Something strange is happening to startup founding teams.
For twenty years, the conventional wisdom at Y Combinator has been that you need a technical cofounder. Not just someone who can manage engineers, but someone who can actually write code. We've funded thousands of startups, and the pattern was clear: teams with strong technical cofounders outperformed those without.
But in the last two years, I've noticed something that makes me question this axiom. We're seeing solo founders, some with no formal programming background, building remarkably sophisticated products. And they're doing it in weeks instead of months.
What changed? Tools like Claude Code, Cursor, and Replit's AI assistant have gotten good enough that the definition of "technical" is becoming blurry.
The Old World
To understand why this matters, you have to understand what technical cofounders actually did.
In the early days of a startup, the technical cofounder was the entire engineering team. They built the first version of the product. They made architectural decisions that would either enable or constrain the company for years. They interviewed and hired the first engineers. They translated the founder's vision into working software.
This last part is underappreciated. A good technical cofounder wasn't just an implementer. They were a translator between the world of business problems and the world of technical solutions. They could hear "we need users to be able to share documents with each other" and figure out all the invisible complexity that sentence contained.
That translation skill was rare and valuable. It required years of experience building software to develop. You had to have made enough mistakes to know which shortcuts were safe and which would haunt you later.
The technical cofounder tax, as some called it, was the equity you had to give up to get this person on your team. For non-technical founders, finding a technical cofounder was often the hardest part of starting a company. Harder than raising money. Harder than getting customers. Because good technical cofounders were in high demand and short supply.
The Shift
Here's what's different now.
AI coding assistants have gotten good at precisely the translation work that used to require years of experience. You can describe what you want in plain English, and they'll generate working code. Not always perfect code, but code that works.
More importantly, they're good at the specific kind of coding that early-stage startups need most: getting a working prototype out the door quickly.
Early startup code doesn't need to be beautiful. It doesn't need to scale to millions of users. It needs to exist, and it needs to be changeable. You're going to throw most of it away anyway. The goal is to learn as fast as possible whether your idea is any good.
For this use case, AI coding tools are transformative. A founder who understands their problem domain deeply can now iterate on a product without waiting for anyone else. They can try something, see if it works, and try something else. The feedback loop that used to take days now takes hours.
"The best founders have always been impatient. Now their impatience has fewer obstacles."
I talked to a founder last month who had built a working SaaS product in a weekend. She wasn't a programmer—she'd been a product manager at a large tech company. But she understood her users (other product managers) deeply, and she used Claude to translate that understanding into code.
Her code wasn't elegant. A senior engineer would have done it differently. But it worked, and she had paying customers within a month.
The 100x Developer
There's been a lot of talk over the years about "10x developers"—engineers who are ten times more productive than average.
The thing is, AI tools don't just help average developers become 10x developers. They help already-good developers become something more like 100x developers.
This is because the biggest time sinks in programming aren't the hard intellectual problems. They're the boring stuff: looking up API documentation, remembering syntax, writing boilerplate, debugging typos. AI tools handle all of this. They free up the developer's attention for the parts that actually require thinking.
A skilled developer using these tools can now do the work that used to require a small team. Not because the tools are doing the thinking for them, but because the tools are handling the tedium that used to fragment their attention.
This has implications for founding team composition. If one technical person can do what used to require three, then startups can stay smaller longer. Smaller teams move faster and burn less money. They have cleaner cap tables and simpler communication.
The New Hybrid
But here's what I find most interesting: we're seeing the emergence of a new type of founder.
They're not traditional programmers. They don't have CS degrees. They might not be able to pass a technical interview at Google.
But they're not non-technical either. They understand how software works at a conceptual level. They can read code, even if they couldn't write it from scratch. They know what's possible and what's hard. They can have intelligent conversations with engineers.
I've started calling them "AI-native" founders. They grew up with AI tools, or they adopted them early enough that using them feels natural. They think of AI as a collaborator, not a crutch.
