Bluesky Attie Site Builder Takes Vibe Coding Into the Social Graph

Bluesky has expanded Attie for the third time in 2026, turning its experimental AI app into a builder for websites, games, quizzes, and other social experiences. The Bluesky Attie site builder is not merely another tool that converts prompts into code. Bluesky wants every resulting project to arrive with identities, participants, and a path to distribution through its social network.

That distinction defines the company’s pitch. A conventional vibe-coding session often ends with a working prototype and an unanswered question: Who will use it? Attie connects each creation to the AT Protocol, or ATProto, the open networking system beneath Bluesky. Friends can sign in, vote, play, RSVP, or leave a message without creating an entirely separate social identity.

The expansion also creates a sharper test for Bluesky. Attie attracted a visible backlash when it debuted as an AI feed builder in March. Bluesky must now show that social vibe coding gives communities useful control without filling the network with disposable AI-generated pages.

The Bluesky Attie Site Builder Moves Beyond Custom Feeds

Attie now turns plain-language requests into small social sites with their own shareable addresses.

Bluesky announced the expansion on October 15 during TechCrunch Disrupt 2026. According to the company’s site builder announcement, Attie users can create pages such as games, guestbooks, surveys, and ask-me-anything sessions.

The product previously focused on custom feeds and reports about activity across Bluesky. Its new site mode broadens the output from something users consume inside a social client to an interactive experience they can share anywhere.

The Attie interface presents several starting points. These include polls, events, guestbooks, games, leaderboards, journals, landing pages, surveys, and dashboards. Users describe what they want in ordinary language, while Attie handles the site-building process.

Every project receives its own address. Friends can then sign in with a Bluesky account to participate. That login connection supplies a ready-made identity layer, reducing the need for each creator to manage accounts and authentication independently.

The initial examples show what Bluesky means by a small social experience. Flappy Butterfly is a casual game where players avoid bad takes and collect likes. Its leaderboard gives the project a reason to circulate among friends.

Another example, Poster Alignment Chart, asks users to respond to ten social scenarios. After receiving permission, it can examine their last 100 posts and compare their self-image with the persona suggested by those posts.

A third example is an Attie team AMA. Visitors can submit questions, including anonymous ones, while selected answers can be posted back to Bluesky. The project turns a familiar social format into a separate interactive page without severing its connection to the original network.

These examples matter because they demonstrate more than automatic page generation. Each one relies on participation, identity, or social context. A generic code generator can build a quiz interface, but it does not automatically provide an audience that understands the references behind the quiz.

Attie remains an early-access product, and its homepage still offers a waitlist to people without access. Bluesky has not presented the feature as a finished replacement for professional development tools. It is positioning Attie as a way for ordinary users to create modest, shareable experiences around existing relationships.

That scope also limits what the announcement proves. Bluesky has shown starter projects and described a creation flow. It has not published evidence about how reliably Attie handles complex applications, heavy traffic, abuse, or long-term maintenance.

The important change is therefore not unlimited software creation. It is the combination of prompt-built pages with portable social identities and an existing community.

Bluesky Social Vibe Coding Starts With Distribution

Bluesky is betting that distribution, not code generation, is the missing layer in consumer vibe coding.

AI coding products have made prototype creation more accessible. A user can describe an interface, request changes, and receive a functioning application without writing every line manually. That process is commonly called vibe coding, meaning software development directed largely through conversations with an AI model.

Yet producing an application and establishing a community around it are different jobs. A new site normally needs authentication, invitations, moderation, notifications, and a reason for people to return. Even a technically successful project can disappear if its creator cannot attract users.

Bluesky COO Rose Wang focused on that gap during the company’s Disrupt interview. People are already building with AI, she said, but distribution remains an unresolved problem. Attie’s advantage, in the company’s view, is access to Bluesky’s audience.

Bluesky had approximately 47 million registered users when the expansion was announced. That number represents registered accounts rather than an independently verified measure of active participation. Still, it gives an Attie creator a much larger starting network than an isolated personal project would usually receive.

This is why Bluesky describes the feature as social vibe coding. CEO Toni Schneider contrasted the familiar solo workflow with software built for a group. Attie is intended for experiences that become useful because friends, followers, or community members participate together.

