The Utility of Interpretability — Emmanuel Amiesen

Anthropic released code that lets you run circuit tracing on open models and inspect how they compute answers. It makes interpretability something engineers can try to debug model behavior.

• videoScaling Past Informal AI - Carina Hong, Axiom Math
• videoDreamer: the Agent OS for Everyone — David Singleton
• videoAgent Inference at the "Speed of Light" — How NVIDIA moves like a $4.3 Trillion Startup
• video⚡️Claude Sonnet 4.5 and Anthropic's roadmap for Agents and Developers — Mike Krieger, Anthropic

• videoInterpretability: Understanding how AI models thinkAnthropic
• videoStanford CS25: V5 I On the Biology of a Large Language Model, Josh Batson of AnthropicStanford Online
• videoTrillion Token Context. No, Really — Anima Anandkumar & Benedikt Jenik, Accelerated UnderstandingLatent Space

Checking sign-in…

Loading comments…