Sky Beyond — a Bluesky custom feed of "likes from the people you like"
Table of Contents
Sky Beyond is a custom feed for Bluesky.
The concept is simple: it shows you the posts liked by the people you like.
The people you like are the people whose taste is close to yours. If they like something, chances are it will land with you too — the classic idea behind collaborative filtering.
The key point: it never looks at the content of a post. No text analysis, no keyword matching, no AI classification. All it sees is "who liked whom" — the connections, the structure of the graph, nothing else. And that alone is enough to build a feed tuned to your taste.
The successor to the popular "Likes of Likes" feed
Sky Beyond is the successor to — and an upgraded version of — Likes of Likes, a feed I ran previously.
The original was fortunate enough to attract a lot of users, but it was built in a way where the more popular it got, the more computation it needed behind the scenes — operating costs skyrocketed and it couldn't be sustained.
Sky Beyond keeps the same concept but was redesigned from scratch. It only ingests events from its own subscribers, and feed contents are computed ahead of time — opening the feed just returns them. As a result, it delivers the same experience at a fraction of the original's cost, and faster. Since costs no longer explode with view counts no matter how many people use it, this time it's built to last.
Bluesky is the place where you can build your own feed in the first place
On X or Instagram, the timeline algorithm is whatever the platform decides — take it or leave it. Even if you wish posts were ranked by some logic of your own, there's simply no way to build it.
Bluesky is different. The underlying AT Protocol opens up the feed mechanism, so anyone can build and publish a feed with their own algorithm. If you're unhappy with the official Discover feed, you can switch to one someone else made — or make your own.
Sky Beyond is one of those. Recommendations are no longer something a platform imposes on you, but something you choose or create — to me, that's the most exciting thing about Bluesky.
What the experience is like
You see beyond your follows. Posts from people you follow already show up in your home timeline, so Sky Beyond deliberately shows only posts from people you don't follow yet
Posts liked by several people whose taste matches yours rise to the top. Not just whatever's going viral, but posts that are well-received in your corner of the network
The ordering is plain chronological. It reads like a natural timeline, without that "the algorithm is making me read this" feeling
The more you like, the more the feed adapts to you. It also supports the official app's "Show more / Show less" feedback
In short: a discovery feed made just for you, seeded from your own like history.
How it works (roughly)
Behind the scenes, it picks up subscribers' likes in real time from Bluesky's event stream and learns an affinity score — "how much do you tend to like this person?" Candidate posts weighted by that affinity are prepared in advance, so when you open the feed it just sorts and returns them — which is why it's light and fast.
To repeat: at no point in this process does it read the content of any post. All it handles are the like edges connecting people to people and people to posts. Zero content understanding, pure graph structure — and yet your "taste" still emerges. That's what makes collaborative filtering fun.
Both affinities and candidates decay over time, so the feed naturally follows your shifting interests.
Try it out!
Just open the feed on Bluesky and like posts as you normally would. The more you use it, the better it gets. If you enjoy it, pin it — and while you're at it, a Like on the feed itself would be much appreciated.
👉 https://bsky.app/profile/so-asano.com/feed/likes-of-likes-2
Developer of: - https://bluecast.app - https://bluefeed.app - https://skymeetsblue.com - https://skybemoreblue.com - Sky Beyond Feed: https://bsky.app/profile/so-asano.com/feed/likes-of-likes-2 He/Him. Contact: so.asano@gmail.com soasano (Discord)
No comments yet