A community organizer could ask members to vote on an event. An artist could run a weekly creative challenge. A small group could maintain a guestbook, leaderboard, or private bulletin board. A commentator could host questions outside the constraints of a standard reply thread.

The resulting page can circulate beyond Bluesky because it has a standalone address. However, the social functions still point back to an Atmosphere identity, meaning an identity that works within the broader ATProto network.

That architecture offers a potential advantage over mini-app systems controlled entirely by one social platform. ATProto is designed to separate identity, data, and applications. In principle, a person’s identity and social relationships are not permanently tied to a single client interface.

Bluesky has spent years presenting this openness as its central difference. Its earlier algorithmic choice work allowed developers to create feeds that users could select instead of accepting one mandatory ranking system.

Attie extends the same idea from choosing algorithms to generating experiences. Instead of asking developers to manually create every feed or community tool, Bluesky wants more users to shape those tools through natural-language instructions.

The strategic pressure falls on standalone AI builders that treat publication as the final step. Attie suggests that publication is only the midpoint. The harder problem is connecting a project to people who recognize one another and have a reason to interact.

Bluesky does not need Attie to outperform dedicated coding platforms across every technical dimension. It needs the social layer to make a smaller class of applications more useful than technically richer projects without participants.

How Bluesky Attie Works Across the AT Protocol

Attie’s real mechanism is a stack of reusable social infrastructure, not simply an AI-generated interface.

Each creator uses an ATProto identity, according to Bluesky. Attie can also suggest ideas based on information available in a user’s public Bluesky profile. That context could help a musician build a voting page for songs or help a community account create a quiz around recurring discussions.

The creator can bring an AI model or use an open-weight option. An open-weight model publishes the numerical parameters required to run or adapt it, although the surrounding training data and code may not be fully open. This choice gives Attie a different posture from tools bound to one proprietary model provider.

Bluesky has not disclosed enough implementation detail to compare model quality, generation speed, or operating costs. It has also not published a comprehensive explanation of which components Attie writes, hosts, or validates. Users should therefore treat broad claims about model flexibility as product statements rather than independently tested guarantees.

What is clearer is how the social pieces fit together. A standalone site needs a way to recognize visitors. Attie can use Bluesky-linked identities, letting people participate with an account they already possess.

A social app also needs data. Today, ATProto was designed largely around public records, including posts, follows, likes, and blocks. That openness allows other services to read and build upon network activity, but it complicates experiences that require private information.

Bluesky is addressing that limitation through Atproto Spaces, a protocol extension for non-public data. The company opened the Spaces alpha in August 2026 with software packages, sample code, and a hosted development environment.

Spaces can support data visible only to permitted participants. The examples include settings, private bookmarks, restricted forums, and subscriber-only publishing. Those capabilities would also matter for Attie projects intended for friends, teams, or invitation-only communities.

The current alpha comes with strong warnings. Bluesky says developers should not use it in production because breaking changes and database migrations remain possible. Attie’s long-term private experiences therefore depend partly on infrastructure that is still being tested.

That dependency explains why the first public examples lean toward small, low-stakes activities. A game score, public quiz, or guestbook is easier to launch safely than a confidential workspace or a service handling sensitive records.

The mechanism also changes what “no code” means. Attie can remove the need to write routine implementation code, but it cannot eliminate product decisions. Someone must still decide what information a page should request, who may see it, how contributions are moderated, and when old data should be removed.

Those choices become more consequential when generated experiences connect directly to a social graph. A flawed standalone demo might inconvenience a few testers. A flawed social app can expose profile information, encourage harassment, or spread rapidly through trusted connections.

Bluesky’s protocol architecture creates opportunities for interoperability, but Attie still needs understandable permission controls. Users should know when a project reads public posts, what data it stores, which model processes that data, and whether participation creates a public record.

The Poster Alignment Chart illustrates the issue. Reading 100 public posts can produce a playful result, yet the user must understand what is being analyzed and how that analysis is used. Public availability does not automatically make every form of automated interpretation socially comfortable.

Attie’s technical promise is therefore paired with a design obligation. If Bluesky wants ordinary users to create software, it must also translate permissions, data handling, and moderation into choices those users can understand.

