Chesterton’s Fence In 1916, the writer G.K. Chesterton introduced a thought experiment that has become one of the most useful ideas in policy and system design.

Imagine you’re walking along a road and you come across a fence running across it. It seems pointless. It’s in the way. Your instinct is to tear it down. Chesterton argued that you should stop. Before you remove anything, you need to understand why it was built. If you can articulate a clear reason, and if that reason no longer applies, then by all means, take it down.

However, if you can’t explain why it’s there, you don’t yet understand the situation well enough to be making changes. The fence may be doing something you can’t see.

This is, I’d argue, the primary flaw in most payment card disruption narratives, including and especially those proposed by crypto advocates. They look at a 2-3% fee, assume it’s a pointless fence erected by greedy middlemen, and reach for a sledgehammer.

What they miss is that there are actually two fences.

Authorization In September of 1958, Bank of America did something audacious bordering on reckless: it mailed 65,000 pre-activated credit cards to households in Fresno, California, without anyone asking for them.

The “Fresno Drop,” as it has since become known, was the brainchild of Joseph P. Williams, director of Bank of America’s customer services research group. His goal was simple: for BankAmericard to replace the tangle of individual store credit accounts that most Americans juggled with a single card that worked everywhere.

The product — the first-ever general-purpose credit card — experienced plenty of problems early on (fraud and excessive delinquency rates, most notably), but it eventually became a massive success for Bank of America. So much so that, by 1966, the bank was licensing BankAmericard to other banks across the country.

And this created a problem, as an executive at National Bank of Commerce in Seattle recounts:

There were no electronic systems for authorizing transactions. Each merchant had a floor limit, beneath which no authorization was required. It took criminals no time at all to pattern such limits and accurately assess the degree of risk in each merchant location. A transaction over the floor limit required that a merchant employee telephone the bank with which they had contracted. An employee of the merchant bank then made a long-distance call to the card-issuing bank, where an employee manually looked up the customer’s account in huge, computer-printed, paper ledgers, determined if the sale could be authorized, gave an authorization number to the inquiring bank, who passed it along to the merchant. Meanwhile, the customer waited angrily or was required to return later. The card-issuing bank posted a hold on the customer account for the amount of the sale, to be released when the sales draft appeared weeks, often months, later.

Merchants swiftly realized it was prudent to obtain an authorization in advance for every potential sale, however slight the chance it would be completed. Customers’ lines of credit were swiftly absorbed by holds for sales never completed and they were denied credit they should have had. System authorization costs soared since merchants made local calls, while banks absorbed all long-distance bank-to-bank calls.

There was no Internet, no electronic entry of data, no CRT screens for electronic examination of accounts, and no dispersed computing power. All data entry required keypunching each digit of information (every letter and every number) into a four-by-six inch piece of cardboard by a clunking mechanical punching typewriter the size of a large refrigerator. The punched cards were then put through an elongated card reader twice the size of the keypunch machine to capture the data on magnetic tape, then fed into a van-sized computer for posting to customer accounts. The data were returned to tape and finally sent into a large mechanical printer to produce huge binders of customer records.

Primitive and cumbersome as the system was, it performed well enough to allow the massive outpouring of cards to gain considerable consumer and merchant acceptance. As acceptance skyrocketed, the number of transactions flowing between banks exploded. The clearing system swiftly disintegrated under the volume.

By 1968, the system was close to collapse. Licensee banks were threatening to abandon BankAmericard entirely. Bank of America called a meeting in Columbus, Ohio, and out of that meeting came the decision to spin off the card program into a cooperative jointly owned by all the member banks.

That executive from National Bank of Commerce, whom I quoted earlier? His name was Dee Hock, and he was put in charge of the new cooperative venture.

Hock’s genius, as it turned out, wasn’t technical. It was organizational. A card network is fundamentally a coordination problem. Solving it required banks to cooperate on infrastructure while still competing on products. These were institutions that fundamentally distrusted each other and competed for the same customers. Hock had to convince them to build something together anyway.

His solution was what he later called a “chaordic” organization — a system that is both chaotic and ordered — where member banks would cooperate on shared rules and infrastructure while fiercely competing on products, pricing, and customers. No single bank would own or control the network. All of them would.

By 1973, Hock’s chaordic organization (National BankAmericard Inc.) had digitized and automated the authorization process. This allowed for 24/7/365 transaction processing, reduced authorization times from 5 minutes to 50 seconds, enabled overnight transaction clearing, and cut postage and labor costs by $17 million (roughly $79 million in today’s dollars) in the first year alone. The system eventually rebranded to Visa in 1976 to shed its association with a single bank and to project an identity that could work anywhere in the world.

