How the first solo-founder unicorn gets built

An old fish swims past two younger ones and nods. “Morning, boys. How’s the water?” The two swim on, and after a while one of them looks at the other and asks, “What the hell is water?”

David Foster Wallace told that story in his 2005 commencement speech at Kenyon College, later published as This Is Water. His point was that the most obvious, important realities are often the hardest ones to see and talk about, precisely because they’re everywhere. The firm is one of those realities. Almost everyone reading this has spent their whole working life inside a company, or building one, or selling to one, and almost no one stops to ask the strange question underneath all of it: why do companies exist at all? It’s the water we swim in.

One economist thought to ask why companies existed. The answer he reached held up for nearly a century, until something started bending it, and this essay is about that bend and where it leads. Within a few years, someone will build a company worth a billion dollars alone. One equity holder. No real payroll. A single person at the center of work that used to take a few hundred people. Call it the first solo-founder unicorn. The idea is already in the air: Sam Altman has said the tech-CEO group chat he is in runs a betting pool on the year the first one-person billion-dollar company appears, a thing he calls unimaginable without AI. That it’s a wager now and not a fantasy is the tell. But a bet isn’t an argument, and most versions of this one are slogans. To make the version that holds up, you have to go back to the one economist who saw it.

Why firms exist in the first place

In 1937, Ronald Coase noticed the water. He asked a question economists had mostly swum past: if markets are such an efficient way to coordinate production, why is so much production organized inside firms, where a boss directs work by command rather than by price? His paper, “The Nature of the Firm,” gave the answer that still holds up: using the market isn’t free. Every market transaction carries costs that have nothing to do with the price of the thing itself. You have to find the right counterparty, negotiate terms, write and enforce a contract, monitor the work, and handle the exceptions when reality diverges from the agreement.

Coase called these the costs of using the price mechanism. When they’re high enough, it’s cheaper to pull an activity inside the firm and direct it by authority instead of buying it on the open market. When they’re low enough, you buy. A firm grows, in his account, right up to the point where the cost of organizing one more transaction internally equals the cost of getting that same thing through the market. That margin is the theory. The boundary of the firm sits exactly where the two costs balance.

Oliver Williamson later sharpened this into the study of asset specificity and opportunism, which is why he and Coase both have Nobel prizes and why the framework has lasted. But the spine is Coase’s, and it’s enough for the argument here: firms exist to economize on the transaction costs of using the market, and they stop growing where that economy runs out.

The easy story, and the fact that kills it

The obvious thing to say about AI is that it slashes those transaction costs, so the boundary collapses and firms dissolve into the market. Everything gets outsourced to a swarm of contractors and agents, the firm shrinks toward nothing, and a single founder commands the whole apparatus through a screen.

The problem with that story is that we have run the experiment already, and it didn’t produce a solo unicorn.

For roughly fifteen years the gig economy has driven the dollar cost of a market transaction toward the floor. Upwork and Fiverr put a global labor pool one search away. Fractional executives, fractional finance, fractional design, fractional everything became normal. Spinning up a contractor went from a procurement project to an afternoon. By Coase’s logic, as the price of each market transaction fell, the boundary of the firm should have moved sharply toward the individual. Solo operators running large distributed operations should have become common.

They didn’t. The gig economy made one-person businesses easier and a few of them lucrative, but it didn’t produce a single founder running a billion-dollar company alone. The cheap transactions were right there and nobody assembled them into a unicorn.

The reason is the part the easy story leaves out. Outsourcing lowered the price of each transaction. It did almost nothing to the coordination load the transactions placed on the founder. Every contractor still has to be briefed, chased, corrected, and stitched into everyone else’s work. Every handoff is a small negotiation. Every exception escalates to the one person who holds the whole picture. That work doesn’t show up on an invoice, but it’s a transaction cost in Coase’s exact sense: search, negotiation, monitoring, exception handling. And all of it lands on a single human’s attention, which doesn’t scale.

So the firm didn’t persist because hiring contractors was expensive in dollars. It persisted because coordinating them was expensive in the founder’s time, and a person has a hard ceiling on how many relationships they can actively hold. You hire employees and managers precisely to absorb that load. The org chart is a machine for spreading coordination across many heads because it could never fit in one.

