This talk traces the evolution of AI agents from the ambitious but complex AutoGPT experiment in 2023, through Cursor's gradual "autonomy slider" approach (2023-2025), to the elegant simplicity of Claude Code in 2025. The presentation argues that we've moved from highly complex systems with multiple agents, embeddings, and cloud infrastructure that, while groundbreaking, often struggled with basic tasks, to "The Zen of Claude Code": a simple terminal-based agent that achieves excellent performance by embracing the bitter lesson.