Every minute, thousands of people post to Bluesky. Most of it disappears into the feed. This catches the patterns as they form.
Bluesky runs on the AT Protocol — an open standard for decentralized social networking. Every public post is a public record, streamed in real time through a WebSocket feed called the Jetstream.
The network generates thousands of posts per minute. Most of them pass through unnoticed — individual observations, reactions, half-formed thoughts, things said to no one in particular. Taken together, they have a shape: what communities are active, what the mood is, what the platform is paying attention to right now.
Live Data Stories pipes that stream through an AI model, which watches a rolling window of posts and writes a short narrative about what it sees — emerging topics, shared moods, unexpected communities, what slice of the internet is awake at this moment. It’s pattern recognition applied to a live social feed.
It updates every 45 seconds. Every story is generated fresh from real posts.
This is a personal side project — one entry in a series of live data narratives exploring what happens when you point an AI at a real-time data source and ask it to find the signal in the noise.