AI Dev 26 x SF | Carter Rabasa: File Systems Are the New Primitive for AI Agents

At Carter Rabasa's session during AI Dev 26 x San Francisco, attendees learned why LLMs are already highly effective at working with file systems — thanks to decades of training on code, operating systems, and file-based workflows — and how to leverage that intuition in agent design. The talk explored how file systems provide a powerful foundation for long-term memory and state, enabling agents to persist, organize, and revisit work far more reliably than prompt-based approaches. It also showed how file systems act as a universal interface for data interoperability and human-in-the-loop collaboration, making them a natural layer for multi-agent and human-agent workflows.