From Investor to Builder with Claude: Andre Retterath (Earlybird) on AI-Native VC

A new €360 million fund, five General Partners, and a generational transition after almost 30 years. Andre Retterath, General Partner at Earlybird, explains how one of Europe's best-known tech VCs managed the handover to its next generation — and why that's a bigger achievement than any single fund.

One Fund, Five General Partners, One Generational Shift

Generational transitions rarely succeed at European VCs. Earlybird tackled the problem proactively: the new generation of General Partners had already been investing together for 8 to 9 years before the actual handover took place. Not a sudden break, but a long runway.

Central to this is a meritocracy structure. Instead of building dependency on individual people, Earlybird relies on processes that work independently of any single person. That keeps the best investors in the firm and makes it more resilient to the loss of key individuals. Andre puts it simply: building an investment firm across 17 fund generations is a bigger achievement than any single successful fund.

Why False Negatives Are the Most Expensive Mistake in VC

A core theme of the episode: institutionalizing investment knowledge instead of leaving it in individual heads. That means distinguishing between False Positives (you invest and it goes wrong) and False Negatives (you don't invest — and miss the next big case).

Andre explains why False Negatives are the most costly mistake in venture capital. The Lovable example shows when structural rules get in the way instead of helping — and why rigid processes can sometimes block exactly the deals you actually wanted to do.

Cognitive Biases in the Investment Committee

To reduce bias in the decision-making process, Earlybird uses concrete mechanisms:

• –Anonymous post-IC surveys to get honest feedback without social pressure
• –The rule "the youngest person speaks first," so hierarchy doesn't pre-shape the discussion
• –Systematic evaluation of subjective signals like university, employer, and founder history — treated as structured data points rather than gut feeling

Dataset bias is also a topic — illustrated by the UiPath story from Romania, which shows how geographic and cultural blind spots can cause good investments to be overlooked.

The Three-Layer Stack: How Earlybird Invests in the AI Era

Earlybird structures its investments in the AI era across three layers, each with different moats:

1. –Deep tech/hardware — e.g. Isar Aerospace
2. –Software infrastructure/frontier models — e.g. Aleph Alpha and Black Forest Labs
3. –Applications — the layer where AI is currently changing the most

Andre explicitly classifies Aleph Alpha as a successful investment — despite all the public debate around European frontier-model providers.

Application Layer: From 20 to 200 Similar Startups

At the application layer, AI is dramatically lowering the barrier to founding a company. Where Earlybird used to see 20 similar startups in a given segment, it now sees 200. Vibe coding and AI-native teams make it easier to build a product fast — but differentiation becomes harder at the same time.

According to Andre, the ideal founding team combines an AI-native founder with a domain expert. Pure technical excellence is no longer enough when a hundred other teams have access to the same tools.

Fundraising Tips for Founders

One practical piece of advice for founders in a fundraising process: find lead investors first. Only once concrete offers are on the table should you negotiate the round size upward — not the other way around.

Andre also stresses that VC remains, at its core, a people business. The investor-founder relationship is decisive for the success of a multi-year collaboration, no matter how data-driven the selection process was beforehand.

Agency: The Most Important Skill of the Coming Years

The episode closes on a broader note: generalists vs. specialists — and why agency will be the most important skill of the coming years. Agency here means the ability to independently find paths forward and to orchestrate AI agents, rather than waiting for instructions.

Andre's concrete advice: stop consuming news — and just start instead. Get two or three people together, use AI tools, try things out, iterate. Less reading about AI, more building with AI.