Why fleets

Nathan Daeila, September 2026

In the fall of 1945, fifteen thousand elevator operators and building workers in New York walked off the job. A million and a half people couldn't get to their desks. The strike cost the city about a hundred million dollars in a week.

Here is the part that matters. The automatic elevator had existed since the turn of the century. Any of those buildings could have installed one. Nobody would ride it. So after the strike, the elevator companies spent the next ten years running ads of grandmothers and small children pushing the buttons themselves. They put a telephone and a red stop button in every car. They added a recorded voice: "This is an automatic elevator. Please press the button for the floor you desire."

When was the last time you thought about any of that?

Last week in Austin, a gold two-seat car with no steering wheel started picking up paying passengers. The launch was a mess. The service area is a few square miles. Federal regulators said they were evaluating the situation. Somebody filmed one of Tesla's robotaxis shoving through a row of plastic bollards. All of that is true, and it's 1950. The machine works and people don't trust it yet, and those are two different problems on two different clocks.

Here's what I believe. Within about ten years, being smart will be cheap and owning things will not be. And the physical world is going to be automated one rung at a time, where each rung is a machine we allow to be responsible for a human body. Cars are the first rung. They're also the only one you can buy today at the price of a car. That is why this firm is called Fleet and not Cybercab.

Thinking is getting cheap

The price of a fixed amount of machine intelligence has fallen roughly ten times a year. Nothing physical moves at that speed. Steel doesn't. Wages don't. Anything a person makes by thinking and typing is heading toward the cost of the electricity it took to produce. People still argue about whether the machines will ever be tasteful. A studio spends five years and a few hundred million dollars on a game. Give it a few years and that's a weekend.

Which raises a question nobody likes. If being smart stops being rare, what makes a person valuable?

Three people who saw this before I did

Sam Altman wrote in 2021 that as AI takes over work, "even more power will shift from labor to capital," and named the two things worth owning: companies and land. Anton Korinek, an economist, put it in a model. Wages are a race between automation and capital accumulation, and when automation wins, wages fall to whatever capital earns. Rudolf Laine put it most bluntly last year: whatever you own at the moment intelligence becomes abundant is roughly what you keep.

Those three don't agree on much else. They were also mostly talking about desks. Software, spreadsheets, code. The part I care about starts where the desk ends.

The ladder

Every rung is the same event. A machine takes responsibility for a body, and we let it.

The first rung is cars, and it's further along than most people think. Waymo has driven 220 million miles with nobody in the driver's seat. Compared to humans on the same roads, it's in 94 percent fewer crashes that cause serious injury. It now drives four million miles a week, which by its own math prevents a serious-injury crash about every eight days. The engineering question is mostly answered. The trust question is 1950.

The second rung is trucks. Since the spring of last year, Aurora has run loaded semis between Dallas and Houston with no one in the cab. A quarter million miles, zero collisions attributed to the system, and a plan for two hundred trucks by the end of this year. Nobody held a launch party. Freight just moved.

The third rung is robots with hands. At BMW's plant in South Carolina, humanoid robots from a company called Figure have worked ten-hour shifts, five days a week, for over a year, handling parts for more than thirty thousand cars at better than 99 percent placement accuracy. What they do is still simple. Carry this, place that. That's where the elevator was in 1900.

The top rung is surgery, and you can see it from here. Last year a robot at Johns Hopkins watched videos of surgeons and then removed the gallbladder from pig tissue with no human on the controls. It was slower than a surgeon. It hasn't been near a person. But someone reading this will probably have a spine repaired by something that never gets tired and never has a bad morning, and they'll think the old way was barbaric.

Each rung makes the next one easier to accept. And under every rung sits an asset that somebody has to own. Trucks. Robots. Operating rooms. The cars are just the rung where the asset is already priced like a car.

The first rung you can buy

A Cybercab costs what a car costs. It holds value the way a car holds value. It can be leased and it can be sold, and banks already know how to underwrite it. It is a rational asset today, before any of the software promises come true.

What changes when they do is the income. Morgan Stanley puts Tesla's cost today at 81 cents a mile, against $1.71 for a rideshare with a human in the front seat. Ark thinks the Cybercab gets to 20 cents by 2030. Any of those numbers works. A car that earns eighteen hours a day is a different asset from one that sits in a driveway for twenty-two.

Uber has said its partners have committed to more than a hundred thousand autonomous vehicles over the next few years. Somebody has to buy them, and somebody has to build the lots they come home to at 4 a.m. We buy the cars and build the depots. Operators run them. When the trucks and robots reach the point the cars are at now, we'll own those too. Cars are the first fleet.

What would make me wrong

Reliability, mostly. The Cybercab launch showed exactly what I'd expect: real cars, real rides, and a technology that isn't ready for people to stop paying attention to it. The elevator companies needed ten years after the strike. If autonomy needs ten years, the cars are still cars. They hold value and they can be leased. Solving autonomy is not our job. The bet fails if self-driving never works or the cars never get cheap. I don't think either is likely.

One more thing

This can sound like a doom bet. I think it's the most optimistic one I know how to make. It pays only if we pull off the hardest thing our species has attempted, which is building something smarter than ourselves and trusting it with our bodies. That trust will be earned slowly, rung by rung, the way it was earned in an elevator. Someday somebody will get into a car with no steering wheel and not think about it at all. Nobody will remember the year.

I'd like to own the cars when that happens. If you would too, my email is below.

Nathan Daeilanathan@fleet-cap.com