ATProto’s Private Turn Could Make It More Than a Social Protocol

ATProto began with a legible promise: give social applications a shared protocol while letting users retain more control over identity and data. Its next experiment is considerably more ambitious. By adding a way to organize non-public information, the technology may be inching toward something closer to a federated cloud substrate—one where applications, people, and perhaps AI agents can collaborate without placing every record on the open web. From public posting to permissioned computing In an [alpha announcement from the ATProto team](https://atproto.com/blog/atproto-spaces-alpha), Spaces is presented as an extension intended to support data that should not be universally visible. That distinction sounds modest, but it changes the protocol’s design problem. Public social networking is comparatively simple at the access layer. A post is published, indexed, cached, and redistributed. Moderation remains difficult, but the basic audience is clear: anyone who can reach the network. Private groups, shared workspaces, family archives, draft collections, and collaborative applications operate differently. They need rules describing who may read, write, invite, export, or delete. T