Every startup guide I read in my twenties said the same thing in a slightly different voice: find a co-founder. Find one before you find a customer. The advice came with a warning built in. The solo founder was a statistical corpse. Investors wouldn't touch one. No single person could hold a company. I believed it, and for the first company I built it was true. Leantree grew to about fifteen people, ran recruiting funnels for companies like Vodafone and TUI, and was sold in early 2026. On almost every day of those six years, someone else in the building knew something I didn't, and that gap was the reason they were in the building.
Today I run six commercial products with no co-founder, no CTO, no head of anything, and fewer decisions landing on my desk per day than I had with fifteen people. The products get built faster than Leantree built anything. I don't say this to boast. It's mostly unsettling. It means the advice I was given was never really about companies. It was about the limits of one person, and those limits have moved.
What the co-founder was for
Strip the mythology off and a co-founder solved two problems. The first was a skill gap. You could sell but not build, or build but not sell, and a company needs both on day one, which is why the phrase "technical co-founder" exists at all. The second was load. There were too many decisions for one head, too many hours in a week where something needed a competent adult, and the equity you gave away bought you a second one.
Almost nobody talks about the third thing, which turned out to be the one that matters. A co-founder was someone whose job it was to tell you that you were wrong. Not a board, which sees you quarterly and mostly sees slides. Not an employee, who is paid to make your idea work. Someone with the same information and the same stake, who could look at your plan on a Tuesday afternoon and say no, and lose money if they were wrong about that.
The first two problems have been solved by software in the last two years. I mean solved in the boring, complete sense, not the demo sense. The skill gap first. I'm not an engineer by training. I ran a marketing company. The operating layer that runs my products today, the thing that watches every service, restarts what falls over, drafts the customer replies and posts the daily digests, was written in conversations with a coding agent over a few months, in the hours a founder used to spend in meetings about the roadmap. I still work with developers on one of the products, but I work with them on a scope, for a period, the way you'd work with a lawyer. Nobody has a title. There's no org chart, because there's nothing to chart.
Load went the same way, only quieter. Most of what a company does in a day is recurring, and recurring work is exactly what an agent on a schedule is good at. The monitoring runs. The lead intake runs. The bookkeeping gets swept by chat once a week. What reaches me is, by construction, the exception, and there are fewer exceptions than there were people in my last company. The second competent adult I used to need is now a set of loops that don't get tired, don't get promoted, and don't need to be told twice.
So the co-founder as a workaround for capacity is gone. Not diminished. Gone. If you're starting a company this year and someone tells you to give away half of it to cover a skill you don't have, they are giving you advice from a world that ended while they weren't looking.
The €800 Tuesday
What isn't gone is the third thing, and I found that out the expensive way.
Earlier this year I set up a batch job that enriched company records with an AI model, one call per record, a few cents each. I checked the first fifty by hand, they looked good, and I pointed it at the full list and went to lunch. When I came back the run had cost about €800, most of it on records nobody was ever going to read. Nothing had gone wrong, technically. The system did exactly what I'd asked it to do, at exactly the speed I'd asked for it. The only thing missing was the question a co-founder would have asked before I pressed enter, which is: how much does this cost if it works?
Nobody asked, because there was nobody there whose job it was to be annoyed by me. I've since written the rule down. Anything over fifty records, I name the cost before I start. It's a good rule. It's also a poor substitute for a person, because a rule only catches the mistake it was written for, and the next mistake will be a different one.
This is the part the guides never put in, and I think I understand why. Disagreement is hard to sell as a feature. "Find someone who will fight you" doesn't fit on a slide next to "complementary skill sets." But when I look back at what my senior people at Leantree actually did for me, the skill coverage is the smaller half. The larger half is the list of things I didn't do because someone I respected said it was a bad idea, and I couldn't prove them wrong in the room.
Rebuilding the argument
So the honest description of running a company alone in 2026 isn't "I don't need anyone." It's that the capacity problem has been solved for me and the disagreement problem has been handed back to me, and I have to build it by hand.