Attie Must Overcome Bluesky’s Earlier AI Backlash

The largest barrier is not whether Attie can generate a page, but whether Bluesky users trust the company’s approach to AI.

Attie’s March debut produced an immediate negative reaction from part of the Bluesky community. Roughly 125,000 users blocked its account within days, according to open-source data cited in coverage of the early backlash.

At the time, Bluesky had around 43 million registered accounts. The block count represented a small share of registrations, but its speed and visibility made the response difficult to dismiss. Attie reportedly became the network’s second most-blocked account.

Some users objected to generative AI broadly. Others worried that Bluesky was following the same product direction as larger platforms they had left. Several wanted the company to focus first on unfinished social features and routine usability improvements.

Bluesky responded by distinguishing its approach from opaque, platform-controlled recommendation systems. Jay Graber, then Bluesky’s CEO and now its chief innovation officer, argued that AI should help people build their own algorithms and software instead of controlling attention for advertisers.

That principle fits Bluesky’s history, but principles do not resolve every implementation concern. A system can be user-directed while still creating low-quality material. It can rely on public data while still surprising the people whose posts are analyzed.

Attie’s site builder raises further moderation questions. Games and polls can become harassment mechanisms. Anonymous question pages can attract abuse. Generated landing pages can support impersonation, scams, or misleading claims.

Bluesky has not publicly detailed a complete moderation model for Attie sites. Important unanswered questions include how users report a generated page, whether Bluesky reviews prompts or outputs, and how enforcement works across independent ATProto services.

The separate attie.site addresses also complicate user expectations. People may assume that anything carrying an Attie address follows Bluesky’s normal content rules. Creators may instead view those pages as independent projects. Bluesky will need to make responsibility visible before disputes emerge.

Quality is another uncertainty. Vibe-coded software can appear polished while hiding weak accessibility, unstable logic, or insecure data handling. Ordinary creators may lack the experience needed to identify those problems before sharing a project widely.

Attie can reduce that risk through constrained templates, safe defaults, and clear permissions. However, heavy constraints would also limit the flexibility behind the product’s appeal. The company must balance creative range with predictable behavior.

There is also a network-level concern. Lowering creation costs increases the number of things people can publish, but it does not guarantee that those things deserve attention. Bluesky could replace a shortage of community software with an abundance of repetitive quizzes, abandoned pages, and automatically generated promotion.

That outcome would undermine the distribution advantage. A social graph is valuable because relationships provide context and trust. If followers learn that Attie links usually lead to disposable content, they will stop opening them.

The strongest evidence for the product will not be a polished demonstration. It will be repeat use by communities that continue maintaining their pages after the novelty fades.

Bluesky should also avoid presenting every objection as hostility toward technology. Some users joined the network because it offered more control over feeds, moderation, and identity. Their skepticism is directly relevant to whether Attie preserves those qualities.

The company’s challenge is to prove that social vibe coding expands user agency without turning every interaction into an AI experiment. That is a product test, a governance test, and a cultural test at the same time.

Attie Turns Bluesky’s Open-Network Strategy Into a Product Test

The site builder converts Bluesky’s protocol argument into something ordinary users can judge through daily use.

Open social protocols can sound abstract to people who do not build software. Portable identities, composable services, and interoperable records matter, but their value is not always obvious inside a familiar scrolling interface.

Attie gives Bluesky a concrete way to demonstrate that value. If a user can create a useful event page, share it with existing followers, and let those followers participate through portable accounts, the protocol becomes visible through an outcome.

This direction follows Bluesky’s stated priorities for 2026. In its product roadmap, the company said it wanted to improve core social features while making feeds more central and helping people discover applications across the Atmosphere.

The roadmap also acknowledged unfinished basics. Bluesky cited missing drafts, limited media handling, weak follow recommendations, and the need for better discovery. Those needs have not disappeared because Attie can now generate games.

That creates a resource tension. Experimental products can attract developers and define a platform’s future, but existing users still judge the service through posting, discovery, reliability, and moderation. Attie will struggle if the main network feels less urgent than its AI experiments.

Bluesky also faces slower growth and uneven engagement. Its registered user count rose from approximately 43 million during Attie’s March launch to 47 million in October. Registration totals alone do not show whether people post regularly, participate in communities, or return after trying the service.

