Sonnet 4.5 & the AI Plateau Myth — Sholto Douglas (Anthropic)

In this episode of The MAD Podcast, Matt Turck interviews Sholto Douglas, a leading AI researcher at Anthropic, about the release of Claude Sonnet 4.5, now considered the best coding model. They discuss the rapid pace of AI releases, the differences between Anthropic's Opus, Sonnet, and Haiku models, and Sholto's journey to Anthropic. The conversation covers frontier AI, the culture of AI labs, the focus on coding, and the potential of AI agents working for extended periods. Sholto explains the significance of reinforcement learning, test-time compute, and the possibility of achieving AGI through current paradigms, emphasizing the importance of measuring progress and preparing for a world with increased individual leverage.