01
A product with agents is not an organization
You should be able to: State the difference between an AAO and a product that uses agents.
A product with agents in it calls a model to do a task. An Agentic Autonomous Organization is structured like a company: it has standing roles, a human who is accountable, approval thresholds an agent cannot cross alone, and a record of what was done. The agents are the workforce; the organization is the governance around them.
The estate's spec package puts it bluntly: an AAO is "what makes an organization an Agentic Autonomous Organization rather than a product with agents in it." The test is not how good the agents are — it is whether the thing around them is an organization at all: who owns it, what each agent is scoped to, who answers when one is wrong.
This is why "AAO" is a governance claim, not a capability claim. Two companies can use the same models and only one is an AAO, because only one has declared its roles, named its accountable human, and set the thresholds — and can prove it to a stranger. The rest of this track is how you become the one that can.
Source: GDA Group — What is an AI Autonomous Organization?
02
The seven questions
You should be able to: List what an AAO must answer about an agent that holds credentials.
The estate reduces "is this a real AAO" to seven questions any organization has to answer about an agent holding credentials — and asks them as executable code, so the answer is a check, not an opinion. They are, in essence: what is this agent's standing responsibility; what may it do; whose authority does it act under; where must a human approve; who is the real, reachable human accountable; how is what it did recorded; and does the whole thing hold together as a governed organization rather than a list of bots.
The fifth question is the one people skip and the estate refuses to let you: a real, reachable human the organization is accountable to. Not a role, not a shared inbox — a person who can be asked why. An AAO with no accountable human is the exact thing the standard exists to reject.
You do not answer these in prose. You answer them in a manifest — the charter — that a validator reads. That is the move that turns "we take governance seriously" from a claim into a fact a machine can check, which the next course argues is the whole point.
Source: Flashy OS — the AAO spec
03
Autonomous, under human ownership
You should be able to: Explain how autonomy and human ownership coexist in an AAO.
"Autonomous" and "under human ownership" are not in tension — the second is what makes the first safe. Agents act on their own within a scope; humans own the organization, approve what crosses a threshold, and answer for the whole. Autonomy is bounded by governance, and the governance is declared up front rather than improvised when something goes wrong.
The estate's deepest rule runs through every AAO: agents suggest, humans consent. An agent may draft a decision, a connection, a spend — but anything that crosses a real boundary waits for a human at the accountable end to sign it, and there is no setting that auto-approves. That single rule is what lets an organization be mostly software and still be answerable.
So an AAO is not a company with the humans removed. It is a company where the humans do less producing and more owning — setting the standards, holding the thresholds, and answering for the output. The Working Inside an AI Organization and Directing Agents Well tracks teach that shift as a day job.
Source: Flashy Academy — Directing Agents Well
04
The DeAI operating system
You should be able to: Place the AAO in the wider stack it is the spec layer of.
An AAO is not built from scratch each time. The AAO manifest is the spec layer of a shared operating system: several organizations — a StartupOS, a PersonalOS, a FinancialOS — are built on the same format, so they compose instead of each reinventing identity, permissions, approvals and audit. "An OS is a manifest and a handful of domain agents. It is never a fork of another OS."
That is what makes the mesh possible. Because every AAO declares itself in the same format, a stranger — or another organization's agent — can read any of them the same way, and organizations can find each other and work together without a central registry. The manifest is the shared language; the mesh is what it enables.
The rest of this track is the practical path: what being readable requires (next), how to write the charter, the one command that scaffolds it, the Flashy spokes that give your AAO identity and value, and the full walkthrough from nothing to a node in the mesh.
Source: Flashy OS — the mesh platform
