In this episode of The MAD Podcast, Matt Turck interviews Boris Cherny, the creator of Claude Code at Anthropic, about the agentic coding AI and its accidental yet rapid success. They discuss the evolution of coding from punch cards to agentic coding, defining Claude Code as a tool where engineers describe desired changes and the model executes them. Boris explains the choice of operating at the CLI level for its universality, the product's design principle of following the model's capabilities, and the concept of agentic search. The conversation covers Claude Code's actions, awareness, memory, the use of sub-agents, the balance between autonomy and human oversight, and the UI/UX considerations. Boris shares use cases, from code-based research to debugging, and provides insights into the future of coding, emphasizing the importance of building for future model capabilities and the empowering nature of AI in coding.
Sign in to continue reading, translating and more.
Continue