Some of that is mechanical. Before anything I write goes out under the company's name, whether it's a letter to a customer or a page on this site, I run it through a panel of agents that have each been given a different reason to dislike it. One reads as the recipient's lawyer. One reads as the recipient's bored assistant. One is told to find the sentence that would embarrass me in six months. The panel killed a sales letter last month that I'd been proud of for a week, and it was right, and I was irritated for exactly as long as I used to be irritated at a partner who was right. Which is, I've decided, the sign that it's working.
Some of it is about time. I write decisions down before I make them, and read them a week later, when I'm a slightly different person with less at stake. The week-later me disagrees with the original about a third of the time. That's roughly the hit rate I remember from good partners.
And some of it is people, just not people on the payroll. I pay specialists by the question now, not by the month. A Norwegian property lawyer for one afternoon, a tax adviser for one structure, an engineer for one review. The difference from an employee is that they have no reason to tell me what I want to hear, and no reason to stay if I stop listening.
None of this is as good as one person who knows everything I know and is sitting in the next chair. I want to be clear about that. It's cheaper, faster, and it doesn't take equity, and it is still a reconstruction of something that used to come built in.
What it's actually like
The upsides are real and they're not the ones people expect. Speed is the obvious one. An idea on a Monday can be a live product with a payment page by Thursday, because there is no one to convince and no one to wait for. Less obvious is coherence. Six products built by one taste look like they came from one company, because they did, and that turns out to matter to the people who find you by googling the name in an imprint. Least obvious of all: stopping things is free. At Leantree, killing a project meant a conversation with the people who'd built it. Now I can switch a product off on a Wednesday because the numbers said so, and nobody's afternoon is ruined. That freedom changes what you're willing to try.
The downsides are real too, and I'll list the ones that don't get mentioned, because the ones that do, loneliness and the bus factor, are true but obvious.
The first is that removing load didn't remove decisions. It concentrated them. Everything that reaches me is an exception, which means my days are made entirely of exceptions. There is no easy Tuesday where you sign off on the routine and go home early, because the routine no longer needs me. I underestimated how tiring a day of only hard calls is, and I don't think there's a fix for it beyond scheduling days with none.
The second is that the outside world has no slot for you. Procurement forms want a CTO's name. Partners want to talk to "the technical side." Banks want to know who runs the company if something happens to you, and "a set of scheduled agents and a folder of runbooks" is a truthful answer that satisfies no one. I've started to think the resistance a company of one meets is mostly this: not that it can't work, but that nobody has a form for it yet.
The third is the one I take most seriously. A company that is one person coasts when that person is gone. The systems keep running. The bookkeeping gets swept, the leads get routed, the customers get answered. But nothing gets decided. If I'm out for a month, I come back to a company that is exactly where I left it, which is a strange kind of stability, and not the kind you'd want if the market had moved in the meantime. The honest answer to "what's your succession plan" is that I don't have one. I have a company that pauses.
The org chart was a map of limits
Here's the generalization I keep arriving at. The org chart was never a map of the work. It was a map of human limits. How many people one person can watch. How much one person can know. How many hours there are in a week. Every box on it was there because some person's capacity ran out at that point, and another person had to begin.
When those limits move, and they have moved, the chart doesn't shrink in proportion. It collapses, unevenly, to the parts that were about people rather than about capacity. Judgment. Disagreement. Relationships, because a Norwegian broker still signs with a person and not a system. Taste. The rest, which is most of the boxes, turns out to have been scaffolding for the fact that one person couldn't hold a company.
Now one person can. That's the new fact, and it's the one I'd most want the founders I meet to actually believe, because most of them still don't. But the parts of the chart that weren't scaffolding don't disappear when the scaffolding does. They just stop having anyone assigned to them, and you find out which ones they were by what goes wrong.
I don't miss having a co-founder. I miss being told no by someone who'd have lost money if I was wrong. That was the part of the co-founder nobody put in the guides, and it's the only part I've had to rebuild by hand.