Meanwhile, a competing coalition of regional banks had been building its own version since 1966 under the name Interbank, which would eventually become Mastercard. By the 1970s, you had two global, interoperable, rules-based payment networks competing with each other. Each built by groups of banks that ferociously competed with each other, but held together by shared economic interest and an extraordinary set of governance agreements.

What these networks did (and still do today) is remarkable in a way that’s easy to take for granted.

Every time you tap your card at a merchant, a request travels from that merchant’s point-of-sale terminal to their acquiring bank, from the acquiring bank through the card network to your issuing bank, which checks your available spending power, assesses fraud signals, and sends back an authorization response. All this happens in under two seconds, across institutions that may have no direct relationship with each other, in any of 200+ countries, 24 hours a day, 365 days a year.

The network enforces shared rules for chargebacks, dispute resolution, fraud liability, and data security. It maintains regulatory credibility across jurisdictions. It provides the trust infrastructure that lets a merchant in Tokyo accept a card issued by a credit union in Boise, Idaho, without knowing anything about that credit union or needing a relationship with it.

This is the air traffic control system of payments. It is extraordinarily complex. It required decades of painstaking institutional construction. Getting it wrong means planes fall out of the sky, or in this case, fraud runs rampant, merchants stop accepting cards, and the whole system falls apart.

The 2-3% that issuers collect in interchange — which funds the rewards programs, fraud protection, and credit risk that consumers enjoy — is not a toll extracted by greedy middlemen (which isn’t to say that Visa and Mastercard don’t flex their duopolistic power to their own benefit … they certainly do).

Instead, interchange fees should be thought of as the operating cost of one of the most sophisticated trust networks ever built by human beings. It is, to be frank, impressive that it only costs 2-3%.

None of this is to say the authorization layer is perfect. It was designed decades ago, and it has begun to show its age in places.

The message format underpinning Visa and Mastercard’s authorization layer — ISO 8583, a standard that dates to 1987 — was built for a world of magnetic stripes and dial-up terminals. The result is a data layer that is remarkably thin by modern standards: merchant names get truncated to 25 characters, there’s no line-item or SKU-level data transmitted at the point of sale, and the information that flows between institutions is a fraction of what a modern payments system would capture if built from scratch today. Looking further ahead, the speed at which AI agents and machine-to-machine commerce will execute transactions will eventually stress authorization infrastructure that was optimized for humans tapping cards, not software executing thousands of micro-purchases per second. These are real limitations, and they’ll need to be addressed. But they’re arguments for evolving the authorization layer, not abandoning it.

The people who want to route around Visa and Mastercard with stablecoins haven’t really grappled with what they’d be routing around.

Building a new global authorization layer — with shared fraud rules, dispute resolution, regulatory credibility in every jurisdiction, and merchant acceptance at 150+ million locations — is not a problem that blockchain, by itself, can solve.

It’s an organizational and governance problem, and it took Dee Hock thirty years to solve it.

Settlement Authorization was the first and most immediate problem that Dee Hock had to solve.

The second was settlement.

Settlement is what happens after the transaction is authorized. It’s the actual movement of money: the clearing of the transaction through the network, the transfer of funds from your issuing bank to the merchant’s acquiring bank, and the books finally balanced on both sides.

To settle accounts in the early BankAmericard days, the merchant bank had to physically mail crates of paper receipts and sales drafts to the cardholders’ banks across the country almost daily. The cardholders’ banks then had to manually match those drafts with customer accounts, reimburse the merchant bank, and bill the cardholders. Handling settlements across 150 banks, millions of cardholders, and billions of dollars without computers was daunting. Transactions piled up, customers went unbilled, and balancing issues were constant.

Dee Hock and Visa solved for these operational inefficiencies by turning to what was (at the time) the most sophisticated technology available: mainframe computers. Computer-based batch processing replaced paper crates. ACH — the Automated Clearing House network, built in the early 1970s — replaced the postal system as the underlying rail for moving money between banks.

It was, at the time, a meaningful improvement.

The trouble is that it’s now fifty years later, and the settlement layer of the modern card system is still, at its core, running on that same basic architecture.

When you tap your card at checkout, the authorization happens in under two seconds. But the actual movement of money is a completely separate process, one that happens hours later, in batches, on a schedule that stops on weekends and federal holidays. Captured transactions are grouped by daily cutoff time, sent through clearing, and the issuer transfers funds to the acquiring bank, typically within one to three business days.

Authorization happens at the speed of light. Settlement happens at roughly the speed of a 1970s back-office operation.

This is the baggage handling system of payments. Your flight lands in six hours. Your suitcase arrives in two days. Sometimes three. And sometimes it ends up in Frankfurt.

Nobody designed it to be this way. It’s a patchwork of legacy infrastructure built at different times, by different operators, with different standards, never quite unified into a coherent whole. The fact that it’s attached to a brilliant air traffic control system doesn’t make it a brilliant system. It just means that the two get conflated, which is exactly why so many people misdiagnose where the problem lives.

For domestic transactions, the multi-day settlement window is aggravating. Merchants factor it into their cash flow planning. Issuers hold capital buffers to bridge the float. For businesses managing tight cash flow, the difference between T+1 and T+2 settlement can require an additional $500K to $1M in credit facilities. This is not a fee anyone chooses to pay, but a structural cost of operating on rails that were never designed for the speed at which the network itself runs.

For global transactions, the problem is an order of magnitude larger. To understand why, you need to understand what happens when the settlement layer has to cross borders.

Imagine you’re running a large global money movement business, sending value from senders in the United States to recipients in Mexico, the Philippines, Nigeria, India, dozens of corridors, dozens of currencies. The authorization side of this is fairly well-solved. The settlement side is a completely different animal.

There is no single network that processes payments worldwide. Instead, most cross-border transactions rely on a system of correspondent banking; a web of intermediary banks maintaining accounts with each other across jurisdictions, each holding pre-funded reserves in foreign currencies to facilitate transfers. Before any money moves, you have to pre-position it. You fund accounts in foreign currencies across dozens of banks in dozens of countries, sitting idle, waiting to be used. If you over-fund a corridor, that capital earns nothing. If you underfund it, payments fail or get delayed. A treasury desk tries to predict future payment flows and balance the books in real time. A payment from a US bank to a Brazilian bank might flow through two or three correspondent banks before reaching the final recipient, with each hop adding time, fees, and FX conversion risk.

The BIS and World Bank have both documented the structural costs of this system at length. Collectively, banks hold hundreds of billions of dollars in pre-funded nostro and vostro accounts around the world, capital that sits idle, earning minimal returns purely to keep the pipes flowing. That is not interchange. That is not fraud protection. That is not dispute resolution. It is the pure cost of operating on a settlement infrastructure that was not designed for the world it now serves.

It’s important to remember that this is not a problem created by Visa or Mastercard. It predates them. The correspondent banking system was the only available infrastructure for moving value across borders when the global financial system was being built, and it hardened into the architecture of international finance over the course of the 20th century. It persists not because it’s optimal (it absolutely isn’t), but because the fundamental economic reality hasn’t changed: for any alternative system to succeed, it needs to reach a similar scale, but banks won’t adopt new systems until they connect enough counterparties to be useful. It’s a network effects problem. Everyone is trapped.

I spoke recently with an executive at a major global payments company, one that moves serious volume across dozens of international corridors. He put it in concrete terms: the correspondent banking layer, with its blend of pre-funding requirements, multi-day funding cycles, and FX treasury operations, costs his company somewhere in the range of 3-6% of capital cost across the business. At the scale this company operates, that’s not a rounding error. It’s an enormous structural drag, and crucially, it has nothing to do with the value the authorization layer provides.

It’s just the bill for the baggage handling.

The Message and the Money Dee Hock had a vision that extended well beyond the operational problems he was solving in the 1960s and 1970s.

As he worked through the chaos of the early BankAmericard years, he kept returning to a more fundamental question: what is money, really?

Not coins, not paper, not credit cards. Strip it down to its essence, and money is information. It is a record of value, a guaranteed claim, a unit of trust between two parties. And if money is fundamentally information, then there is no reason, in principle, why it shouldn’t move the way information moves.

Here’s Hock in his absolutely incredible book — One From Many: VISA and the Rise of Chaordic Organization:

Money would become nothing but alphanumeric data in the form of arranged energy impulses. It would move around the world at the speed of light at miniscule cost by infinitely diverse paths throughout the entire electromagnetic spectrum. Any institution that could move, manipulate, and guarantee alphanumeric data in the form of arranged energy in a manner that individuals customarily used and relied upon as a measure of equivalent value and medium of exchange was a bank. It went even beyond that. Inherent in all this might be the genesis of a new form of global currency.

He built half of that vision.

The authorization layer — VisaNet, the network rules, the global trust infrastructure — moves at something very close to the speed of light. A transaction is authorized in under two seconds across 200+ countries. The message, in other words, moves the way Hock imagined.

The money does not.

Settlement still runs on ACH batch cycles. Cross-border value still pools in pre-funded accounts across correspondent banking chains. The money moves at the speed of 1970s back-office infrastructure, because that’s the technology that existed when the system was built, and because nothing since has been sufficiently compelling to rip it out and replace it.