Bluesky’s CEO Jay Graber recently announced Attie, a new app that lets users build custom feeds using AI. My first thought? Finally, someone gets it. We’ve been screaming into the void about algorithmic transparency for years, and here’s a platform actually putting the tools in our hands.
As someone who builds bots for a living, this hits different. Attie isn’t just another “personalization” feature where you tick some boxes and hope for the best. It’s a proper feed-building toolkit that uses AI to help you craft exactly what you want to see. No more fighting with an opaque algorithm that thinks you need to see your uncle’s political rants because you accidentally liked one post three months ago.
What Attie Actually Does
Here’s the practical breakdown: Attie lets you create custom feeds on Bluesky using natural language. You describe what you want, and the AI helps build the filtering logic. Want a feed of just tech news from verified developers? Done. Need to filter out all posts containing certain keywords while boosting content from specific communities? You can do that too.
The beauty is in the execution. Instead of requiring users to learn complex query languages or boolean logic, Attie translates your intent into working feed algorithms. It’s like having a junior developer who specializes in content filtering, except it responds instantly and doesn’t need coffee breaks.
Why This Matters for Bot Builders
From a technical perspective, this is fascinating. Bluesky’s AT Protocol already supports custom feeds through Feed Generators, but they required coding knowledge. Attie democratizes that capability. Now anyone can experiment with algorithmic curation without touching a line of code.
For those of us building bots and automation tools, this opens interesting doors. The patterns users create with Attie reveal what people actually want from their social feeds. That’s valuable data for designing better content filtering systems, recommendation engines, and moderation tools.
I’m particularly interested in how Attie handles edge cases. Feed curation gets messy fast when you start combining multiple filters, time-based rules, and engagement metrics. The AI needs to understand context, handle contradictions, and produce something that actually works. If Bluesky pulled this off cleanly, they’ve solved some genuinely hard problems.
The Bigger Picture
This move fits into Bluesky’s broader 2026 roadmap, which includes improvements to their Discover feed and real-time features. But Attie stands out because it flips the script on how social platforms typically operate. Most platforms guard their algorithms like state secrets. Bluesky is saying “here, build your own.”
That philosophical difference matters. When users control their feeds, they’re not products being optimized for engagement metrics. They’re participants shaping their own experience. It’s a fundamentally different relationship between platform and user.
Compare this to how other platforms handle AI. We’re seeing tools like Talat focusing on privacy-first AI meeting notes, and Particle using AI to clip interesting podcast moments. These are all useful applications, but they’re still doing things to users or for users. Attie gives users the actual tools.
The Technical Reality Check
Let’s be honest about the challenges. Building good feeds is hard. Even with AI assistance, users will create feeds that don’t work as expected. The AI might misinterpret vague instructions. Feeds might become too narrow or too broad. There will be a learning curve.
But that’s okay. The point isn’t perfection out of the gate. It’s giving people agency and the ability to iterate. You build a feed, see what it produces, adjust your parameters, and try again. That feedback loop is how you learn what you actually want.
For bot builders and developers, Attie represents a template worth studying. It shows how to make complex technical capabilities accessible without dumbing them down. The AI acts as a translation layer, not a replacement for the underlying system.
What Comes Next
I’m watching to see how the community uses Attie. Will people share their custom feeds? Will popular feed configurations emerge? Could we see a marketplace of feed templates? The social dynamics around user-created algorithms could get interesting fast.
More importantly, this could pressure other platforms to open up their algorithms. If Bluesky demonstrates that users can handle algorithmic control without the platform collapsing into chaos, it undermines the justification for keeping everything locked down.
For now, I’m excited to get my hands on Attie and see what’s possible. Building bots taught me that the best tools are the ones that extend human capability rather than replace it. Attie looks like it might actually do that. And in a space full of AI features that feel like solutions searching for problems, that’s refreshing.
🕒 Published: