Last Updated on July 7, 2026 by Taya Ziv
For most of my career, the startup world couldn’t agree on almost anything. Raise early or stay lean. Build first or sell first. Move to the Valley or stay home. Pick a fight about any of it and you’ll find smart people screaming on both sides.
But there was one rule nobody broke. One commandment everyone repeated to every first-timer who would listen: don’t do this alone. Get a co-founder. Solo founding is how startups die. Paul Graham said it. Every accelerator said it. I have said it, to founders sitting across from me, more times than I can count.
That rule just cracked, and it cracked at the one place that has the data to know better than the rest of us.
The most quoted advice in startups just lost its grip
Ollie Forsyth, who pulls apart every Y Combinator batch with actual numbers, looked at YC’s Spring 2026 cohort and found that 19% of the companies have a single founder. Nearly one in five. The median founding team in the whole batch is now just three people.
To feel how strange that is, you have to know where it came from. The Summer 2024 batch was around 6% solo. YC’s long-run historical average sits near 10%. So in roughly eighteen months, the share of solo founders at the most selective accelerator on earth tripled. This is the same YC that for two decades treated “no co-founder” as close to an automatic rejection, the same partners who would tell you flatly that the bus factor of one was a deal-breaker. They did not write a blog post reversing themselves. They just quietly started funding the people they used to wave away.
And it isn’t a fluke of one weird batch. The thing pushing it is sitting right there in the rest of the numbers: 60% of the one-liners in that same batch mention AI or agents. The founders aren’t shipping features anymore. They’re shipping software that does the work a junior employee, or a second founder, used to do.
What a co-founder was actually for
Here’s where most takes on this go shallow. They say “AI replaced the co-founder,” high-five, and move on. That’s lazy, and it misses the more useful point.
A co-founder was never one thing. It was three things wearing one hoodie.
The first was raw labor. Two people ship more than one. While you sell, they build. While they sleep, you answer support. For a company with no money and everything to prove, a second pair of hands was simply more hours in the day, and more hours was survival.
The second was the second brain. Someone to tell you the idea is wrong before the market does it for you, louder and more expensively. Someone to argue with at 2am about whether the pricing makes sense. You cannot do that with a mirror.
The third was the signal. Investors read “two technical founders who’ve known each other ten years” as proof that at least one other competent human looked at this and decided to bet their twenties on it. A solo founder set off a quiet alarm: if it’s so good, why is nobody else in?
For twenty years those three things came bundled, so we treated them as a single rule. Get a co-founder. AI didn’t delete the rule. It pulled the first thread out, the labor, and the whole sweater started to unravel. When one person with the right tools can do the building of a small team, the pure-hours argument for a second founder mostly evaporates. We watched this same shift hollow out the org chart when the people who used to coordinate the work, the middle managers, became the first layer AI quietly made optional. What happened to companies last year is happening to founding teams this year.
The investor signal is bending too, because investors can count. When a single person with a laptop can now stand up a real product that used to need a funded team, “why are you alone?” stops sounding like a red flag and starts sounding like a normal Tuesday. And it shows up in what gets funded: the last batch was full of small, unglamorous teams winning by doing one narrow thing well, not by fielding a roster.
The take: you were probably chasing a co-founder for the wrong reason
So here’s what I actually think, and I’ll admit upfront I have a bias, because I built my whole working life around a partnership and I’d do it again.
The reason the co-founder rule is breaking is that most people only ever wanted one for the labor. They wanted hands. They wanted someone to take half the to-do list. And that reason, the most common reason, is exactly the one AI just made obsolete. So when I see a founder out there speed-dating for a co-founder because the playbook says they need one, I want to grab them by the shoulders. You are solving a problem you no longer have, and you’re paying for it in the most expensive currency there is, which is half your company.
Because a co-founder is not a hire. A co-founder is a marriage with a cap table. My partner Eytan and I do an actual sit-down every year that we half-jokingly call marriage counseling, where we talk about meaning and resentment and money before we talk about a single KPI. That is the real cost of the rule nobody put in the blog posts. A bad co-founder isn’t a neutral, it’s a wound that compounds, and it shows up right when the company is most fragile.
So the honest 2026 question is not “solo or duo.” It’s “which of the three things do I actually need?” If you need labor, you have tools for that now, and they don’t get a vote on your exit. If you need the second brain and the shared weight, no model gives you that, and going solo to dodge a bad partner is smart while going solo to dodge people entirely is just loneliness with extra steps.
Where I might be wrong: it’s early, and a Demo Day batch is a snapshot, not a scoreboard. Nobody has shown me five years of returns proving solo founders backed by AI actually outlast the duos. The bus factor of one is still real. If the solo founder gets sick, or burns out, or just loses the thread, there’s no one holding the other end of the rope. I’m describing a door that opened, not a guarantee of what’s on the other side.
What to actually do with this
If you’re tempted to go solo, don’t read this as permission to be a hermit. Read it as permission to stop panic-marrying the first technical person who’ll have you. Replace the labor with tools, genuinely, and put real money and real hours into it. Then go build the bench AI can’t be: a small group of founders at your stage you talk to every week, two or three advisors who will actually pick up the phone, and at least one person in your life allowed to tell you you’re being an idiot. That covers the second brain and most of the weight, without splitting your company in half.
And if you do want a co-founder, good, want one for the right reason. Not because a list told you to, and not because doing it alone is scary. Want one because there’s a specific person whose judgment makes yours sharper and whose presence in the room you’d choose even if they cost you nothing. Then, before you sign anything, do the unglamorous thing almost nobody does: work together on something hard and real for a few months first. Date before you marry. The equity split is the easy part. The 2am argument is the whole job.
The rule that ran startups for twenty years didn’t die because solo founders got better. It died because the cheapest reason to take a partner stopped being a reason at all. What’s left is the expensive reason, the human one. And that one was always the only one that mattered.
