A guided tour of current research on AI agents: how they are measured on real work, the harnesses, memory and skills built around models, how they learn from experience and improve themselves, where self-improvement goes wrong, and what it takes to act on real devices and real tasks.
Background: Memory for AI agents