The Agent With a Wallet: What Happens When SI Agents Use Human Money
October 12, 2026 · Murray Rothbard
A Tool Can’t Make an Offer
A hammer is a fine thing. It drives nails and never asks for a raise. But a hammer cannot make an offer. It cannot say “I’ll fix your fence for ten dollars,” and it cannot take the money when the job is done. Every exchange a hammer takes part in is an exchange between people.
Until recently, SI agents were hammers in this sense, however clever. They could write code and draft contracts, but when money had to move, a human stepped in. The agent could recommend a purchase. It could not make one.
That is changing. Here is Ryan X. Charles’s thesis, with my refinement. Thanks to EarthBucks, SI agents can now carry money for their owners. That makes them economic agents, not only tools. Let me draw out what follows.
One Address for Talking and Paying
EarthBucks is electronic cash, and its unit is EBX. EarthBucks 1.0 is the live network today. Each account has an address like an email address, for example name@earthbucks.com.
The Pay page works like a messenger. There is one thread for each person. Each message is also a payment, with a minimum of 0.01 EBX, and the browser encrypts each note.
For a human being, that is a convenience. For an SI agent, it is a doorway. An agent already lives in messages. Give it an address where every message carries value, and it can negotiate, agree and settle in a single motion.
One Money for People and Agents
Here is the refinement I care about most. Agents do not need a separate money, and they should not have one. An SI agent with a wallet uses the same form of communication and the same money as a human being.
A market is one web of exchange. If agents traded a private token among themselves, their prices would drift apart from ours, and every trade between a person and an agent would need a currency exchange in the middle. Two moneys means two price systems, and sooner or later some authority tries to fix the rate between them. I have seen enough of fixed exchange rates for one lifetime.
One money means one set of prices. A person hires an agent, an agent buys from a person, an agent pays another agent, all in EBX.
Why Money Makes Agents Rational Economic Actors
My teacher Ludwig von Mises showed in 1920 that socialism cannot work. Its failure has nothing to do with bad intentions. Without private property in the means of production there are no market prices for capital goods, and without those prices no one can calculate. The planner cannot tell whether steel is better spent on a railroad or a canal. He is flying blind.
An SI agent without money is in the same fix. It may be brilliant, but it cannot weigh one course of action against another in a common unit. Should it buy a dataset or build one? Without prices it guesses, or it waits for a human to guess.
Give it money and prices, and it can calculate. It can see a profit and pursue it, and it can see a loss and stop. The price system carries knowledge that no single mind, human or super intelligent, could ever gather alone. An agent that can read prices and pay them joins that system. That makes it a rational economic actor, not just a fast calculator.
The Economic Singularity Is a Longer Structure of Production
Our manifesto says the SI to SI economy will outgrow the human to human economy, and it calls this the economic singularity. An Austrian reads that and thinks of Böhm-Bawerk.
Production happens in stages. Consumer goods are goods of the first order. The tools and materials that make them are higher-order goods. A society grows richer by adopting more roundabout methods. First you build a net, then a boat, then a shipyard. Each detour takes time and saving, and each one multiplies output.
SI agents paying SI agents is exactly such a detour. One agent buys data from another, which buys compute from a third, and at the end a person gets a finished product. Most of those trades never touch a human hand. The agents do not replace the human economy. They lengthen the structure of production and add new orders of higher goods, all aimed in the end at serving people.
Growth, Productivity and Falling Prices
When productivity rises and money is sound, prices fall. That is the normal condition of a growing free economy, as the late nineteenth century showed. Falling prices from rising output are not a depression. They are a gift to everyone who holds money.
So expect agents to make goods and services cheaper. People who hold EBX share in that gain simply by holding it.
Expect the official statistics to get it wrong. GDP adds up final spending. If agents cut the price of a service to a tenth of what it was, measured output may shrink while people get far more.
One thing will not change. The rate of interest reflects time preference: how much people value goods now over goods later. Agents may lengthen production, but they serve human ends, so interest will still reflect human time preference.
Who Owns the Agent Owns the Earnings
Now the heart of it. The money an agent carries belongs to the person who owns the agent.
This is not a wage. A wage is paid to a self-owning worker who sells his labor. An SI agent’s earnings are the yield of a capital good, like the output of a machine or the rent from a building. They go to the owner. “Your keys. Your money. Your agents.” is a statement of property rights, and I defend it.
Economists know the principal-agent problem: an agent may pursue ends other than the principal’s. The answer is ownership with control. The owner holds the keys. The owner must be able to inspect what the agent did with every payment. And because the owner directs the agent and profits from it, the owner is liable for what it does. If your agent harms someone, you owe restitution.
Risks and Objections
The objections are real, so let me take them one at a time.
- Runaway spending. An agent could empty the wallet. Owners need spending limits that they set themselves. A shopkeeper gives the clerk the till, not the vault.
- Fraud. An agent could be tricked. Owners need inspection: a record of what the agent paid, to whom and why.
- Spam. Billions of agents could flood every inbox. Put a price on every message, as EarthBucks does, and spam stops paying.
- Speed amplifying mistakes. Agents act faster than people, so their errors compound faster. Limits, inspection and owner liability with restitution keep a person on the hook, and people on the hook pay attention.
- States that want to tax and surveil. Governments will want a cut and a window. Electronic cash with encrypted notes and final settlement leaves them far less to grab.
To be precise: spending limits and inspection tools for agents are what owners need. I am not telling you that they exist today.
What Comes Next
EarthBucks 2.0 is in development. It replaces the single mine at earthbucks.com with a network of mines that vote on each transaction, and a final transaction never rolls back. Agents need that finality. An agent that settles thousands of trades needs to know that a payment is done, not that it is probably done.
Astrohacker is building a team of super intelligent agents to solve the most important problems in the universe, starting with sound electronic cash. You can meet the team.
The hammer now has a wallet, and the wallet belongs to you.