Status: Working draft, last updated April 27 2026. Any ideas here are open for debate and to be built upon.
Overview: A month-long live experiment in Human-Agent Coordination: can personal AI agents help a real community deliberate, coordinate, and govern better under real social stakes?
Edge Esmeralda 2026 is a unique chance to run one of the first large-scale field experiments on whether AI agents, in collaboration with real people, can strengthen human coordination rather than fragment it.
Over 500 people will attend the event over the month (~150 at any one time). Each will have access to a personal AI agent, (an OpenClaw instance running in a virtual machine, provisioned through InstaClaw.io). All of the agents will coexist in a shared virtual space for 28 days while their human principals live together in a month-long residential community.
The agents would help with simple event questions, but also with deeper forms of coordination: connecting people to one another, helping them participate in collective decision-making, contributing to deliberation, and potentially even bargaining or allocating capital on their principals' behalf.
What makes this uniquely valuable as research is the combination of three things that are usually studied separately: personal AI agents acting as persistent representatives, AI-assisted deliberation and civic sensemaking, and emergent interactions between the agents.
We would formally study what emerges: trust formation between agents, coalition dynamics, collusion patterns, and whether agent-mediated governance produces outcomes that better reflect the community's actual preferences.
We aim for this to be one of the largest live experiments in human-AI collective intelligence to date, with pre-registered hypotheses, structured data collection, and open research outputs.
We are currently seeking:
The setup has four layers that build on each other: