Post by @shom
Algorithm is math not magic
Wonderful example of how to surface information to an user while respecting their privacy, with transparency, and still providing a self-curated experience.
Edit: what happens if I turn a quote post into a blog? Also, did it spam everyone with a notification of an edit? (Sorry, but testing account gotta be testing)
@stefano
@fiend_unpleasant no AI features here. Good old deterministic computing.
It's a simple algo that will consider the people you follow. At the end of the page, there's a complete explanation:
Why am I seeing these posts?
Each post scores on signals this instance already holds, and the score halves as the post ages. Nothing is fetched from anywhere else, and nothing learns from what you open:
+2.5
per account you follow who boosted it
+1.0
× log of the favourites seen from here
+0.8
× log of the boosts seen from here
+0.6
× log of the replies seen from here
+1.2
you favourited, boosted or replied to this author in the last month
+1.0
carries a hashtag you follow
+0.8
continues a conversation you took part in
-1.0
the post is a reply
12h
the score halves for every this many hours since the post reached your feed
Your own posts, and anything you already favourited, bookmarked, boosted or replied to, are left out. At most two posts per author. Twenty entries, then the page ends: there is nothing more to scroll.
Your blocks, mutes and filters apply exactly as on every other timeline.
@stefano I'm doing weird things to find the edges. Here's one. If I take a quote post and then turn it into a blog two observations:
- The first line is treated as the blog title, which in this case was just one line and it got cut off. This makes sense for how blogging needs to work but an observation.
- The blog post retains no connection to the original post being quoted. While there are a lot of considerations around embedding any of that content, I do think it should post a link to the original post in the blog as a top level metadata type field. What do you think?