In 1991, a physicist at CERN published a document that almost nobody read.
It was titled Information Management: A Proposal.
His supervisor wrote the words "vague but exciting" on the cover page and handed it back.
The physicist's name was Tim Berners-Lee.
The document became the World Wide Web.
What Berners-Lee built in that first iteration was a web of documents. Static pages. Text and images sitting on servers, available to anyone with a browser and a connection. You could read them. You could follow links between them. You could not do very much else.
This was Web 1.
Then something shifted.
The pages began to talk back.
Forums appeared. Comment sections. Email inboxes accessible from a browser. The web stopped being a library and started being a conversation. Content was no longer something you consumed. It was something you produced, shared, argued about, and built identity around.
This was Web 2.
It gave us social media. It gave us platforms. It gave us the attention economy, the creator economy, and a version of the internet that most people alive today have never known the world without.
And then the web began to own things.
Blockchain introduced the concept of native digital scarcity. For the first time in the history of the internet, something could exist online in a single verifiable instance. Not a file that could be copied infinitely. An asset with provable ownership, transferable without an intermediary, recorded permanently on a ledger that no single authority controlled.
This was the beginning of Web 3.
But Web 3 had a problem.
It was built for humans.
And the next wave of the internet is not being built for humans at all.
THE AGENT ECONOMY
Something is happening right now that has no clean historical parallel.
AI agents are entering the economy.
Not as tools that humans use to accomplish tasks. As participants. Entities that receive instructions, make decisions, execute actions, and interact with services, APIs, and other agents autonomously, with minimal human involvement between the instruction and the outcome.
A founder tells an agent to research three acquisition targets, prepare a briefing on each, schedule introductory calls with the most promising one, and book the private aviation for the site visit.
The agent does not ask clarifying questions at each step.
It moves.
It interacts with research databases, calendar systems, aviation booking platforms, and communication tools, completing a sequence of tasks that would have consumed most of a working day, in a fraction of the time, without a single human touchpoint between the instruction and the completed output.
This is already happening.
The infrastructure supporting it is being built at a speed that most observers are significantly underestimating.
And it has surfaced a problem that nobody had fully solved.
When an agent needs to pay for something, how does it do that?
THE PAYMENT PROBLEM
Human payment infrastructure was built for humans.
Credit cards require a cardholder. Bank accounts require an account holder. Payment processors require a merchant relationship and a settlement window measured in days. Even the most modern fintech infrastructure assumes, somewhere in its architecture, that a person initiated the transaction and a person will receive the funds.
AI agents are not people.
They do not have credit cards.
They cannot open bank accounts.
They cannot call customer service when a payment fails.
And yet they need to pay for things constantly.
API calls. Data access. Compute resources. Research tools. Booking services. Every task an agent executes on behalf of its principal involves accessing resources that cost money, in amounts ranging from fractions of a cent to thousands of dollars, at a speed and frequency that human payment infrastructure was never designed to handle.
The current workaround is clumsy.
Developers pre-load agents with API keys tied to human accounts. They set spending limits and hope the agent stays within them. They build custom payment logic for every service the agent needs to access. They reconcile the charges manually at the end of the billing cycle.
It works, in the way that a bucket works when the pipe is broken.
It is not a solution.
It is a patch on an infrastructure gap that will become more expensive to ignore with every passing month as the agent economy expands.
x402
The protocol worth knowing this edition is x402, an open standard built on the HTTP 402 status code that has existed in the web's architecture since 1991 but was never formally implemented.
The 402 status code was reserved for future use. Its name was Payment Required.
For thirty-four years it sat in the web's specification like an empty room with a door and no furniture.
x402 furnishes the room.
The protocol enables any HTTP resource, any API, any piece of digital content or compute service, to require a payment before granting access, with the payment settled in stablecoins on a blockchain, instantly, without a merchant account, without a payment processor, without a billing relationship, and without any human involvement in the transaction.
An AI agent encounters a resource it needs.
The resource returns a 402 response containing a payment request.
The agent pays.
The resource grants access.
The entire sequence takes milliseconds.
No invoice. No net-30 settlement window. No API key tied to a human account. No reconciliation meeting at the end of the month.
Just a machine that needed something, a resource that required payment, and a protocol that connected them at the speed of code.
This is what internet-native money looks like.
Not money that moves through the internet.
Money that is native to it. Money that behaves like data. Money that flows between machines the way packets flow between servers, without friction, without intermediaries, without the assumption that a human being is standing somewhere in the middle authorizing the movement.
THE CIVILIZATIONAL SHIFT
The founders and leaders I work with operate in worlds where the quality of infrastructure is directly legible in the quality of outcomes.
A private aviation client who books through a network with genuine access to the right operators, not just the listed ones, moves faster and pays less than one who navigates the market without those relationships. A yacht acquisition client who works with someone who understands the full ownership architecture, not just the purchase transaction, builds a position that generates value rather than consuming it.
Infrastructure is invisible until it fails.
And then it is the only thing you can see.
x402 is infrastructure.
Right now it looks like a developer tool, a protocol specification, a piece of technical architecture that belongs in engineering conversations rather than boardrooms.
In five years it will look like the moment the agent economy found its payment rails.
In ten years it will look like what credit cards looked like in 1970.
Something that the previous generation got along without and that the next generation cannot imagine operating without.
Berners-Lee's supervisor called his proposal vague but exciting.
The World Wide Web was neither vague nor, in retrospect, simply exciting.
It was the infrastructure of an era.
x402 is the infrastructure of the next one.
THE FINAL WORD
Every significant shift in the web's history has been a shift in what the web could do with the resources flowing through it.
Web 1 moved information.
Web 2 moved attention.
Web 3 began to move ownership.
What is being built right now moves value at machine speed, between agents, across borders, through services, in amounts too small for traditional payment infrastructure to handle and at frequencies too high for human oversight to track.
The founders who understand this earliest will build the next generation of businesses on top of it, the way the founders of the early 2000s built on top of broadband and the founders of the 2010s built on top of the smartphone.
The infrastructure is being laid.
The room that sat empty for thirty-four years has been furnished.
And the agents are already moving.