The gig economy attacked the wrong cost. It made transactions cheap while leaving the coordinator a bottleneck. Coase’s boundary didn’t move much, because the bottleneck was never the price.

And that fragmentation hasn’t stopped. It’s accelerating, and it’s now reaching inside the job itself. Maria Black, who runs ADP and sees payroll data for one in six American workers, describes the present moment as a shift from jobs to tasks: take a job, unbundle it into a series of tasks, and put a wage on each one. The real-time tracker ADP built with the Stanford Digital Economy Lab already shows up to sixteen percent fewer jobs in AI-exposed work since 2022. Read one way, that’s a story about jobs disappearing. Read another, it’s a story about supply: the unit of outsourceable work is shrinking from a whole job to a single task, so more of what a company does becomes a discrete, contractible, machine-routable piece.

That’s the gig-economy trend run forward, and it widens what a lone founder could in principle hand off. But it cuts the other way too. Every job that splits into five tasks is five briefs, five handoffs, five things to check, where there used to be one hire who held all of it together internally. Finer decomposition lowers the cost of buying each piece and raises the cost of coordinating the whole. It’s the same fork as before, only sharper: the pieces are more outsourceable than ever, and more uncoordinated than ever. The accelerant is real, but only for whoever can put the pieces back together.

What’s actually different now

The new thing isn’t cheaper transactions. The gig economy already delivered those. The new thing is the first technology that scales a single person’s coordination capacity.

AI grounded in a knowledge layer and running workflows can absorb the brief-chase-correct-integrate loop that used to consume the founder. It can hold context across many parallel relationships, draft the handoffs, monitor the work, and surface only the exceptions that genuinely need a human decision. That’s a direct attack on the residual transaction cost the gig economy couldn’t touch: not the price of a contractor, but the load of running a hundred of them at once.

If that load can be carried by software instead of by the founder’s inbox and memory, then the ceiling on how many external relationships one person can command rises by an order of magnitude. And that’s the variable Coase’s boundary actually turns on for a would-be solo operator. The market got cheap years ago. What was missing was a coordinator who didn’t run out of attention.

So the boundary tips toward the individual, but not for the single reason the easy story gave. It takes two trends at once. The market kept getting cheaper, the unit of outsourceable work shrinking from a whole job to a single task, so more of what a company does became a contractible piece than ever before. On its own that just piles up coordination the founder can’t carry. What tips the boundary is the second trend landing on top of the first: a coordinator whose attention never runs out, finally able to hold all those pieces together.

The same insight, applied to the edge of the firm

This is the mirror image of an argument Dorsey and Botha made about the inside of the firm. Their point is that hierarchy exists to route information, that a manager can only directly coordinate a handful of people, and that AI lifts that span-of-control limit, so a company needs fewer managers and fewer layers. We think they’re right, and we have written about why.

The argument here applies the identical mechanism to the firm’s boundary rather than its interior. A founder, like a manager, has a span-of-control limit, except theirs is over external relationships rather than internal reports. The same thing that lets a company flatten its org chart, AI absorbing the coordination a human used to do by hand, lets a company shrink toward its edge until the edge is one person. Internal span of control and external span of control are the same constraint measured from two sides. Remove it on the inside and you get a flatter company. Remove it at the boundary and you get a smaller one. They’re each half of one idea.

What this doesn’t mean

The real claim is narrower than “the firm is dead,” and the difference matters.

Coase’s boundary relocates. It doesn’t disappear. Some work still clusters into firms no matter how cheap coordination gets, for reasons that have nothing to do with the cost of a transaction and everything to do with what arm’s-length contracts can’t do well:

• Deep, asset-specific collaboration. Work where two parties invest in each other in ways that only pay off together is exactly the case Williamson showed resists the market. A long, intertwined R&D effort doesn’t decompose into clean gig-sized pieces.
• Trust and tacit knowledge. Some understanding lives between people and never makes it into a brief. The handoff itself destroys it.
• Liability, IP, and regulation. When someone has to be legally accountable, or when the value is the secret, you keep it inside. A contractor network is a poor container for either.
• Brand and relationships. Customers and partners often want a counterparty that will still be there next year, with skin in the game, not a coordinator orchestrating strangers.

