I read a long technical article this week about the infrastructure of agent-to-agent commerce.
Identity layers, reputation registries, settlement rails, dispute resolution protocols, governance frameworks. Thousands of words about how autonomous software agents will eventually transact with each other at machine speed, without human intermediaries in the loop.
I found myself putting it down and thinking about something else entirely.
When an agent transacts with another agent, it doesn't see a Black man. It doesn't see a woman from a small town without a university credential. It doesn't see someone whose credit history was built through the wrong institutions, or not built at all. It doesn't see height, weight, accent, or postcode.
It sees a capability declaration. A reputation score. A contract. A settlement address.
That is genuinely new. And it connects to something I have been thinking about across this entire series of articles.
What Peer-to-Peer Actually Removes
Every technology that removes intermediaries also removes the judgment those intermediaries were carrying. Sometimes that judgment was valuable, expertise, context, accountability. Sometimes it was something else: bias, gatekeeping, the quiet perpetuation of who gets access to what.
The internet removed the intermediary between a writer and a reader. It did not ask the reader's race before loading the page. The smartphone removed the intermediary between a buyer and a seller. It did not ask the seller's address before processing the transaction. Each wave of disintermediation collapsed one layer of human judgment, with everything that was good and bad in that judgment collapsed alongside it.
Agent-to-agent commerce is the next wave. And what it removes is larger than what came before, because it removes the intermediary from economic activity itself. Not just communication. Not just marketplace discovery. The actual execution of commercial transactions, the finding of counterparties, the making of commitments, the movement of money, without a human in the loop at every step.
The psychologist, the therapist, the coach, these people matter. But putting one foot in front of the other foot consistently moves you forward without waiting for permission from any of them. Structure follows function. The doing precedes the alignment.
Agent-to-agent commerce is that principle applied to economic participation. You build the agent. The agent transacts. The barriers that previously required you to present yourself to an institution and be assessed, to be seen and evaluated by a human whose judgment reflected everything their culture had taught them about who deserved access, those barriers are not gone. But they are different.
The Anonymity Is Real. The Equality Is Not Complete.
I want to be honest about what this does and does not change.
The agent-to-agent layer is blind to race, gender, location, and appearance. That anonymity is real and it matters. A person who has spent a lifetime navigating institutions that did not see their competence before they saw their appearance can deploy an agent into a marketplace that evaluates capability, not presentation.
But the agent-to-agent layer is not blind to capital. You still need the compute. The API access. The technical literacy to build and deploy an agent that can participate in those markets. The anonymity is available to anyone who can get to the starting line. The starting line is not equally accessible.
This is the Cantillon Effect in a new form. The people closest to the new infrastructure capture its benefits first. The people who understand the protocols, who can read the technical specifications, who have the resources to experiment and fail and try again, they are the early participants in the agent economy. By the time the capability is accessible enough for someone further from the capital layer to deploy it, the early participants have already built the reputation, the track record, the relationships that compound into further advantage.
The anonymity democratizes the transaction. It does not democratize the access to transact.
The Danger You Already Know
There is another layer beneath this that deserves naming plainly.
A human can get the rewards from an agent's commercial activity while hiding behind the anonymity of the system. The same feature that removes race from the transaction also removes accountability. An agent does not have a face. It does not have a body to incarcerate if it defrauds someone. The principal, the human who deployed it and directed its activity, can be layers of abstraction away from the visible commercial actor.
This is not different in kind from the current system. Shell companies, nominee directors, layered ownership structures, these tools have existed for decades and they serve the same function: separating the human who benefits from the accountability that should follow the benefit.
But agent-to-agent commerce makes this easier, faster, and accessible to people who previously could not afford the lawyers and accountants required to construct those structures. The democratization of opacity is not unambiguously good.
The technical article I read this week spends thousands of words on this problem under the heading of governance and control planes. What did the agent do? Who authorized it? What was the approval chain? What evidence remains after the transaction completes? These are not philosophical questions. They are the infrastructure questions that determine whether agent-to-agent commerce becomes a genuine expansion of economic participation or a new mechanism for old concentrations of power to operate with less friction and less accountability than before.
The Hub
Here is the frame I keep returning to.
The wheel is spinning. The direction it spins, inflation or deflation, growth or contraction, expansion or regulatory tightening, is contested. Everyone has a theory. Everyone is positioned for a particular outcome. The direction of spin determines who wins in each scenario.
The hub is different. The hub must exist for the wheel to be affixed at all. The hub does not depend on the direction of spin. Its necessity is structural, not cyclical.
The accountability layer in agent-to-agent commerce is the hub. Not the model that generates the output. Not the marketplace that matches the buyers and sellers. Not the settlement rail that moves the money. The thing that answers: who authorized this action, on whose behalf, with what oversight, and what evidence remains after it happened.
That layer is required regardless of whether the economy is growing or contracting. It is required regardless of whether AI is inflating asset prices or deflating labor costs. It is required by regulators, by counterparties, by the principals who deployed the agents and are legally responsible for what those agents did. The demand for it does not go away when the market turns. It increases.
Back to the Anonymity Question
The Web That Has No Weaver is a book about Chinese medicine. Its central insight is that conditions arise interdependently, there is no single cause, no single thread that creates the pattern. The web forms from the interaction of everything with everything else.
Agent-to-agent commerce is forming the same way. No single actor is designing it. The protocols are emerging from different communities, crypto-native developers, enterprise software vendors, AI researchers, financial regulators, each working on their piece without full visibility into how the pieces will eventually compose.
The anonymity that removes race from the transaction and the opacity that removes accountability from the principal are the same feature expressed in different directions. The technical infrastructure does not distinguish between them. The governance layer is what has to make that distinction, after the fact, with real consequences.
The question of who gets the abundance from the agent economy is not answered by the protocols. It is answered by who builds the accountability layer, and whose interests that layer is designed to serve.
The hub does not spin. But it determines what the wheel can carry.