The Second Telegraph
How AI Agents Made Bitcoin Inevitable
Brian Connelly
In 1844, Samuel Morse tapped out “What hath God wrought” and broke money. Not immediately, not obviously, but irreversibly. The telegraph did something no technology had done before: it separated the promise of payment from the act of payment. You could wire instructions to transfer gold at the speed of electricity, but the gold itself still traveled by ship, by rail, by armed courier. For the first time in the history of commerce, the commitment to pay and the fulfillment of that commitment existed in different time zones.
That gap needed a custodian. Banks stepped in. Clearinghouses stepped in. Correspondent networks, settlement windows, reserve requirements: an entire institutional architecture arose to manage the space between “I will pay” and “you have been paid.” The system worked, more or less, for a century and a half. It also extracted rent at every node, because that’s what custodians do when they hold the gap between promise and delivery.
Lyn Alden traced this history with precision. The telegraph didn’t just speed up communication. It restructured the monetary system around intermediaries who were never supposed to be permanent but who became, over time, the system itself. The gold eventually stopped moving altogether. The promises became the money. And the custodians got greedy, because there was no longer any external reference point to discipline them. Bretton Woods, Nixon’s closing of the gold window, the full float of fiat: each was a logical consequence of that original separation. The gap between commitment and fulfillment didn’t just persist. It became the product.
The Anatomy of the Gap
In the mid-twentieth century, philosophers of language, J.L. Austin in the 1950s, John Searle in the 1960s and 70s, established a framework for understanding what happens when people speak. Language doesn’t just describe the world. It acts on it. “I promise,” “I accept,” “I declare this settled”: these are speech acts, performances that change the state of affairs between people. Every economic transaction is a sequence of these acts: a request, a promise, a fulfillment, a declaration of satisfaction. When the loop completes, the transaction is done. When it doesn’t, you have an open commitment, a thread left hanging, consuming attention, generating risk, requiring someone to track it.
The pre-telegraph economy was slow but its loops were clean. You handed someone gold, they handed you goods. The speech act and the physical settlement were the same event, in the same place, at the same time. The loop closed on contact.
The telegraph opened every loop. You could make the promise at wire speed, but fulfillment still moved at the speed of matter. And once the loop was open, someone had to stand in it. Banks became proxies for fulfillment. They said, in effect, “trust us, the gold is coming.” That’s a promise about a promise, a meta-commitment. And the entire architecture of modern finance is built from increasingly abstract meta-commitments stacked on top of that original gap.
Fractional reserve banking is a promise to pay that assumes not everyone will ask for fulfillment at the same time. Rehypothecation is using someone else’s promised collateral as the basis for your own promise. Derivatives are promises about the future value of other promises. Credit default swaps are promises to cover the failure of promises. Every layer adds distance between the original speech act and actual settlement. The system doesn’t complete conversations for action. It trades them. The loops never close. They just spawn more loops, more intermediaries, more unfulfilled commitments.
This is where Alden’s historical analysis and speech act theory converge. The problem wasn’t just that intermediaries got greedy. It’s that the commitment architecture itself became structurally incapable of closure. The system was designed to manage open loops, not close them. Closing loops would eliminate the need for custodians. The custodians had no interest in that.
Bitcoin: The Loop That Closes
Bitcoin, understood through this lens, is not primarily a store of value, a hedge against inflation, or a speculative asset. It is a commitment protocol. A Bitcoin transaction is a speech act that completes at the protocol level. The request, the promise, the fulfillment, and the confirmation of satisfaction are compressed into a single cryptographic event. There is no meta-commitment. There is no proxy standing in for settlement. No one holds the gap because there is no gap. The loop closes. Period.
This is not a feature of Bitcoin. It is the definition of Bitcoin. Satoshi didn’t set out to build faster payments or cheaper wire transfers. He built a system where commitment and fulfillment are the same event, something that hadn’t existed since before Morse’s first tap.
But here’s the part that hasn’t been said clearly enough: for human-speed commerce, the unclosed loops of fiat finance were a tax, not a fatal flaw. You paid the rent, you tolerated the delay, you absorbed the counterparty risk, because the friction was small relative to the pace of human economic life. A two-day settlement window is annoying when you’re making a handful of transactions a day. The pathology was real but livable. People accommodated. They always do.
Which is why Bitcoin adoption, for all its philosophical elegance, remained a slow burn for fifteen years. The people it served best, the unbanked, the politically targeted, the citizens of collapsed monetary regimes, were at the margins. The center held. The custodians kept extracting. And the loops stayed open, because the humans making the commitments were slow enough that the gap didn’t kill anyone.
The Second Telegraph
Now we have a new telegraph. It’s called artificial intelligence, and it compresses decision-making, negotiation, and settlement into microsecond-scale, millions-per-transaction-per-second frames that traditional banking cannot comprehend, much less process.
Jordi Visser, the macro strategist and former Morgan Stanley derivatives trader, described the transition this way during a recent conversation at Blockworks: the shift from the chatbot era to the agentic era was probably the most important transition of our lives. A chatbot waits for a human to ask a question. An agent acts autonomously, negotiating, executing, acquiring resources, settling transactions, twenty-four hours a day, at machine speed, across millions of simultaneous threads. The compute demand, Visser argues, has already increased by a thousand times over what the chatbot era required.
Think about what that means for the commitment architecture. A single AI agent making thousands of transactions per day is generating thousands of speech acts that need to close. Each unclosed loop is a thread consuming memory, consuming compute, introducing uncertainty into the agent’s decision model. An agent that has to maintain state on thousands of pending settlements, waiting for ACH clearing, waiting for wire confirmation, waiting for compliance review, waiting for a human somewhere in a back office to approve something, is an agent burning resources on commitment tracking instead of productive work.
