Jesse Hu

This talk draws parallels between building self-driving cars and developing AI agents, emphasizing that agents, like robots, operate in complex environments and require more than just sophisticated models. The core thesis is that the 99% of work beyond the AI model—encompassing infrastructure, tooling, feedback loops, and robust offline stacks—is crucial for reliable agent development and deployment.