The Hussh Mega Map

TL;DR: The Hussh Mega Map is one diagram that shows the entire platform at a glance: a seven-layer stack of what each part is and why, plus ten end-to-end user journeys that each read as a single left-to-right sequence. Every story is its own lane, so the connections are fully traceable and never tangle.

Status as of 2026-06-10: see body.

Relations

• Three-layer architecture — the protocol, platform, and fund framing the map expands into detail.
• PCHP — the consent, export, and audit boundary that forms the Trust layer and Flow 7.
• One — the orchestrator that sits at the top of the Agents layer.
• Kai — the finance specialist and the AlphaAgents investing flow.
• Nav — the privacy and consent guardian among the agents.
• BYOA — Bring Your Own Agent, the developer-extension story the map calls Flow 5.

The Hussh Mega Map is a single picture of the entire platform. Most architecture diagrams force a choice between what the system is and how it actually works — the Mega Map shows both, in one view, without turning into a tangle of arrows.

It answers two questions at once:

• What is the platform? A clean seven-layer stack. Each layer holds its real components, and every component says what it is, why it exists, how it works, and gives a concrete example — so each claim is something you can question, not decoration.
• How does it connect? Ten end-to-end journeys, one per user story. Each journey is a single left-to-right sequence of real steps. Because every story has its own lane, the connections are fully traceable and no two lines ever cross. That is the whole trick behind "connections without spaghetti."
• Consent Reverse-Auction — the Intention Economy on PCHP + UCP + AP2 — the consent reverse-auction — flipping data value to the user

How to read it

1. Decode the words first. The map opens with a plain-English key for every acronym — personal knowledge model, consent handshake, bring-your-own-key, zero-knowledge, and the rest. Nobody should have to already know the vocabulary to follow the picture.
2. Walk the stack top to bottom. Each layer hands off to the one below it, and the short sentence between layers states that dependency in words: experience reaches people, the trust gate earns consent, services bring data in, the encrypted memory holds it, agents act on it, channels distribute the results, and the infrastructure carries it all.
3. Trace a journey left to right. Pick a story at the bottom and follow its steps. Each step is color-tagged to the layer it touches, which is what ties the two halves of the map together: the same colors you met in the stack reappear in the flows. Reading each component. Every box in the stack now shows two things at once: its plain role (a small tag — accounts connector, agent runtime, identity provider) and, beneath, what · why · how · e.g. so each claim is debatable, not decorative. Status is shape-coded for colorblind readers — ● shipped · ◆ approved · ○ future — so you never rely on colour alone.

The seven layers — what it is

1. Experience and interaction. Where a person or an AI meets Hussh: the web app, the native iOS and Android app, an on-device runtime on the Mac, and external AI hosts. The same product truth shows up on every surface.
2. Channels and ecosystem. The governed ways results and capabilities reach people, developers, and partners: a developer API, a hosted consent-tool server, a marketplace where advisers can share strategies, certification tiers that tell users which agents to trust, and partner integrations that receive only narrow, approved fields — never a copy of your private memory.
3. Intelligence and agents. The reasoning layer that acts inside scoped permission and never touches raw keys: an orchestrator that coordinates, a privacy guardian that enforces consent, a finance specialist, a multi-agent investing debate that produces a clear Buy / Hold / Reduce decision, and a developer kit so anyone can build new agents under the same rules.
4. Data, knowledge, and the personal memory. The heart of the system: one encrypted store that the user genuinely owns. The server only ever holds ciphertext. Personal data is organized into 24 domains across six human families — who you are, what you know, who you relate to, what you have, what you want, and how you delegate.
5. Trust, identity, and consent. The gate every action passes through: sign-in, a biometric vault unlock where the key lives only in memory, short-lived scoped permission tokens, a six-phase consent handshake, and an encrypted, audited release of exactly the fields you approved.
6. Core platform services. The backend that enforces the rules and brings chosen data in: the consent routes, an AI-memory importer, a receipts connector, a read-only accounts connector, and a public-profile research engine.
7. Infrastructure. The governed foundation underneath everything: compute, the database, the secret store, authentication and push, and a release pipeline with safety gates.

The ten journeys — how it connects

Each is one left-to-right sequence on the map, with an honest status.

1. Build your personal memory (shipped) — sign in, unlock the vault, the app encrypts a domain on your device, and only ciphertext is stored.
2. Import past AI conversations (future) — connect an AI provider, parse the export, map it into the personal-memory domains.
3. Claim your public profile (approved) — the research engine verifies and assembles a profile from public sources; you choose what to keep.
4. From debate to a trade (approved) — ask the finance specialist, a three-agent debate weighs the case, and you get a clear decision; placing the order is future state.
5. Build your own agent (approved) — use the kit, get certified, publish to the marketplace, and run under the same consent contract everyone else does.
6. Keep a partner system in sync (future) — a partner receives only the narrow fields you approve, with a consent receipt logged — never a mirror of your memory.
7. Consent through the protocol (shipped) — an outside service discovers what it can ask for, makes a scoped request, you approve with biometrics, and an encrypted, audited export is delivered.
8. Same truth on web and native (shipped) — one action resolves through either the web or native path to the same backend, so every surface agrees.
9. On-device edge (future) — the Mac runs the agent locally with on-device inference and developer tools, still acting under the same consent.
10. Advisers share strategies (approved) — an adviser publishes picks to chosen contacts through a consented relationship, and those investors see them at home.

Why it stays calm

The map earns its readability from two disciplines. The flows are true swimlanes, so a connection is always a single horizontal line that structurally cannot cross another. And the layout is balanced by arithmetic rather than by eye — the content is centered, every layer fills the width evenly, all ten lanes share the same step span, and each handoff marker sits at the exact middle of the gap between layers. The result reads as one coherent document instead of a wall of boxes.

Sources

• Three-layer architecture
• PCHP