Big Technology Podcast Anthropic Chief Product Officer: Why AI Model Development Is Accelerating — With Mike Krieger
353 snips
Oct 8, 2025 
Mike Krieger, Chief Product Officer at Anthropic and co-founder of Instagram, discusses the rapid advancements in AI with the recent launch of Sonnet 4.5. He explains how customer feedback has accelerated development and shares insights on agents, which are proactive collaborators rather than mere tools. Krieger emphasizes the importance of memory in AI models for personalization. The conversation also touches on Anthropic's product philosophy of augmentation over automation, and the strategic partnerships necessary for enterprise AI adoption.
AI Snips
Chapters
Transcript
Episode notes
Customer Feedback Enabled Faster Releases
- Faster releases came from tighter customer feedback loops and streamlined release operations.
- Mike Krieger credits user-driven urgency and improved rollout processes for accelerating model cadence.
Scale Gains Are Engineering-Heavy
- Scale helps but running large training jobs reliably is primarily an engineering challenge.
- Krieger says algorithmic ideas and engineering to maximize compute use must come together to deliver model gains.
Claude Acting Like A Coworker
- Anthropic uses Claude as an active collaborator inside the company, not just a coding autocomplete.
- Internal tools like "Claude on call" let the model triage incidents and dive into data in Slack channels.