Scale that to Visser’s vision of trillions of agents and the system doesn’t degrade gracefully. It collapses. The fiat commitment architecture wasn’t designed for that volume of open loops. It was designed for human-scale transaction frequency with human intermediaries managing the gaps manually. There’s no human to call. There’s no relationship manager to expedite. There’s no judgment call to make about whether this particular counterparty is trustworthy. The agent needs the loop to close or it can’t proceed.
The first telegraph created the need for trusted third parties. The second telegraph, AI, destroys the viability of trusted third parties, because no human intermediary can process, verify, and settle at the speed agents operate. The banking system isn’t just slow relative to AI. It is architecturally incapable of participating in agentic commerce. It’s like asking a horse-drawn carriage to merge onto an interstate.
The Flywheel
This is where the relationship between AI and Bitcoin stops being a correlation and becomes a flywheel.
AI agents proliferate. They need trustless, high-velocity settlement to function. Bitcoin provides it, the only monetary protocol where the commitment loop closes at the protocol level, at machine speed, without intermediaries. More economic activity flows through Bitcoin. Network value increases. That attracts more capital and infrastructure investment. That infrastructure, the data centers, the semiconductor supply chains, the energy systems, supports more AI compute, which produces more capable agents, which need more settlement capacity. Each revolution of the wheel strengthens both sides.
The energy-money thinkers who preceded Satoshi, Wilhelm Ostwald, Frederick Soddy, Henry Ford, the Technocrats, argued that money should be grounded in physical reality rather than political abstraction. Bitcoin accomplished that with proof of work. But AI agents add a new dimension: they don’t just need money grounded in physical reality. They need money grounded in computational reality. Proof of work isn’t just an energy anchor. It’s a shared language between AI systems and the monetary network. Both speak compute.
This is not speculation. It is already being built.
Block: The Proof of Concept
Jack Dorsey’s Block, Inc., the company formerly known as Square, has quietly assembled the entire stack.
On the AI side, Block deployed Goose, an internal agentic AI framework, across the company. They reduced their workforce from 12,000 to under 6,000, not by outsourcing, not by offshoring, but by restructuring the company’s commitment architecture around agent capabilities. Block’s CTO described it in terms that would be immediately recognizable to anyone familiar with speech act theory: they stopped thinking of Block as an enterprise with separate tools and departments, and started treating everything, taking payments, moving Bitcoin, issuing invoices, managing code, as capabilities. Then they put an agent middleware layer on top. Every capability became a commitment that an agent could initiate and close.
On the Bitcoin side, Block automatically enabled Bitcoin payments via Lightning Network for millions of eligible U.S. merchants in March 2026. Zero processing fees. Near-instant settlement. The commitment loop closes at the speed of the network, not the speed of a compliance department.
And in between, connecting both sides: Block contributed Goose to the Agentic AI Foundation alongside Anthropic’s Model Context Protocol and OpenAI’s AGENTS.md, the open infrastructure layer that lets agents talk to tools, to each other, and to settlement rails.
The contrast with PayPal is instructive. PayPal is attempting to bolt agentic commerce protocols onto an organization whose commitment network was designed for human-speed, intermediary-dependent settlement. They’re trying to run new software on old architecture. Block rebuilt the architecture first.
This is exactly the pattern that separates enterprises that will capture value in the agentic era from those that won’t. The ones that understand their own commitment networks, who is promising what to whom, where the loops open and close, where the intermediaries add friction rather than value, can restructure around agents and settle on Bitcoin. The ones that don’t will automate their dysfunction at machine speed, which is worse than not automating at all.
The Discounted Cash Flow Problem
Visser makes one more observation that ties this together. The standard method for valuing a company, discounted cash flow analysis, depends on projecting what a business will earn years into the future with reasonable confidence. AI has made those projections meaningless for software companies. If you can’t say what Salesforce or Adobe looks like in three years because agents might replace the entire workflow their products support, the model produces fiction. The inputs are unknowable, so the output is noise.
This is actually the same problem, expressed in financial terms. A DCF model is a series of commitments projected forward: the company commits to generating these cash flows, and investors commit capital based on those projections. When the commitment architecture of entire industries is being restructured by agentic AI, those forward commitments become unreliable. The loops don’t close in three years because you can’t know what the loops will look like in three years.
Bitcoin doesn’t have this problem because Bitcoin was never a commitment about future cash flows. It is a protocol. Protocols capture value through usage, not earnings. TCP/IP doesn’t have cash flows either, but the economy that runs on it is worth tens of trillions. Bitcoin’s value proposition isn’t “we will generate returns.” It’s “we will close your commitment loops at machine speed, without intermediaries, forever.” In an economy where every other commitment is becoming unreliable, the protocol that guarantees closure becomes the most valuable thing in the system.
What Comes Next
The first telegraph separated commitment from fulfillment and created a century and a half of intermediaries. The second telegraph, artificial intelligence, makes that separation unsurvivable. The agents are too fast, too numerous, and too relentless for human-speed custody to function. Every open loop is a cost. Every intermediary is latency. Every settlement delay is compute burning without productive output.
Bitcoin is the only monetary protocol where the commitment loop closes at the protocol level. Not because it’s philosophically superior, though it is. Not because it’s a better store of value, though it is. But because it’s the only money that works at the speed the new telegraph demands.
The flywheel is already turning. Block is proof. The agents are already eating, as Visser says, and they need to be fed. The question of today is not whether Bitcoin becomes the settlement layer of the agentic economy. It is how long it takes everyone else to see what Dorsey already built.