Attie can help if it gives inactive users a new reason to gather. A recurring game, local event page, or community poll creates an interaction that a normal timeline may not support well. It can turn a loose follower network into a group organized around an activity.

The feature can also help ATProto developers prototype ideas. A creator might validate demand with an Attie page before investing in a more specialized application. Successful formats could later become independent services across the Atmosphere.

That pathway resembles Bluesky’s experience with custom feeds. The company supplied the infrastructure, while independent developers produced many of the feeds that gave the network its character. Attie attempts to widen that creator pool beyond people comfortable writing code.

However, AI generation changes the scale of experimentation. Custom feed development still required enough technical effort to discourage many casual or malicious projects. Prompt-based creation removes more of that friction.

Bluesky must therefore build discovery systems that reward sustained value instead of raw output. Popularity alone can favor novelty, outrage, or manipulation. Community endorsements, transparent permissions, and visible maintenance histories would provide stronger signals.

The company must also clarify how open an Attie creation really is. A page running on ATProto is not automatically portable in every practical sense. Users need to know whether they can export their project, move it to another host, change models, or continue operating if Attie closes.

Those details determine whether Attie strengthens an open network or simply places a proprietary builder on top of one. Bluesky’s philosophical commitment will matter less than the controls creators actually receive.

The opportunity remains significant. Social platforms traditionally decide which product surfaces exist, then release the same set to everyone. Attie offers a different model where communities can create narrower tools around their own habits.

If that works, Bluesky becomes more than a microblogging alternative. It becomes a distribution and identity layer for many small social applications. If it fails, Attie risks becoming a collection of AI demos orbiting a network still working on its core experience.

What Comes Next for the Bluesky Attie Site Builder

Three signals will show whether Attie is becoming social infrastructure or remaining an experiment.

The first signal is sustained community adoption. Initial creation counts would reveal curiosity, but Bluesky should be judged by returning participants, repeat creators, and projects that remain active for weeks.

A useful Attie page needs more than a launch-day burst. People must revisit a leaderboard, respond to another poll, attend an event, or contribute to an ongoing journal. Persistent use would support Bluesky’s claim that social distribution changes the value of vibe-coded software.

If most pages receive one round of traffic and then disappear, the evidence will point toward novelty rather than a durable product category. Bluesky has not yet published retention or participation metrics for the site builder.

The second signal is the progression of Atproto Spaces from alpha toward a stable release. Non-public data will determine whether Attie can move beyond public games and surveys into experiences for smaller, trusted groups.

A stable permission system would strengthen Bluesky’s vision of software built for friends and communities. Delays, security problems, or confusing access controls would constrain Attie to lower-stakes uses.

The third signal is the quality of Bluesky’s safety and governance controls. Users need simple reporting, visible permission requests, reliable identity information, and clear rules for anonymous or generated content.

Bluesky should explain how it handles malicious pages, impersonation, data retention, and model access. It should also show whether blocking an Attie creator, project, or account changes how that project can interact with a user.

Strong safeguards would answer part of the concern raised by Attie’s March reception. Weak controls would reinforce the argument that Bluesky expanded generation faster than it expanded accountability.

Competitor responses also deserve attention, but they are secondary to these internal tests. Standalone vibe-coding products can add community features, while established social platforms can introduce their own prompt-built mini apps. Bluesky’s defensible advantage depends on whether ATProto identity and distribution work better in practice.

For developers, Attie is worth watching as a low-friction entry point into open social software. For creators, it offers a way to test interactive formats without first assembling a full application stack. For community leaders, it promises tools tailored to a group rather than another universal feature imposed by a platform.

None of those benefits is automatic. Creators still need judgment, moderation plans, and a reason for people to return. Users still need control over how their identities and public activity enter an AI-generated experience.

The Bluesky Attie site builder has made the company’s wager unusually clear. Generating software is becoming easier, so relationships and trusted distribution become more valuable. Bluesky already has the beginnings of both, but it must prove that they survive contact with large-scale AI creation.

Anyone experimenting with prompt-driven products should document what worked, what failed, and which permissions users found confusing. A structured prompt library can help teams preserve those lessons instead of repeating the same tests.

The question for the next few months is not how many sites Attie can generate. It is whether Bluesky communities adopt a small number of them as places they genuinely want to return to.