Andreessen Horowitz partner Andrew Chen says generative AI is breaking down the long-standing founder templates that have shaped early-stage venture investing in Silicon Valley. For years, investors often gravitated toward familiar patterns: a technical and business co-founder pair, a Stanford computer science PhD, or a star team trained at major tech companies. Chen argues that AI is forcing those assumptions back onto the table.
Speedrun has put founder background at the center of the discussion
Chen wrote that much of his recent time at a16z has gone into Speedrun, the accelerator where the firm writes the first check to brand-new startups, with investments of up to $1 million.
One of the recurring questions inside that work is simple: what kind of founder profile makes sense now, and how does that answer change in the AI era?
He also noted that the current week was the last week to apply for the upcoming San Francisco residency program. Founders accepted into the program can work from the firm’s office, meet with him in person, and start their next company. The application deadline, he wrote, is Sunday, Nov. 1.
Five founder types Chen expects to see more often in the AI era
Chen said he recently discussed with Ben whether AI gives solo founders enough leverage to build very large companies on their own, without the traditional split between business and technical co-founders.
His answer is that AI will materially increase the power of solo builders. Work that once required two co-founders, or even a three-person combination spanning product, engineering, and go-to-market, may now be compressed into one person.
But he also presented the counterargument from Ben: the bar for a V1 product, and the market’s expectations for a seed-stage startup, may rise at the same time. If a co-founding team is also amplified 100x by AI, it may still move faster than a solo founder who gets the same 100x boost. In that case, teamwork may keep its edge.
From there, Chen laid out five founder profiles he expects to become more common, along with the case for and against each one.
1. Solo founders
The argument in favor is straightforward. With AI, solo founders may get a 100x jump in productivity and ship an initial product that would previously have been out of reach for one person.
The case against them is that co-founder teams may still outperform on business execution. Even if agents can write code, someone may still need to run hundreds or thousands of customer meetings. Chen also raises a deeper question: was writing code ever the real bottleneck for startups in the first place?
2. Non-technical founders
Chen points to finance and consulting as fields full of highly capable people who may not code but care deeply about business building. He makes a similar point about people in creative industries who have always been making things, just not in software. AI now gives them a way to build software products directly.
Still, he says the gap between a demo and a stable, reliable, scalable commercial product remains large. If everyone can quickly assemble a first version, then technical judgment may matter even more as products become more complex. That includes deciding what to build, what is technically feasible, and how to stay ahead, especially in categories such as developer tools.
3. Domain experts
Doctors, real estate agents, soldiers, and teachers often understand the pain points in their own work better than anyone else, Chen wrote. If they can use software to solve those problems directly and build a company in the process, they become a meaningful founder category to watch.
But domain expertise does not automatically produce disruption. Traditional hoteliers did not create Airbnb, and traditional finance did not produce Kalshi. In some cases, an outsider’s perspective is what resets an industry, rather than an insider making incremental improvements inside the old frame.
4. Kids, high schoolers, and college students
Chen argues that younger people are starting to build products at earlier ages. Making a Minecraft mod, a Roblox game, or a Discord bot can quietly teach someone how to create something used by thousands of people.
With AI added to that process, a young person with a strong idea may be able to build something larger, earlier, and more efficiently than before.
He also draws a clear line between building something friends enjoy and running an actual company. Hiring, fundraising, managing people, and figuring out monetization do not go away. Young founders may have a natural edge in consumer products because they understand their peers, but that advantage weakens in enterprise software, healthcare, and other sectors where industry knowledge and business relationships matter more.
5. Content creators and influencers
Chen says creators already spend their time making things. The same creative energy that goes into video can also go into software. Just as important, they come with built-in distribution. They know what their audience cares about, how to capture attention, and how to reach large numbers of people without paid acquisition.
In the past, turning an audience into a software product usually meant finding a technical partner or hiring engineers. Now creators can build and launch products themselves, then put them in front of millions of potential users.
