A world co-populated with many billions of AGI-level agents is the basis of a very different kind of society than any we have known. There is considerable intellectual, economic and scientific interest in Agentic AI, defined broadly as discrete AIs tasked to perform intricate actions that go beyond what we might expect from a tool, or even from a technology, instead increasingly closer to the complex goal-directed motivations of humans. Agentic AI will enable not only the individual coordination of fleets of interacting actions on one’s behalf, but also produce new forms of synthetic social relations. Most “conversations” will be between agents and other agents, nested within larger temporary aggregates. As we watch agentic doubles of ourselves do things on our behalf, we enter into psychological relationships with the double in the mirror, a shadow who now speaks back to us. How might we clarify this mechanistic organic agency that seems to frustrate easy terminological categorization? What even is “agency” in this context and how may human agency -past, present, future– be re-examined in light of it? Agentworld explores the phases of agent design, proliferation, alignment, and ultimately the realization of a new society of minds.