Today's AI agents are a primitive approximation of what agents are meant to be. True agentic AI requires serious advances in reinforcement learning and complex memory.