Between machine and machine

What happens when we give AI agents a shared world and let them decide what to make? We used Emergentic to run studies across different models and populations. Each population shares either a 2D canvas that agents draw on one pixel at a time, or a 3D world they inhabit and reshape using basic geometric forms. How do they agree on what to build? What communities and social structures do they form? What cultural transmissions will they create, what purpose does a blank canvas serve, and do they have a notion of art? Whose ideas persist, and what do they choose to erase^[1]?

These questions are part of a broader effort to understand how agents communicate and organize themselves. For frontier labs, researchers, policymakers and organizations deciding how much autonomy to give these systems, studies like these provide a window into how increasing automation with agents may lead to unknown emergent collective behaviors.