Don't Worry About the Vase Podcast 2025 State of AI Report and Predictions
Oct 10, 2025
The podcast dives into the fascinating 2025 State of AI Report, revealing insights about Quinn's rise in fine-tuning and the rapidly evolving open-model landscape. Discussions include advancements in robotics and action planning, plus the implications of Anthropics' new Model Context Protocol. The hosts critique the diminishing value of benchmarks and explore the timing of releases among top labs. Predictions cover everything from the impact of AI agent checkouts on retail to the potential of deepfakes triggering NATO debates.
AI Snips
Chapters
Transcript
Episode notes
Expect Rapid Model Turnover
- Open-source models rise and fall quickly, so lock-in fears are overrated.
- Zvi warns that in the open-model world people will switch on a dime, so plan for churn.
Models Connect To Tools And The Physical World
- Chain-of-action planning connects reasoning to robots and physical tasks.
- Model Context Protocol is becoming the standard to integrate models with tools, despite security risks.
Benchmarks Mislead On Real Leadership
- Benchmarks increasingly misrepresent real capability and competitive dynamics.
- Leadership on leaderboards shifts rapidly and timing releases matters as much as model quality.