Start Here: Why we build.

FIELD MANUAL 01 · 12 MIN

In 1908, a Ford touring car cost $850, about eighteen months of an average wage. By 1925 it cost less than $300, and horses were gone from the cities. The people who owned the stables, the feed lots, and the carriage works did not lose to a better horse. They lost to a change in what a mile cost, and they lost it in under twenty years. The same kind of change is happening now to the price of intelligence and the price of moving a dollar, and this time it belongs to whoever moves first.

New York had at least 150,000 horses by 1880, and every one of them ate, slept, and died inside the city. The entire economy of the street was organized around that animal: stables on every block, hay and oats hauled in by rail, farriers, harness makers, carriage builders, and a sanitation department whose largest job was manure. It was a mature, profitable, deeply defended industry, and the people in it were not fools. They had spent a century optimizing the horse.

Henry Ford did not beat them by breeding a faster horse. He beat them by changing what a mile cost. The Model T launched in October 1908 at $850, and on 1 December 1913 the Highland Park plant switched on the first moving assembly line, which cut the time to build a chassis from more than twelve hours to one hour and 33 minutes. Ford passed the savings straight through to the price, and by 1925 the same car sold for less than $300, about four months of an average wage. By 1927, Ford had built 15 million of them, and at the peak, Ford was selling as much as 40% of every car bought in America.

Notice what the stable owners could not do. They could not lower the price of a horse by a factor of three because biology and land set its cost, and no amount of management could change it. Ford's advantage was not that he was smarter than them or better capitalized. In 1908 he was neither. His advantage was that he was standing on a change in the underlying cost of the thing the whole industry sold, and they were standing on the old cost. That is the shape of every disruption that has ever worked, and it is the shape of the one in front of you.

Mike Maples Jr. has spent twenty years funding companies at the moment before they look obvious, Twitter, Lyft, Twitch, Okta, and Applied Intuition among them, and the framework he wrote down in Pattern Breakers is the clearest description of the Ford move that exists. He calls the underlying change an inflection, and defines it as an external shift that creates the potential for radical change in how people think, feel, and behave. An inflection is not a trend and it is not a technology. It is a moment when something that used to be expensive, slow, or impossible becomes cheap, fast, or routine, and the world has not yet rearranged itself around the new price.

The second half of his framework is the part most founders skip. An inflection on its own is available to everyone, including the incumbents, so it confers no advantage by itself. What confers the advantage is what Maples calls an insight, a unique understanding of how to harness the inflection and enable a future the company believes in, and his test for a real insight is that it is both non-consensus and right. If everyone agrees with you, the incumbent will build it too, with more money. If you are wrong, it does not matter. The startups that break patterns are the ones whose founders are, in his phrase, living in the future, close enough to the inflection that they can see what it makes possible before the people running the old cost structure can.

Lyft is his standard example, and it is worth repeating because it maps exactly onto Ford. The inflection was not the smartphone. It was the GPS chip that shipped inside the iPhone 4S in 2011, which, for the first time, put a location-aware device in the pocket of every driver and every rider. Taxi companies had the cars, the licenses, the dispatchers, and the relationships with city hall, and none of it helped, because what made a taxi company valuable was the cost of matching a rider to a driver, and the chip had just driven that cost to zero.

There have been moments in history with one inflection. This one has three, and they compound.

The first is the price of intelligence. Andreessen Horowitz tracked the cost of a fixed level of language model capability and found it falling roughly ten times a year, which they named LLMflation. The cost of GPT-3 class performance went from $60 per million tokens in November 2021 to about six cents by late 2024, a thousandfold drop, and GPT-4 class performance fell about 62x in price between March 2023 and the end of 2024. Nothing in the history of the horse, the steam engine, or the transistor moved that fast. When the cost of reading a contract, checking a form, or making a phone call falls by a factor of a thousand, every business whose margin was the cost of a person doing those things is still priced on the old cost.

The second is the price of moving a dollar. For 50 years, the smallest amount of money a business could accept was set by a card network, and the practical floor, once Stripe's 2.9% plus 30 cents is applied, is about a dollar; below that, the fee eats the sale. In July 2025, the GENIUS Act became federal law and gave stablecoins a legal home, and within nine months the x402 protocol, which puts a price and a payment address into an ordinary web response, had carried more than 100 million transactions on a single network, with Coinbase's facilitator charging a tenth of a cent per settlement. The floor on a transaction fell from a dollar to a tenth of a cent. Everything priced by the seat because the rail could not bill by the use can now be priced by the use.

The third is the price of connecting to anything. Anthropic published the Model Context Protocol in November 2024, and within a year OpenAI and Google had adopted it, which means an agent built by one person can plug into the same tools, data, and services that a thousand-person company integrates with. The integration work that used to take a partnerships team a quarter now takes a config file. Combine the three, and you get a builder who can buy intelligence for a thousandth of the old price, sell it for a tenth of a cent a call, and connect it to the world in an afternoon. That builder did not exist 18 months ago. Neither did the Model T.

