Minus One Anthropic CTO on AI Safety in the Age of Exponential Growth | Rahul Patil
Apr 9, 2026
Rahul Patil, CTO of Anthropic and veteran engineering leader focused on building dependable, safe AI infrastructure. He talks about prioritizing safety over speed, designing trustworthy AI and infrastructure, tradeoffs between scale and responsibility, and why culture and mission matter when chasing exponential impact.
AI Snips
Chapters
Transcript
Episode notes
Work Where Durable Exponentials Are Happening
- Exponentials warp thinking: working in areas with durable exponential growth (CPUs, memory, data) changes what success looks like.
- Rahul picks domains he can 'be part of' rather than let the exponential just happen to him.
Make The Fastest Path The Safest Path
- Build systems where the fastest path is also the safest by automating tests, deploys and guardrails.
- At Stripe Rahul made 'touchless deploy' the default so 80% of deploys were automatic, fast and safe.
Run Teams Like Scientific Experiments
- Treat teams like scientists: start experiments with explicit hypotheses and evaluate outcomes without attaching identity to single results.
- Rahul asks daily: was I lucky or was I good? to separate luck from skill during growth.