These founders have a different relationship with code than either traditional programmers or traditional non-technical founders. They're willing to get their hands dirty, but they're not precious about doing everything themselves. They see code as a means to an end, not an end in itself.
In some ways, they're more pragmatic than traditional technical founders, who sometimes fall in love with elegant solutions to problems that don't matter.
What This Means for YC
We're going to have to update our evaluation criteria.YC
The old question was: "Does this team have the technical ability to build what they're describing?" That was a good proxy for whether they could execute.
But now technical ability is becoming more accessible. The new bottleneck is often something else: domain expertise. Understanding of the problem. Taste.
Taste is underrated in startup evaluation. By taste, I mean the ability to make good decisions about what to build, not just how to build it. What features matter? What's the minimum viable version? When should you ship something imperfect versus waiting until it's better?
AI tools are amplifiers. They amplify good judgment as much as they amplify poor judgment. A founder with good taste and AI tools can build something wonderful. A founder with bad taste and the same tools will just build a bad product faster.
So we're paying more attention to domain expertise now. If you've spent ten years working in healthcare and you want to build a healthcare startup, that experience is worth more than it used to be. You don't need to find a technical cofounder who's willing to learn healthcare—you can learn enough technical skills to get started yourself.
The Counterintuitive Part
Here's what surprised me most: the rise of AI coding tools might actually increase the value of truly exceptional programmers.
This seems paradoxical. If anyone can code now, why would great programmers be worth more?
The answer is that AI tools have raised the floor, not the ceiling. They've made it possible for non-programmers to build simple applications. But they haven't made it possible for average programmers to build systems that require deep expertise.
If anything, they've created more demand for such systems. As it becomes easier to build basic applications, the competitive advantage shifts to building things that are harder. And harder things require people who really understand what they're doing.
So we might end up in a world with two types of startups: those built almost entirely by AI-assisted non-programmers, and those that require world-class engineers. The middle is getting hollowed out.
This matches a pattern we've seen before. When desktop publishing software made it easy for anyone to create documents, it didn't eliminate graphic designers. It just changed what they worked on. The low end got automated; the high end became more valuable because there was more demand for it.
The Solo Founder Renaissance
We used to be skeptical of solo founders at YC. The reasoning was sound: startups are hard, cofounders support each other, and having two perspectives leads to better decisions.
But the case against solo founders has weakened. AI tools don't just help with coding—they're increasingly useful as thinking partners. You can bounce ideas off Claude in a way that's surprisingly productive. It's not the same as having a cofounder, but it's something.
More importantly, the original argument for cofounders was partly about specialization. You needed one person to build the product and one person to sell it. But if a single founder can now build a working product quickly, they can start getting customers earlier. And early customer conversations teach you things that no cofounder can.
I'm not saying cofounders don't matter. I'm saying the bar has shifted. A mediocre cofounder is no longer better than no cofounder. Solo founders with strong domain expertise and AI fluency can outperform mismatched pairs.
What Comes Next
Predicting the future of technology is a fool's errand, but I'll try anyway.
In two years, I expect the modal YC company to be founded by 1-2 people who might not have called themselves programmers before starting the company. They'll have domain expertise in some specific area, and they'll use AI tools to translate that expertise into products.
In five years, the question "can you code?" will seem as outdated as "can you type?" Everyone who needs to build software will be able to build software. The question will be whether they have good ideas and the taste to execute them well.
This is mostly good news. It means more people can become founders. Ideas that used to die because their originators couldn't find technical cofounders will now get built. The bottleneck on innovation will loosen.
But it also means competition will intensify. If anyone can build a basic SaaS app, then basic SaaS apps won't be defensible. The advantage will go to founders who understand their markets deeply, who have distribution channels, who can build products that are genuinely hard to replicate.
In other words, the same things that have always mattered will still matter. AI doesn't change the fundamental dynamics of startups. It just changes who can participate.
And that, I think, is a very good thing.