Ex-GitHub CEO Thomas Dohmke Raises $60M for a Startup Without a Product

Thomas Dohmke used to be CEO of GitHub. Now he's building Entire, the next developer platform — built for the era of AI agents. The numbers behind it are unusual: a $60 million seed round at a $300 million valuation, a team that grew from 3 to 40 people in under a year, and 400 AI agents effectively multiplying that team's output.

In this conversation, Thomas explains what it actually means to build a startup AI-native — and why that's about far more than just using AI tools.

What building AI-native really means

For Thomas, being AI-native means rethinking every single process through the lens of AI — consistently and without exception. Not for today's requirements, but with an eye on a generation that won't even leave school for another 16 years and will grow up treating AI as a given.

This mindset shows up clearly in how far ahead companies plan. Thomas points out that OpenAI plans no more than eight weeks in advance — product managers who show up with classic six-month roadmaps get sent back. In a field moving this fast, planning certainty over multiple months is simply an illusion.

How fast this principle plays out in practice is visible at Entire itself: the company grew from 3 to 40 employees — and those 40 people work alongside 400 agents. The result is a company that looks significantly bigger from the outside than its headcount would suggest.

Building software with agents

A central point in the episode: the real breakthrough isn't that agents can write code — it's that they use actual developer tools directly on the local codebase. Grep, search, writing tests, fixing bugs. In other words, they work like a developer with a full toolbox, not like a chatbot offering suggestions.

Thomas also describes how so-called RAV Loops work: agents get assigned a task along with a clear evaluation criterion, and they keep looping until that goal is met. This, he argues, is exactly where developer competence is shifting — away from writing code itself, toward correctly defining evaluation criteria. He's convinced that agent-written code today already surpasses the quality of the average developer.

Differentiation: why technology alone isn't enough

A core theme of the episode is that technology today is table stakes — a baseline requirement, not a competitive advantage. What actually matters is brand, positioning, and network effects. Thomas draws on his own past to make the point: pretty much anyone could have written GitHub's code — but the whole world knows the Octocat.

He explains why a product has to be ten times better to break established user habits. As examples, he cites Tesla completely removing the traditional speedometer, and the original iPhone, which didn't even support copy-paste — and still fundamentally changed user behavior. The most dangerous mistake a startup can make, according to Thomas, is being just a clone of an existing product that competes on price alone.

Fundraising and the 10-year vision

How do you close a $60 million seed round for a startup that doesn't even have a finished product yet? Thomas explains that his reputation and a clear vision made the decisive difference. At the same time, he's convinced that AI startups today need at least $10 million just to get started seriously.

He breaks down three budget categories that are central for AI-native startups: people, marketing, and tokens. Token costs in particular are often underestimated — they can add up to millions very quickly.

Toward the end, Thomas describes his approach to Entire as an infinite game. Individual launches, to him, are just minigames within a much bigger story. Winning isn't about the next PR win — it's about still being relevant in ten years.

The full episode offers deep insight into building an AI-native startup — from team structure to the technical architecture behind working with agents to strategic market positioning.