The MAD Podcast with Matt Turck Sonnet 4.5 & the AI Plateau Myth — Sholto Douglas (Anthropic)
415 snips
Oct 2, 2025 
Sholto Douglas, AI researcher at Anthropic and former Google engineer, delves into the innovations of Claude Sonnet 4.5, claiming we're mere years away from AI matching human capabilities. He explains how reinforcement learning has suddenly made a breakthrough and how agents can maintain coherence during long coding sessions. Sholto also discusses the cultural differences across major AI labs, Anthropic's focused approach to coding, and the profound implications of AI's upcoming exponential progress, especially in economics and robotics.
AI Snips
Chapters
Transcript
Episode notes
Two-Paradigm Progress And Compute Cycle
- Progress accelerated because research now updates along multiple frontiers: pretraining and reinforcement learning.
- Increased compute availability and commissioning lead times mean releases come faster as infrastructure scales.
Fencing Taught Deliberate Practice
- Sholto credits fencing and a world-class coach for learning deliberate practice and mentorship.
- He reached top-50 globally, which taught him how mentorship and repetition produce mastery.
Showcase Work With Deep Technical Posts
- Do publish world-class independent artifacts like deep technical blog posts to signal ability.
- High-quality guides (e.g., CUDA kernel optimizations) can open doors to top research roles.