Mike Krieger, Instagram CoFounder & Anthropic CPO: Where Will Value Be Created in an AI World?|E1265

· ai · Source ↗

Watch on YouTube ↗ Summary based on the YouTube transcript and episode description.

Anthropic CPO Mike Krieger admits the company underinvested in first-party products and says AI is still day one for most workers’ daily lives.

  • Novartis cut clinical trial report time from 15 weeks to 20 minutes using Claude — cited as a concrete near-term AI productivity step-change.
  • Krieger’s biggest mind-change in 12 months: first-party products matter as much or more than the API; he says Anthropic was too slow to act on this.
  • Claude Code found a tool-use flaw within one week of internal deployment; the fix shipped directly in Claude 3.7 Sonnet — first-party dogfooding as model R&D.
  • DeepSeek’s real lesson was product velocity and narrative, not architecture; Krieger wishes Anthropic had shipped chain-of-thought UI first, calling it a missed novelty moment.
  • Models diverge in character over time rather than converge; Anthropic deliberately targeted coding via reinforcement learning after seeing traction, not by accident.
  • AI is still day one for most workers — Krieger says it is not yet indispensable to the daily work of most people, and staying power will go to whoever closes that gap.
  • Agent-to-agent privacy and discernment — agents holding sensitive context but lacking judgment about what to reveal — is the most underappreciated and under-researched coming risk.
  • Cursor’s breakthrough came from years of iteration before the right model unlocked the product; startups should beat against current model limits rather than wait for perfection.

2025-03-03 · Watch on YouTube