Straight answers
Frequently asked questions
If an answer here ever stops being true, the page is wrong and gets fixed — not the other way around.
Where does my relationship data live?
In a local store you hold — files on your machine, git-guarded so they cannot be committed by accident. Nothing syncs anywhere by default. “Nothing leaves without consent” is the architecture, not a settings toggle.
Do I upload my contacts?
No, and this is doctrine rather than a missing feature: “upload your contacts so someone can mine them” is the product shape that dies, and deserves to. You record trust edges — which are yours — and the router traverses them under your control.
How is this different from LinkedIn?
LinkedIn is a public directory with a feed; its unit is attention. Magician has no feed, no followers, no posting. Its unit is the consented introduction and the sealed outcome. LinkedIn tells you Michael knows Sarah; Magician knows what that connection is made of, whether it is current, and what it has produced — and shows none of it to anyone without a yes.
What happens when someone declines to make an introduction?
The path becomes “no longer available” to the requester — indistinguishable from a path that never existed. Who declined, and that anyone declined, is never shown. A visible refusal would leak exactly the relationship the refusal was protecting.
Can I see who is in someone else’s graph?
No. You can learn that a path exists through your own consented hops, one reveal at a time. A private edge is indistinguishable from a missing one — existence is itself relationship data.
How does scoring work? Can I pay or post my way up?
There is no single score, and nothing you can do alone moves your standing. Per-domain authority comes from what others assert about you; the one measured figure is sealed outcomes your edges helped route. There is a test in the engine proving unilateral activity moves nothing.
What is a Circle?
A sovereign graph with an owner — a person, a fund, a family office, a university. Circles federate by exposing permissioned capabilities to each other under scoped, revocable grants, without ever exposing membership.
Is the demo showing real people?
No. Every demo persona is fictional and marked as such in the data itself. A demo graph that could be mistaken for a real one would be a fabricated claim about real people’s relationships — the worst thing a trust product could ship.
When does the protocol open?
trust/1 and introduction/1 publish under Apache-2.0 when the first Circle we do not control adopts them — and not before. Captive adoption is evidence about a relationship, not a format. The roadmap page states this gate; the gate is the promise.
What does Magician cost?
Individuals: free or nearly so — your own graph must be valuable solo, on day one, or the product does not deserve a network. Institutions pay for graph intelligence, succession, federation and outcome analytics. Pricing is published when the pilot gate opens; unverifiable “contact sales” theatre is not the plan.
How does Magician fit the Flashy estate?
It is the human-trust layer of the same thesis. FlashyOS answers “which organisation can do this”; Magician answers “which trusted human path unlocks it”; Flashy Network seals and measures the record. One doctrine — consent enforced in code, outcomes sealed, no paddable numbers — at two layers.
What is an AAO, and is Magician one?
An Agentic Autonomous Organization: an org with a published charter — named roles, a human accountable by email, approval thresholds agents cannot cross. Magician’s charter is served at /.well-known/flashyos-charter.json and validated by the estate’s own conformance tooling in this repository’s test suite.
Will agents ever make introductions automatically?
Agents will do almost everything else — detect latent fits, prepare cases, negotiate calendars. The introduction itself always waits on a named human’s yes. That rule is inherited from the estate’s charters and is not staged to be relaxed; it is why the system deserves graphs at all.
How do I join the pilot?
The join page lists the current rungs. Stage 0 runs on the founding graph; Stage 1 opens to a small set of operators whose intents are real; Circles come after the loop proves itself. The honest state of each stage is on the roadmap page.
Is Magician open source?
The repository is licensed Apache-2.0, and the protocol formats publish openly at the adoption gate. What is never open is anyone’s graph: the code is shareable precisely because the data model makes the data unshareable by default.
How is Magician different from AI superconnector services?
The category — AI-mediated warm introductions — is real and being validated in the market, which is good news for everyone in it. The centralized model asks every person and partner network to join one global matching pool that gets smarter as it absorbs more relationships. Magician makes the opposite promise: your graph remains yours. Sovereign graphs — a fund’s, a university’s, a government ecosystem’s — cooperate through permissioned discovery (beacons and per-hop consent) without any of them uploading their relationship intelligence anywhere. One model optimizes for the network it owns; this one is infrastructure for networks their owners keep.
Can two organizations discover a trusted path without sharing their graphs?
Yes, and this is the product’s hardest acceptance test, kept as a named test in the engine’s suite: one circle publishes a beacon/1 — per-capability reachability with hop bounds, trust bands and registers, no identities, no membership — and another probes it entirely locally against its own edges to that circle’s gateway. The answer is “a trusted route exists — request an introduction.” Nobody’s graph moved; the beacon’s parser refuses a document carrying so much as one person id.
Does Magician train AI models on my relationship data?
No — by default and by design. The store is local-first with no telemetry, no sync, and no export path into any model pipeline. Any future opt-in would be per-graph, explicit, revocable, and logged as a consent event. A trust product that quietly trains on who trusts whom has misunderstood its own dataset.
Will Flashy Gold rewards be part of trust scoring?
Never as an input. “Make introductions → earn rewards” manufactures activity, which is exactly the behavior Magician exists to refuse — the product rewards quality outcomes, and standing moves only on sealed non-none outcomes. Any future community incentive sits strictly after an outcome, outside the reputation algorithm, bounded and reversible; the standing module’s invariant test is the tripwire that keeps it out.
Why should institutions trust the sealed records?
Because they verify offline. An introduction/1 record hashes over canonical JSON; anyone can recompute the digest with no server, no account and no trust in us. The same sealing discipline runs in production across the estate — from settlement records at flashyos.com to the public ledger verdict at claimyour.gold.
What is the concierge in the corner of this site?
The Mesh Concierge — the estate’s standard door, dogfooded here. Type what you need; it drafts a filing to a named human at this property through your own mail client. At layer zero it makes no network calls, keeps nothing, and tracks nothing — a test greps the served bundle for all three promises. Any property or partner can adopt the same door in an afternoon: the /concierge page documents the Activation.