rob toews

Anthropic's trajectory from a 2019 OpenAI spinout raising at a few billion valuation to a trillion-dollar company demonstrates that exceptional founding teams in 'neo labs' can justify seemingly crazy entry valuations, providing an existence proof for the category.

The firm evaluates nearly every 'neo lab' but declines the vast majority because billion-dollar seed valuations for pre-product companies with unclear business models and technical direction skew risk/reward unfavorably. They only invest when the team is objectively world-class (e.g., Discovery Loop) pursuing a massive market, acknowledging another Anthropic-scale outcome is possible but rare.

Discovery Loop's founding team (Jeff Dean, Oriol Vinyals, Sanjay Ghemawat, Quoc Le) represents arguably the most accomplished AI founding team ever, pursuing recursively self-improving AI systems for scientific discovery — a vast, 'infinite' market compared to finite enterprise AI agents.

The founding team (Jeff Dean, Oriol Vinyals, Sanjay Ghemawat, Quoc Le) is objectively the most accomplished in AI history with top citation rankings in both AI and distributed systems, and they have a track record of shipping products used by billions. Their vision of recursively self-improving AI systems applied to fundamental scientific challenges (climate, health, semiconductors) represents a vast, infinite opportunity set where most long-term value will be created.

Losing four of its most important AI leaders is a blow, but Google's deep talent bench, massive training data, compute scale, full-stack platform (TPUs, cloud), and search cash machine provide durable structural advantages that will keep it successful in the AI race.

Radical Ventures evaluates neo labs rigorously, passing on the vast majority due to skewed risk/reward at billion-dollar entry valuations, but invests when teams are objectively world-class pursuing infinite markets — citing Anthropic's trillion-dollar outcome as existence proof.