The current architecture of digital social life is centralized. A small number of platforms — controlled by a small number of companies, governed by their terms of service, shaped by their business models — mediate the majority of online friendship formation and maintenance. The friendships formed there carry the platform's imprint: its norms, its information architecture, its incentives. The data underlying those relationships belongs to the company, not the people in them. When platforms change or die, the relationship infrastructure they hosted may not survive.
Decentralized social protocols represent a different architectural premise. Instead of a single company running a platform that users access, decentralized systems distribute infrastructure across many servers (instances) operated by different people and organizations, with open protocols that allow interoperability between them. The most developed example is ActivityPub, the protocol underlying Mastodon and a growing ecosystem of applications including Pixelfed, PeerTube, and others — what is collectively called the Fediverse. Bluesky operates on a different decentralized protocol, AT Protocol, with a different set of architectural tradeoffs. Nostr is a further alternative. Each represents a different answer to the question of how digital social infrastructure should be governed.
For friendship, the implications are real but not yet fully realized. In decentralized systems, users can move their accounts between servers without losing their social graph — the people they follow, and who follow them, are not locked to a single corporate instance. Identity is more portable. Data can in principle be owned and managed by users or small communities rather than by corporations. The power to set moderation norms, community standards, and interaction rules shifts from a central corporate authority to the people running individual instances.
These changes matter because the architecture of social infrastructure shapes the relationships that form within it. A friendship that forms in an environment owned by the people using it — or by a small community that has control over its own norms — is structurally different from one that forms in an environment owned by a surveillance-advertising corporation. The terms are different. The incentives acting on the relationship are different. What the platform is extracting from the interaction is different.
The honest caveat is that decentralized social remains, as of now, a small fraction of total digital social life. The network effects that make large centralized platforms powerful are real: people go where other people are, and most people are still on centralized platforms. Mastodon has millions of users; Meta has billions. For decentralized social to shape friendship futures at scale, it would require either significant migration from centralized platforms or the gradual establishment of decentralized infrastructure as the substrate for new communities.
The significance of decentralized social for this manual is less about current adoption rates and more about what it demonstrates: that the architecture of digital social life is a choice, not a law of nature. The centralized platform model is not inevitable. Different architectures produce different social conditions. Making those conditions visible — understanding what centralized ownership does to the friendships that form within it — is the first step toward choosing differently.