Don't Worry About the Vase Podcast On Dwarkesh Patel's 2026 Podcast With Dario Amodei
9 snips
Feb 16, 2026 A rapid-fire tour of AI progress, from how code corpora turbocharge models to debates over continual learning. They weigh compute buying, profitability, and why labs scale cautiously. Security, export controls, and international rivalry come up alongside tensions between fast capabilities and slow real-world impact.
AI Snips
Chapters
Transcript
Episode notes
Fast Scaling Will Produce Powerful Systems
- Dario expects extremely rapid AI capability gains driven by compute, data, training length, and objectives.
- He thinks 'geniuses in a data center' are likely within a few years and verification shapes what gets built.
Conversations Are Good; Tooling Still Limits Automation
- Current LLMs can match well-educated humans in hour-long conversations on many topics, but task automation lags due to tooling gaps.
- Zvi and Dario expect task-level automation to grow rapidly as computer use and tooling improve.
Internal Tools Drive Rapid Revenue Growth
- Internal use of LLMs creates massive hidden productivity gains that often aren't obvious externally.
- Anthropic saw rapid revenue growth linked to internal tooling and agentic coding adoption.