AT Protocol defines a JSON-based data network where posts, likes, follows, and profiles are typed, hyperlinked records anyone can own.
Key Takeaways
All social data (posts, likes, follows, profiles) is stored as strongly-typed JSON with shared Lexicon schemas, making records composable and portable.
Every record and user has a URL; content-IDs create strong cross-user links that survive account migrations.
bsky.storage automates periodic account data backups to a storage network and adds a UI for PLC identity backup and recovery – no self-hosted PDS required.
The architecture separates identity (DID/PLC) from hosting (PDS), so users can recover accounts without depending on any single provider.
Rich text facets and embeds are expressed as inline JSON structures, keeping the wire format inspectable and toolable without proprietary encoding.
Hacker News Comment Review
Commenters drew comparisons to pre-platform-era internet protocols, framing ATProto’s open data model as a structural return to decentralized norms rather than a incremental feature.
The bsky.land Etherpeg-style visualizer surfaced as a concrete example of what becomes possible when the firehose is open and typed – third-party tooling the mothership never planned.
One commenter noted that HTTP(S) is not the only transport for structured data, implicitly questioning whether ATProto’s JSON-over-HTTP choices are necessary constraints or just defaults.
Notable Comments
@danhon: points to bsky.land as an ATProto-native Etherpeg – a live packet-visualization callback to open-internet nostalgia.
@thesuitonym: “you can send information through the internet without HTTP(S)” – a pointed question about whether the protocol layer choices are deliberate or habitual.
