A good small community contains knowledge that nobody has written down. People know who is joking, who always overpromises, who can repair a boiler, whose silence means discomfort, and which old disagreement explains today’s careful wording.
Software cannot simply turn that knowledge into a database and retain the community. The knowledge works because it is situated, reciprocal and costly to fake.
This is why “scaling community” is often the wrong goal. Scale the ability of communities to cooperate. Do not scale one room until nobody recognises it.
Local trust is cheap for a reason
In a small group, consequences attach naturally. A bad contribution is seen by people who understand its context. Help creates memory. Disputes can be resolved by people who expect to meet again.
At large scale, those conditions weaken. Moderators cannot know participants. Universal rules replace judgement. Metrics stand in for memory. A central team becomes responsible for translating millions of local situations into a single enforcement language.
The usual response is more machinery: identity checks, reputation scores, automated moderation, global policy. Each layer can solve a problem, but it also moves authority toward whoever builds the machinery.
The federation move
Federation accepts that trust is local and translation is lossy.
Imagine many communities with their own membership, norms and memory. They cooperate through explicit bridges. A bridge does not announce that “Ana has a reputation of 0.87.” It says something narrower:
This open-source group has worked with Ana for a year and considers her reliable for code review. We do not know whether that assessment transfers to financial custody.
The limitation is part of the signal.
Elinor Ostrom’s work on durable commons showed that successful self-governance often relies on clear local boundaries, community-shaped rules, monitoring, graduated sanctions, inexpensive conflict resolution and nested layers. This is neither the fantasy of no governance nor the claim that one authority must govern everything.
A yoga class and a protocol
The principle is ordinary. A yoga teacher may trust one student to open the studio, another to help beginners, and a third to handle the mailing list. Nobody needs a universal rank. When the studio collaborates with a neighbourhood group, somebody translates between the two contexts.
A hacker project works similarly. Commit access, release authority and social standing are related but not identical. Another project may accept a signed release without importing the entire hierarchy.
Good federation makes these bridges visible and replaceable. Bad federation hides a global platform behind local branding.
The test is not how many users the system can place in one graph. It is whether groups can remain meaningfully different, cooperate anyway, and leave a captured bridge without losing themselves.