2026-07-11 –, Open Social Space
"Decentralization" can imply particular network topologies while bundling together a set of related values like credible exit, data sovereignty, privacy, permissionless participation, and censorship resistance. This talk teases those values apart and makes the case for pursuing them directly, framed by what they enable rather than what they prevent.
People have been arguing about what "decentralization" means for years. It points at a bunch of values worth caring about (credible exit, data sovereignty, permissionless participation, censorship resistance) but it also smuggles in assumptions about network topology, often inherited from peer-to-peer work that don't always map onto what we're really trying to achieve. It's worth pulling those threads apart, naming what we actually want, and framing our work around what it opens up rather than what it guards against.
This talk is inspired by designing and building atproto (AT Protocol), which borrows primitives from the p2p world (content-addressed data, cryptographic identity, signed user repositories) and transposes them to the server so they can hold up at global social network scale. Along the way we've had to keep questioning our assumptions about what "decentralization" means, avoid prescriptive framings, and search for the properties that actually matter for the systems we're building.
Long time hacker in the dweb/p2p space. I now lead design/development of atproto at Bluesky, trying to apply what I learned to the open social web.
Daniel Holmgren is Head of Protocol at Bluesky, where he leads the design and development of the AT Protocol. He's worked in the DWeb space for eight years, previously founding a startup for p2p collaboration on scientific data, then serving as the founding engineer at Fission, where he built an in-browser framework for user-controlled data.