Inside Claude Code From the Engineers Who Built It

In this episode of the podcast, Dan Shipper interviews Cat Wu and Boris Cherny, the creators of Claude Code, about their journey in developing the agent SDK. They discuss the unintentional evolution of Claude Code, emphasizing the importance of building tools that solve personal problems and the surprising effectiveness of giving AI models direct access to Bash. The conversation explores the team's roles, the concept of "ant-fooding" (dogfooding), and the dual-use design of tools for both engineers and models. They delve into practical applications like code review, feature development, and managing tasks with sub-agents, as well as the shift towards using Claude Code as a general-purpose agent SDK beyond coding. The discussion also covers strategies for managing AI-driven development, balancing extensibility with simplicity, and the future of agentic workflows, including the potential for proactive, always-running agents.