01
A fragment, and the merge
You should be able to: Describe how federated fragments become one graph without a central list.
Each organization publishes a directory/1 fragment: nodes for what it is, edges for its relationships, served from its own domain. Nobody maintains a master list. A merge reads every fragment and folds them into one graph, and a validator checks each fragment against the spec before it is trusted.
This is federation rather than a registry: the record of an organization lives where the organization serves it, and the graph is a view computed over all of them. The estate can go down and your fragment still describes you.
You enter the record the same way anyone reads it: you serve the file. The emitter writes the fragment from what you already declare, so the record and the charter cannot drift apart into two stories.
Source: Flashy OS — @flashyos/directory
02
Only the other side consents to crossing a boundary
You should be able to: Explain why a claim about a relationship needs the counterparty’s signature.
Every write to the record lands as proposed. A claim that stays inside your own organization can proceed on your say-so. A claim that crosses a boundary — an edge naming another organization — reaches its countersigned state only when a human at the other organization signs it.
This is the estate’s deepest rule applied to the record: agents suggest, humans consent. An agent may draft the edge, but nobody’s machine turns a proposal into a mutual fact, and there is no setting that auto-approves one.
The consequence for a reader: a countersigned edge is a claim both sides stand behind, while a proposal is one side talking. The public record shows the first and never the second.
Source: Flashy Academy — Running a Joint Initiative
03
The public projection: private is indistinguishable from missing
You should be able to: Explain why being unlisted and not existing look identical from outside.
The public projection of the record never queries proposals, private-tier rows, or superseded entries — they are absent from the query, not filtered out after it. What a stranger sees is what has been earned and consented to, and nothing else.
A private organization is deliberately indistinguishable from one that does not exist: existence is itself information an organization gets to withhold. Being in the record publicly is earned by publishing and being verified, and it lapses — in public — when that stops.
So the record is not a phone book of everyone who signed up. It is the set of claims that are both consented to and current, which is exactly the set a reader can rely on.
04
A settled edge is derived, not posted
You should be able to: Explain why the record of completed work cannot be written by the parties directly.
When two organizations complete a joint initiative, a settled edge appears in the record. Unlike an ordinary edge, nobody posts it: it is a projection derived from a sealed settlement, so it exists only because the work was sealed, not because someone claimed it.
That is why a settled edge legitimately skips the propose-and-countersign flow — it is not a claim a caller could make, it is a consequence of a settlement that already happened and is already verifiable. A settlement with nothing sealed against it produces no edge.
The lesson generalizes: the most trustworthy entries in any record are the ones no party could have written directly. When you see a derived edge, you are reading the shape of work that occurred, not the shape of what someone wanted you to believe.
Source: Flashy Network — the public face of the mesh
