Bluesky’s Attie Expansion Turns the AT Protocol Into an AI Research Layer

Bluesky is turning its AI assistant into a research tool for its entire network, not just its own app. On July 24, 2026, TechCrunch reported that Bluesky added a new feature called Quests to Attie, the AI assistant that already lets users build custom feeds without writing code. Quests lets people ask Attie open-ended questions about what is happening across the Atmosphere, Bluesky’s name for the wider set of apps built on the AT Protocol, the open standard that Bluesky itself runs on.

The feature is a modest beta rollout with a waitlist. The more interesting part is what it reveals: Bluesky is building a proprietary AI product on top of a public, federated data layer that other developers can, in principle, read from just as easily. That combination, an open protocol underneath a company-owned AI assistant, is going to keep showing up as more social and collaboration tools get built on decentralized standards, and it is worth understanding how the pieces fit together.

What Quests Adds to Attie

Attie launched in March as a separate Bluesky product that lets users design their own feed algorithms and custom feeds without programming knowledge, with an eventual goal of letting people “vibe-code” their own small apps on top of the network. Quests extends that into open-ended research: instead of only building a feed, a user can now ask Attie to summarize trending topics, or identify the most influential accounts in a given field or location, drawing on activity from across the Atmosphere rather than just the Bluesky app.

Bluesky chief innovation officer Jay Graber framed the feature as a defense against misinformation rather than a novelty. “Attie’s AI is a tool that helps you make sense of the internet, helping you fight back against the proliferation of misinformation and noise by giving you the tools to find the truth yourself,” Graber wrote, according to TechCrunch. “Our bet is that tools that help people make sense of the world will prove to be more useful than those that fill it with confusion.”

Graber, who stepped down as Bluesky CEO in March and was succeeded by Toni Schneider, the former founding CEO of Automattic, also said Attie is being redesigned with new accessibility features and a new logo. The new Attie experience, including Quests, is in beta, with users invited off a waitlist over the coming weeks, and Bluesky has described more changes as still to come.

Why This Only Works Because the Protocol Is Open

The detail that makes Quests worth a closer technical look is the scope TechCrunch described: Attie can research “not only Bluesky, but also any app that integrates with the underlying AT Protocol.” That is only possible because of how the AT Protocol (short for Authenticated Transfer Protocol, or atproto) is built. According to the protocol’s own documentation, atproto is explicitly designed as “a decentralized protocol for large-scale social web applications,” and the Bluesky app is only one client built on top of it.

Personal Data Servers, Relays, and the Firehose

In the AT Protocol’s architecture, each user’s data lives on a Personal Data Server, or PDS, which the documentation describes as “your home in the cloud”: it hosts your data, distributes it, manages your identity, and orchestrates requests to other services on your behalf. Relays then collect updates from many different PDSs into a single combined stream, which the documentation calls a firehose. App Views, the third core service, consume that firehose to provide aggregated application data for the entire network, including large-scale metrics like likes, reposts, and followers. Bluesky’s own app is one App View among potentially many, and account data moves between these services over standard HTTP and WebSocket connections rather than a proprietary, closed database.

That architecture is exactly why Attie’s Quests feature can look beyond Bluesky’s own app: the firehose it draws from already contains activity from every PDS on the network, whether or not the account posting that activity ever opens the Bluesky app itself. In principle, a user who posts exclusively through a third-party AT Protocol client could still show up in an Atmosphere-wide trend that Attie surfaces, since that activity flows through the same firehose.

The Same Firehose Is Open to Competitors, Too

The flip side of that openness is that Bluesky does not have exclusive access to the data Quests relies on. Because relays and the firehose are part of the open protocol rather than a proprietary Bluesky system, any developer with the technical ability to run or subscribe to a relay can, in principle, build a similar research assistant over the same activity stream. Bluesky’s advantage is not data exclusivity; it is being the reference implementation with the largest user base, the product design lead, and the distribution to put a feature like Quests in front of tens of millions of people first.

That is a meaningfully different competitive position than an AI feature built on top of a closed platform’s API, where a company can restrict, meter, or revoke third-party access at will. Anyone evaluating an AI assistant layered on an open protocol should ask what is actually proprietary about it: usually the answer is the interface and the model, not the underlying data.

The Business Tension Behind an Open AI Feature

Bluesky needs Quests, or something like it, to work commercially. TechCrunch reported that the network’s user growth has slowed sharply: after doubling to 40 million users in the year leading up to October 2025, Bluesky now reports around 45.6 million registered accounts, a far slower pace of growth. The company raised $100 million in a Series B round that closed in 2025, funding it is now using to support experiments like Attie, and it has already discussed the idea of eventually charging for the assistant or offering a premium subscription tier.

That puts Bluesky in a genuine bind. TechCrunch noted that a meaningful share of Bluesky’s own user base dislikes AI features on economic, political, and environmental grounds, the same community whose growth the company is trying to reaccelerate. Layering a monetizable AI product on top of an avowedly open, community-governed protocol means balancing two audiences that want different things: users who chose Bluesky partly because it is not run like a closed, ad-optimized platform, and investors who need to see a path to revenue from the $100 million already raised.

What This Means for Anyone Building on Open Protocols

Quests is a small beta feature today, but the pattern behind it applies well beyond Bluesky. Teams building on open, federated standards such as the AT Protocol, ActivityPub, or Matrix should assume that any AI feature built on public firehose-style data can eventually be replicated by someone else with access to the same stream. Durable advantage has to come from product execution, trust, and distribution rather than data access alone. Anyone evaluating a vendor’s AI assistant that sits on top of an open protocol should ask the same question directly: is this useful because of something proprietary, or because the vendor got there first?

There is also a moderation problem worth planning for before it becomes urgent. A tool that can identify trending topics and the most influential accounts across an entire federated network is just as useful for coordinating manipulation as it is for the misinformation-fighting use case Bluesky is pitching. Teams building similar research assistants over open social data should treat abuse cases, like coordinated inauthentic activity discovery, as a design requirement from the start, not a feature to bolt on after the first incident.