Essay
Consent-Gated Discovery
The system can know a path exists without anyone learning who is on it. That inversion is the product.
Ask a superconnector for help and watch what they do: they think, they weigh, and then they say “let me make a call.” They do not hand you their phone book. The call is the point — each person in the chain decides, with context, whether this particular request deserves their credibility. Every networking tool ever built has flattened that judgment into a searchable directory, which is why superconnectors do not use them.
Magician keeps the judgment and automates everything around it. When the router finds a path from your intent to a capability, you see your own first hop — you already know them — and a veil over everything beyond: “someone within two trusted hops can satisfy this.” Each subsequent node is revealed only when the person whose edge it crosses says yes, with your intent and your reason in front of them.
The rule underneath has a precise form, inherited from the estate’s consent doctrine and enforced in code: a private edge is indistinguishable from a missing one. This has a consequence that surprises people — a *declined* introduction is also indistinguishable from a path that never existed. The requester sees “no longer available,” never “X declined,” because a visible refusal leaks exactly the relationship data the refusal was protecting. There is a test in the protocol suite whose only job is proving those two states render identically.
Consent-gated discovery is what makes institutional participation possible at all. A fund, a family office, a government office — none of them can expose a membership list, and all of them can answer “does anyone in your graph have this capability, and will they take a meeting?” one consent at a time. That is the Circles model, and it is the only shape of federation that survives contact with a compliance department.