The bedroom producer: why the founding team is collapsing to one or two people

Here are three numbers that do not fit the model of a company most of us still carry in our heads. Cursor is generating in the region of forty million dollars of revenue for every person on staff, on two billion dollars of revenue with about fifty people. Midjourney does roughly two hundred million dollars a year with something like eleven employees. And as of the second quarter of this year, solo founders were sixty-three percent of the new companies incorporated in America, an all-time high.

Read those together and something obvious falls out that almost nobody has fully absorbed. The company is shrinking. Not the ambition, not the revenue, not the reach. The headcount. The unit that produces a serious business is collapsing from the institution to the individual, and the entire mythology of founding, the one that says assemble a team, raise a round, scale your headcount, was built for a world that is quietly ending.

I want to argue that this is not a curiosity or a handful of freak outliers. It is a phase change in what a company is, and the clearest way to see where it goes is to look at an industry that already went through the exact same collapse a generation ago.

What happened to the record

Making a record used to require an institution. A studio, which cost a fortune to build and more to hire by the day. A label to finance it. Engineers who understood a mixing desk with a thousand knobs. Session musicians, a producer, a room full of people and a wall of equipment that no individual could own or operate alone. If you wanted to make music that reached anyone, you needed to get inside that institution, and the institution decided who got in. The gatekeeping was not snobbery. It was economics. The tools required all those hands.

Then the tools collapsed into software. A digital audio workstation on a laptop absorbed the mixing desk, the tape machine, the rack of outboard gear, the session players, one instrument at a time, until a single person in a bedroom could produce a record that competed with anything out of a million-dollar studio. And they did. A large share of the music you have heard in the last few years was made by one person in a room, on a machine that cost less than a month of studio time used to.

The unit of music production shrank from the institution to the individual, and it happened fast, and it did not make music worse. It made vastly more of it, from vastly more people, and it broke the grip of the gatekeepers who had controlled access when access was expensive. That is the template. Company-building is now running the identical process, and the digital audio workstation is AI.

The company as bedroom studio

Look at what a founder can now do alone that used to require a department. The engineering that needed a team ships from one person with agentic tools. The design, the copy, the first marketing, the customer support, the data analysis, the legal first drafts, the financial model, each of these was a hire or an agency, and each is now a capable person with the right AI doing the work of several. The founder is the bedroom producer, and the AI is the rack of gear that used to need a studio and a staff to operate.

The numbers are simply this fact showing up in the accounts. When the top AI-native companies do several million dollars of revenue per employee, five or six times what a strong traditional software company manages, that is not a productivity tweak. That is the same collapse the record industry lived through, expressed as a revenue-per-head figure that would have looked like a typo a decade ago. Cursor at forty million a person is a bedroom studio that happens to be selling to the whole world.

And the sixty-three percent solo-founder number is the other half of it, the supply side. When the tools no longer require an institution, people stop assembling one before they start. They just start, the way a teenager with a laptop now just makes the track instead of trying to get signed first.

The bottleneck moved, it did not disappear

Here is the part the excitable version of this story misses, and it is the part that matters most if you are actually founding something. The collapse does not make founding easy. It moves the bottleneck, and it moves it somewhere less forgiving.

When a company needed two hundred people, the constraints were legible and external. Capital, hiring, coordination, the sheer logistics of many hands. You could always, in principle, solve them with more of something. When a company is two people, all of those external constraints fall away and what is left is the two people. Their taste. Their judgment about what to build and what to refuse to build. Their range across domains that used to be someone else's job. Their stamina. The bottleneck stops being the size of the team and becomes the quality of the individual, and that is a far harder thing to fix, because you cannot hire your way out of your own limitations when the entire premise is that you are not hiring.

The bedroom producer taught this lesson too. When the studio no longer gate-kept, the gear stopped being the differentiator, because everyone had the same gear. What separated the one-person records that reached millions from the millions that reached no one was taste, the ear, the judgment about what to keep and what to cut. AI is doing the same thing to founding. It is handing everyone the same near-infinite capability, and in doing so it is making the scarce thing not the capability at all, but the judgment of the person wielding it. More gear never gave anyone taste. More AI will not either.

What this changes about founding now

If you are founding something in this environment, a lot of the inherited advice is now actively wrong, and it is worth being blunt about which parts.

Hire fast is wrong. Every hire in a bedroom-studio company should be either a capability you cannot get from yourself and your tools, or an admission that you failed to automate something you should have. Headcount used to be a proxy for progress and a thing investors rewarded. It is now, increasingly, a cost and a coordination tax that the best companies are conspicuously avoiding. Staying small is not a phase you endure before the proper company arrives. For more and more businesses it is the shape of the finished thing.

Raise big is wrong, or at least no longer automatic. The reason to raise a large round was to fund the team and the runway the building required. When the building requires two people and a stack of subscriptions, the capital case changes entirely, and raising more than you need buys you dilution and a headcount expectation you did not want in exchange for money you cannot usefully deploy.

Let the ambition outrun the headcount. The mental shift is to stop sizing the company to the problem in people and start sizing it in capability. The question is no longer how many people would this take. It is what could two people with these tools actually reach, and the honest answer is now startling.

The real risk, named honestly

I am not going to pretend the tiny company is free. Its weakness is the mirror image of its strength. A company of one or two is fragile in ways a larger one is not: the bus factor is brutal, the loneliness is real, and the breadth demanded of each person is punishing, because there is nobody to hand the thing you are bad at to. The bedroom producer burns out, or hits the edge of their own range, in ways the old studio system absorbed across many people.

So the scarce resource in this new world is not more hands. It is the right one or two people, with the taste to know what to build, the range to cover the ground, and the judgment to protect themselves from the specific failure modes of being tiny. That is precisely the bet we make at Moonlabs. We back the one or two person founding team, because we think that is the shape of the company now, and we spend the twelve weeks making sure the individuals at the centre of it are strong enough to be the whole bottleneck, because in a company this small, they are.

The record industry did not go back to needing a studio once the bedroom producer arrived. It will not, and neither will this. The institution is not coming back as the default unit of a company. The founders who understand that they are now the bedroom producer, that the gear is handled and the taste is everything, are the ones who will make the records that reach the whole world with a team you could fit in a car.

James Freestone is a co-founder of Moonlabs, the operator-led AI incubator and academy, and previously built the home.co.uk, Homemove and homedata.co.uk stack. Moonlabs backs tiny founding teams building at a scale that used to need an institution. Site: moonlab.ventures.