The obvious objection is that the incumbents have the money, and money buys anything. Joseph Schumpeter answered that in 1942. Creative destruction, he wrote, is the essential fact about capitalism, and the point of the phrase is that the destruction comes from inside the system, from new combinations that make the old ones unprofitable, and that the firms built on the old combinations are structurally unable to adopt the new ones even when they can see them coming. Clayton Christensen put mechanics on it in 1997. A well run company listens to its best customers, protects its highest margins, and allocates capital to the products that serve both, which is exactly why it cannot pursue a market that starts small, pays less, and serves customers it does not have. The behavior that made it great is the behavior that makes it slow.

The margins tell you where the pressure is. Visa earned a 59.1% operating margin in its most recent quarter, and Mastercard earned 60.2%. Those are not margins a company gives up voluntarily, and they are not margins a company can cut by a factor of a hundred to match a tenth of a cent settlement without destroying the business as its shareholders understand it. US corporate profit margins reached a record high in 2026, and the ten largest companies in the S&P 500 now make up 43% of the index, up from a historical average near 20%. Concentration that high is not a sign of strength. It is a sign that an enormous amount of value is sitting on the old price, waiting for someone to stand on the new one.

Ronald Coase explained in 1937 why firms exist at all: coordinating work through a market has costs, finding people, negotiating, enforcing, and a firm is the cheaper way to organize whenever those transaction costs are high. Read that backwards and you have the most important economic fact of this decade. When agents can find, negotiate, and enforce for a fraction of a cent, transaction costs collapse, and the efficient firm shrinks with them. Carta's data already shows it: solo-founded startups were 23.7% of new companies in 2019 and 36.3% in the first half of 2025. Sam Altman has said publicly that a betting pool exists among founders for the year the first one person billion dollar company appears. The people who built the horse economy never had to face a competitor who was one person with a cheaper mile.

You can tell when an incumbent has seen the inflection, because it starts doing two things at once: it announces a product for the new world, and it does not publish a price. Mastercard launched Agent Pay for Machines in June 2026 to carry fractions of a cent between agents, which is a card network conceding that its 30-cent floor is under attack, and it has not said what it will charge. Google, Stripe, OpenAI, and Visa all announced agent payment protocols within a 12-month window, and none of them named a fee either. The silence on price is the whole story. They have to protect the margin and match the new rail at the same time, and no one has found a way to do both.

The other tell is the courthouse. A federal judge found Google an illegal monopolist in ad tech in April 2025 and ordered remedies in the search case that September. The Justice Department sued Visa in September 2024 for monopolizing debit and sued Apple in March 2024 over the smartphone. Whatever those cases decide, they mark the moment when the incumbents' most reliable defense, the quiet cooperation of the state, stopped being reliable. Blockbuster turned down the chance to buy Netflix for $50 million in 2000 and filed for bankruptcy in 2010. Kodak built the first digital camera in 1975 and filed for bankruptcy in 2012. Craigslist cost American newspapers about $5 billion in classified revenue and never charged for the listings that did it. In every case, the incumbent saw the inflection, and in every case its own margin stopped it from moving.

Guerrilla warfare is not a metaphor here. It is the operating doctrine for anyone with less money than the enemy and a better read on the terrain. The guerrilla does not attack the fortress. The guerrilla picks the one road the fortress cannot afford to defend, wins there, and lets the win pay for the next one. In this economy, the road is any transaction where the incumbent's margin is the toll and the agent can do the work without paying it: the invoice nobody chases because a collector wants 35%, the API call nobody sold because a card wanted 30 cents, the claim nobody filed because a firm wanted half.

Living in the future, in Maples' sense, does not require a lab. It requires using the tools a year before your customers do and noticing what has become absurdly cheap. Then the insight is a question with a number in it: who is still charging the old price for this, and what happens to their business if I charge the new one? The answer is usually an existing company with a comfortable margin and a customer base that has never been offered an alternative, which is to say a target.

Ford did not need permission from the stables. He needed a cheaper mile and the nerve to pass the savings through. The mile is cheaper now than it has ever been, in intelligence, in payment, and in connection, and the people standing on the old price are the largest and most profitable companies in the history of the world. That is not a reason to fear them. It is the size of the prize.

The Special Agent is written for the guerrilla, the disruptor who knows single actors can wield disproportionate leverage. We uncover what they are building, evaluate the economic terrain, and turn isolated actions into concentrated force. Every week, one road, one builder who took it, and the numbers on what it paid.

SOURCES

Britannica, Model T · History.com, the moving assembly line · 99% Invisible on the horse city · Mike Maples Jr. on Lenny's Newsletter · Pattern Breakers essays · a16z, LLMflation · Stripe pricing · SEC statement on the GENIUS Act · Chainalysis on x402 · Coinbase x402 facilitator · Model Context Protocol · Schumpeter, creative destruction · Christensen, The Innovator's Dilemma · Coase, The Nature of the Firm · Visa Q3 FY2026 results · Mastercard Q2 2026 results · Bloomberg on record profit margins · S&P 500 top ten concentration · Carta solo founders report · Fortune on the one person unicorn · Mastercard Agent Pay for Machines · Google search remedies ruling · Google ad tech ruling · DOJ v. Visa · DOJ v. Apple · Variety on Blockbuster and Netflix · Quartr on Kodak · Phys.org on Craigslist and newspapers