a16z partner Andrew Chen published an essay on October 27, 2026, titled “New founder archetypes in the AI age,” in which he argues that AI is paving the way for five new types of startup founders: solo founders, non-technical specialists from business and creative industries, domain experts, children and students, and content creators whose new “content” is becoming apps. The full version of the essay is available as a paid article on Substack.

image
image

What happened

On October 27, 2026, a16z partner Andrew Chen released the essay “New founder archetypes in the AI age,” discussing which types of founders AI will unlock. The announcement post collected 96.2 thousand views, 241 likes, and 79 replies, while the full version was published as a paid article on his Substack. Chen identifies five archetypes. The first is solo founders: in his assessment, AI provides “100x” leverage for one person to ship the first version of a product. The second is non-technical founders from finance, consulting, and creative industries. The third is domain experts, among whom Chen lists doctors, real estate agents, soldiers, and teachers. The fourth is children, school students, and college students. The fifth is content creators, for whom apps become the new “content.” For each archetype, the author analyzes the arguments “for” and “against,” and among the examples cited are Airbnb, Kalshi, Minecraft mods, and Discord bots.

Context

Andrew Chen is one of the most cited partners at a16z; he leads the Speedrun accelerator, where a16z writes the first check of up to $1 million to new startups. The essay challenges the classic venture capital pairing of “technical founder plus business co-founder”: the author believes that AI shifts founder selection criteria toward solo teams, non-technical specialists, and domain experts. The key thesis of the text is the hypothesis that code may never have been the primary bottleneck for startups, and that founder time actually goes to customer meetings, distribution, and hiring. The second pillar of the argumentation is the significant gap between a demo and a scalable product: this aligns with known limitations of agentic coding, where models are strong at generating the first version but weaker on long horizons such as refactoring, testing, migrations, and operations. From this, Chen concludes that the scarce asset in the early cycle is not code, but distribution, domain expertise, and technical judgment—understanding what to build and what is actually possible.

Why this matters for the industry

For the venture industry, this is a signal to reconsider which profiles to look for in the early cycle: funds and accelerators receive a ready-made framework that challenges the necessity of the classic “technical plus business” pair, and a reason to rewrite pitches for solo teams, non-technical founders, and domain experts. The flip side is that as the cost of the first version decreases, the moats of ordinary apps narrow, and early-stage competition intensifies: the demo becomes a mass phenomenon. For builders of devtools and agentic environments, a market opens for “first version in a day” tools tailored to new founder archetypes. In the coming months, a second wave of texts in this genre from funds and the growth of vertical starter kits for specific practices like clinics, real estate businesses, or schools are likely, but real value will come not from new manifestos, but from factual data, such as survival statistics for solo founders across accelerator cohorts like Speedrun.

Why this matters for users

For the reader, this is a practical framework for self-assessment: can their product be built by one person, a specialist without a technical background, or a domain expert like a doctor, real estate agent, or teacher? The essay honestly shows the boundaries: AI covers the creation of the first version, but not scaling, hiring, fundraising, and technical judgment. A verifiable action to take now is to attempt to ship v1 alone using current AI tools and record where the process breaks down: deploying to production, handling failures, observability, and controlling inference costs still require engineering practices that most new builders do not have. A separate skill worth honing is a skepticism filter for bright quantitative claims not supported by measurements: this habit helps to soberly evaluate both this text and the subsequent wave of similar publications.

What is currently unknown / limitations

This is an essay by a venture partner, not a research study. The central claim about “100x” AI leverage for a solo founder is not accompanied by methodology, measurements, or benchmarks, so it remains a rhetorical assessment; the hypothesis that “code may never have been the primary bottleneck” is also not supported by data in the text. There is a structural conflict of interest: the author leads the Speedrun accelerator with a first check of up to $1 million, and the narrative that “non-technical, domain, and solo founders can now” expands the incoming flow to his own program—this does not refute the arguments, but reduces a priori trust without independent data. It is also unknown whether the trend will be confirmed: the fork in the road depends on whether agents will learn to reliably maintain and scale codebases, including tests, migrations, and operations; if not, the essay’s conclusions may turn out to be a premature extrapolation.

Sources

Author

Look at AI, editorial team