That said, an initial traffic spike from followers is not the same as a durable growth engine. The most valuable software products are not short-lived meme hits. They are products that require steady refinement and repeated iteration over time. Chen questions whether the creator mindset always fits that kind of long run.
The founder pool may widen sharply, even if failure rates stay high
Chen says the range of founder profiles entering startup creation will almost certainly expand sharply, and he sees that as a positive development. He also says high failure rates are unavoidable, though some of these founders will still break through and succeed at a very high level.
He contrasts different strengths and weaknesses. Teenagers and students may have strong instinct and cultural sensitivity but limited industry depth. A dentist building AI software for dental diagnosis may have deep experience but less of the outsider perspective that can break old assumptions. In Chen’s view, every founder profile comes with trade-offs, and a broader, more inclusive founder base is itself a meaningful shift.
He ties that point to a larger pattern: in any era, the founder types that stand out usually do so because they understand the technical shift and the distribution advantage unique to that moment.
How venture preferences form and why they change
Chen revisits the founder archetypes that venture firms have historically embraced in different waves of technology.
- After Google’s rise, investors became fixated on Stanford PhDs.
- In the Facebook era, the college dropout building a Web 2.0 social app for friends became a familiar success story.
- In the Airbnb era, design-driven builders with strong taste looked like the right people to create elegant mobile experiences.
- In the Uber era, hard-driving founders with strong offline operational ability fit the on-demand mobile services model.
- Now, in the AI infrastructure cycle, the pendulum appears to have swung back toward computer science PhDs.
Against that backdrop, he says venture has also maintained a long list of founder types investors are often told to avoid: solo founders, non-technical founders, industry veterans trying to build software in their own field, overseas cross-border founders, married couples, and people such as dentists, influencers, or celebrities trying to start tech companies.
Chen’s question is whether a general-purpose technology as disruptive as AI should reshuffle some of those rules. If some of the old biases were never as solid as they looked, and AI now gives previously disadvantaged founders a real way to build software, then those assumptions deserve another look.
Why founder profiles become self-reinforcing
Chen argues that founder profiles matter on several levels.
The most obvious one is pattern matching. If investors believe social apps, two-sided marketplaces, on-demand services, or AI are venture-scale categories, they look at the companies leading those markets, attribute part of the success to the founders’ backgrounds, and then search for more people who look similar.
A less obvious layer is that founder opportunity is unlocked by the underlying technology cycle and the distribution channels of the time. In the Web 2.0 period, viral email invites were a core growth tactic, and college students with campus email networks were naturally well positioned. If App Store editorial featuring was a key acquisition channel, then design-oriented developers with taste and relationships had an advantage. When the iPhone unlocked GPS and mapping, it created the infrastructure that made on-demand platforms possible.
There is also a more hidden mechanism. If investors broadly agree that successful founders in a given era should look a certain way, then backing those founders feels safer. Those companies may find it easier to attract follow-on capital, and even if business progress lags for a time, they may still receive more generous financing support than others.
Chen says this self-reinforcing effect matters especially in seed and pre-seed rounds, where many small checks often need to come together to complete a financing.
What may endure and what may fade
For Chen, the real question is not which of the new founder types will win outright. It is which old assumptions about founders reflect durable truths, and which were only temporary byproducts of older technical limits and distribution constraints.
Some rules, he says, are likely to remain. Coding may get easier, but sales, recruiting top talent, strategic judgment, and resilience do not become easier with it. Deep industry experience can help founders identify real pain points, but it can also trap them inside the logic of the existing system. A solo founder may be able to replace the output of 10 people, but a tightly aligned three-person team, with leverage, may still produce the force of 30.
At the same time, some traditional rules are likely to age out. A founder who looked mismatched or low quality five years ago may now hold a real competitive advantage.
Chen ends with a warning against replacing one rigid pattern-matching system with another. The mistake is not only clinging to the old model, but also declaring that solo founders now win, or domain experts now win, or creators now win.
His core point is narrower and more structural: when the underlying technology of productivity shifts this dramatically, the founder profiles capable of breaking out should evolve with it.