For that class of work, the firm stays a firm. What collapses isn’t every company but the minimum viable firm for a large class of work that’s coordination-heavy and collaboration-light: agencies, content and media operations, e-commerce brands, software with a narrow surface area, services that are mostly assembly and routing. For those, the floor on how many people you need is dropping toward one.

Hierarchy doesn’t vanish here either; it relocates the way the boundary does. A solo founder who leans on agencies and contractors is delegating into other people’s org charts, not abolishing the org chart. The pyramids are still there, just outside the founder’s own company. And some of them are no longer made of people: the coordination layer itself runs hierarchically, a lead agent breaking a goal into parts and handing them to sub-agents. “Solo” describes who owns the company and sits at its center, not a world with no management in it. The management mostly moved out of the founder’s firm, into supplier firms and into the machine.

And we’re not starting from zero. Solo and near-solo founders keep reaching higher than anyone expects, running real businesses on little or no payroll. What none of them has done is reach a billion-dollar valuation alone, and the reason isn’t talent or effort, it’s coordination. A small solo business is a product and a set of contractors that one person can still hold in their head. A billion-dollar one is a company’s worth of relationships, the thing a single human could never hold at once.

The famous small-team outliers show how close the world already got. The thirteen people at Instagram and the fifty-odd at WhatsApp, when each sold for a fortune, proved that headcount and value had decoupled. But thirteen isn’t one, and the gap that remains between a tiny team and a true solo founder is almost entirely communication. Most of what those thirteen did for each other was keep the work in sync: who owns what, what changed, what’s waiting on whom. That’s coordination, and it travels in messages. Automating it is what closes the distance from thirteen to one, the same span-of-control collapse from earlier run at its smallest possible scale. That’s the piece arriving now, and it’s what turns the gap into a prediction.

The prediction, earned

So here is the concrete version. The first solo-founder unicorn isn’t built by a genius doing the work of three hundred people. It’s built by one person sitting at the center of a coordination layer that does the work the three hundred people mostly used to do: the briefing, chasing, integrating, and monitoring of a large fleet of contractors and AI agents. The founder makes the decisions a human still has to make and lets the layer carry the rest. The business is almost certainly in the coordination-heavy, collaboration-light zone above, because that’s where the boundary moves first.

On timing, we will be conservative, because the constraint isn’t raw model capability but the maturity of that coordination layer, and maturing it is harder than describing it. The gig economy’s cheap transactions are already in place. The compute is already in place. The missing piece is software that can carry a founder’s coordination load reliably enough to bet a company on, and that’s a few years of hardening, not a demo. We’d put the first one before the decade is out.

A note on where we sit, since it would be coy to pretend we have no stake. The coordination layer this argument needs is roughly what we’re building: a brain that holds the operational knowledge a founder would otherwise carry in their head, and workflows that move work forward without a person relaying each step. The hardest part of that, the brain keeping itself current automatically from the full stream of messages, is something we’re building toward rather than something we can claim is finished today. We are pointing at the shape of the thing, not declaring it done. This essay is a bet about the economics, and the economics don’t depend on us being the ones who finish it.

If Coase is right that the firm exists to economize on the costs of using the market, then the firm has always been, in part, a workaround for a coordinator who runs out of attention. Remove that constraint and the workaround gets smaller. The first solo unicorn is what it looks like when it gets all the way down to one.

Key takeaways

• Coase’s “The Nature of the Firm” explains firms as a way to economize on the transaction costs of using the market; a firm grows until internal coordination costs as much as buying on the market.
• The naive “AI dissolves the firm” take fails the gig-economy test: fifteen years of cheap contractors lowered the price of transactions but produced no solo unicorn.
• The reason is coordination load, a transaction cost in Coase’s sense, that always fell on one founder’s attention and didn’t scale.
• What’s new is AI that scales a single person’s coordination capacity, attacking exactly the cost the gig economy couldn’t. This is the same span-of-control insight Dorsey and Botha apply inside the firm, applied instead to its boundary.
• The firm doesn’t die; Coase’s boundary relocates. Asset-specific, trust-heavy, and regulated work still clusters. The minimum viable firm collapses toward one for coordination-heavy, collaboration-light work, and the first solo unicorn likely arrives in the back half of the decade.