21. Solving for Sybil: Four Families

For the first time in this story, the materials exist.

That sentence has to do a lot of work, because for twenty chapters the answer to "why doesn't anybody just fix this?" has been some version of the pieces aren't there yet. The shared banlist reached for portable reputation before the tools to build it existed; EigenTrust prescribed a cost-of-entry layer that nobody had built; the field spent two decades treating Douceur's proof as a tombstone instead of a specification. Every defense we have surveyed failed at the same join: it could see behavior, content, velocity, network shape — everything except who is actually behind the account, and whether that someone can be held to anything. The identity column was empty. You cannot ban an actor you cannot name.

The thing that has changed, and the reason this chapter can exist at all, is that the column is no longer empty. Over roughly the last thirty years — slowly at first, then in a rush over the last five — a set of standards, protocols, and services has matured into a family of approaches to the Sybil problem. The last chapter named the four families at altitude. This chapter walks each one at ground level: what it is, who is building it (inside Web3 and well outside it), what property it provides, where it fails, and — the part the vendors never get to — how it composes with the others.