable recommendations.

Thing #2 is to help users create linkblogs. Back when I wrote this newsletter regularly I would sometimes go through my liked articles on Yakread and pick a few to share in the newsletter. It could be nifty to automate that process. Users could opt-in to have a public Yakread profile, and then their liked articles would show up there. Or maybe there'd be a separate share button which sends articles to your profile. Profiles would have an email signup form, and subscribers would get a weekly automatic email with your liked/shared articles—no need to compose or send the newsletter yourself; you just use Yakread like normal.

There could be some more doodads. You could pin a few articles to share each week, which could be nice for sharing your own articles that you've written. There could also be a “from the archives” section at the end with some randomly/algorithmically picked shared articles from previous weeks. Maybe you could write some optional commentary for the articles when you feel so inclined.

I think the biggest potential benefit of this is the connections you could make. I like the idea of Bluesky, but in practice my feed is mostly people talking about politics. No shade if you're into that! But I get my politics fix from a couple newsletters I subscribe to; I'd rather have my social media experience be more of a scenius-type thing. Could Yakread's focus on longer-form writing do the trick?