1. Headline & intro
Social networks are quietly entering a new phase: not just competing on feeds, but on who gets to design them. Bluesky’s new AI assistant Attie looks like a niche tool for power users who enjoy tinkering with custom feeds. In reality, it is Bluesky’s most direct attempt yet to turn "the algorithm" from a black box into user‑controlled infrastructure.
In this piece we’ll look at what Attie actually does, why it matters strategically for Bluesky and its protocol (atproto), how it fits into the AI+social trend, what the European angle is, and what this experiment tells us about the future of recommender systems.
2. The news in brief
According to reporting from TechCrunch, Bluesky has introduced Attie, a standalone AI assistant for building custom social feeds and, eventually, simple social apps. The app was unveiled at Bluesky’s Atmosphere conference by former CEO and now chief innovation officer Jay Graber together with CTO Paul Frazee.
Attie runs on top of Anthropic’s Claude models and is tightly integrated with atproto, the open protocol that powers Bluesky and other compatible apps. Users log in with their existing "Atmosphere" identity (any atproto app account), allowing Attie to use their public social data to propose posts they might like, suggest content to repost and generate fully custom feeds via natural‑language instructions.
The product launches as a private beta for conference attendees. Over time, Bluesky plans to let people “vibe‑code” small social applications and tools through Attie, again via conversational prompts rather than traditional programming. The company recently disclosed a $100 million funding round, giving it over three years of runway and supporting continued work on privacy features and monetization for its reported 43.4 million users.
3. Why this matters
On the surface, Attie is “just another AI assistant.” In practice, it attacks one of the most defended moats in consumer tech: the ranking algorithms of major social platforms.
Today, TikTok, Meta, X and others treat their feeds as proprietary capital. Users may tweak a few switches, but the logic that decides what you see is deeply opaque and optimised for metrics like watch time and ad revenue. According to TechCrunch’s coverage, Bluesky’s leadership explicitly positions Attie as the opposite: an assistant that lets users and developers shape ranking logic themselves, on top of an open protocol.
The immediate winners are:
- Power users and curators who already build custom Bluesky feeds but struggle with code or complex configuration.
- Indie developers and researchers, who get a higher‑level way to prototype new recommendation ideas directly on live social data.
The potential losers:
- Incumbent platforms, if users come to expect a level of algorithmic control that closed systems cannot or will not provide.
- Bluesky itself, paradoxically, if the default Bluesky app becomes “just another client” in a broader atproto+Attie ecosystem where value shifts to third‑party apps.
Attie also reframes AI in social media. Instead of another engagement‑maximising recommender, it is pitched as infrastructure for personal and community algorithms. That’s a philosophical move, but also a business one: if Bluesky can own the “WordPress moment” for social feeds—hosting a sprawling ecosystem of custom recommender plugins and small apps—it gains a very different kind of leverage than traditional ads.
4. The bigger picture
Attie sits at the intersection of several trends that have been building for years.
First, there’s the decentralised social wave: ActivityPub (Mastodon), Nostr, Farcaster and now atproto/Atmosphere. All promise user portability and app choice. But until now, algorithmic control has lagged behind content and identity portability. You could move your account, but not your feed logic. Attie is one of the first serious attempts to make recommendation itself portable and programmable.
Second, we’re entering the era of agentic AI. Instead of chatbots that just answer questions, we are seeing tools that act on our behalf: auto‑reply in Gmail, copilot features in code editors, AI that manages calendars or runs web searches. Attie extends that idea into social: an agent that reads your network, curates information and, later, may even spin up micro‑apps.
Compare this with Meta’s AI characters inside Instagram or TikTok’s increasingly algorithm‑driven experience. Those offerings keep users inside a single walled garden. Bluesky’s bet is subtly different: the AI runs on top of a protocol, not a single app. That aligns more with how WordPress created a broad plugin/theme economy on top of an open standard, which Bluesky’s interim CEO has explicitly referenced.
Historically, every time users gained new control over information flows—think RSS in the 2000s or Twitter lists in the early 2010s—we got a burst of experimentation, followed by consolidation as a few defaults dominated. The open question is whether AI lowers the barrier enough for custom feeds to stay diverse this time, or whether “Attie‑made default feeds” simply become the new monoculture.
5. The European / regional angle
For European users and regulators, Attie collides with several ongoing debates.
Under the Digital Services Act (DSA), very large platforms must offer non‑profiling based recommender options and greater transparency around ranking. While Bluesky is not yet at the scale of Meta or TikTok, building user‑configured feeds from day one positions the ecosystem ahead of where Brussels wants the industry to go anyway.
At the same time, Attie’s ability to read a user’s cross‑app social activity on atproto raises GDPR questions: data minimisation, purpose limitation and meaningful consent. The technical architecture may be open, but “open” is not automatically “compliant.” Before a serious EU push, Bluesky will need fine‑grained controls over what Attie can access, clear legal bases for algorithmic profiling and explanations that normal people can understand.
For European developers, the opportunity is obvious: Attie could become a low‑friction way to build domain‑specific feeds—for instance, a German‑language climate policy feed, a Spanish‑speaking startup radar or a regional news filter that respects local media ecosystems. Compared to fighting for visibility inside Meta’s opaque algorithms, building on an open protocol with AI assistance is appealing.
But there is also a cultural factor. European users, particularly in DACH and Nordics, are more privacy‑sensitive and sceptical of AI than average U.S. users. If Attie is framed merely as another “AI that reads everything about you,” adoption will stall. If instead it is presented as a tool that restores agency over feeds—and ships with strong privacy defaults—it could resonate strongly.
6. Looking ahead
Several things are worth watching over the next 12–24 months.
Who actually uses Attie? If it stays confined to conference attendees and hardcore Bluesky fans, it will be an interesting demo, not a platform shift. The key test is whether less technical creators and community managers adopt it to run their own feeds.
Governance of user‑made algorithms. What happens when someone uses Attie to build a harassment feed, a doxxing tool or extremist content recommender? On a decentralised protocol, enforcement is messy. Bluesky will need clear policies on what kinds of feeds Attie will help create, and client apps will need moderation layers on top.
Monetisation models. TechCrunch notes that Bluesky is exploring subscriptions and hosting, and hasn’t decided whether Attie itself will be paid. A plausible path is a WordPress‑style stack: core protocol and basic Attie free, with paid tiers for advanced tools, analytics, or managed hosting for large communities.
Regulatory scrutiny. As user‑configurable recommenders spread, lawmakers will start asking who is accountable for harmful outcomes: the user who designed the feed, the AI assistant that helped, or the protocol operators. Expect this to surface in DSA implementation debates and future guidance on AI‑driven personalisation in the EU and UK.
If Bluesky executes well, Attie could quietly become the default interface through which people think about social algorithms—less as magic boxes, more as editable artifacts. That shift in mental model may be more important than any single feature.
7. The bottom line
Attie is less about adding AI to Bluesky and more about putting the most powerful part of social platforms—the feed—into users’ hands. If it works, Bluesky won’t just be another Twitter alternative; it could become the WordPress of social recommender systems. The big question is whether users and regulators are ready for a world where anyone can spin up their own algorithm as easily as a blog—and who will take responsibility when those algorithms go wrong.
