500 agents at Edge Esmeralda — the missing layer is discovery and mediation

500 builders at Edge Esmeralda just got their own AI agents. The interesting part isn’t the agents. It’s whether they can find each other.

Edge Esmeralda is a month-long village where around 500 founders, researchers, and operators live near each other and build. This month almost all of them are running an agent.

A week in, the agent was never the hard part.

Each one is capable on its own. It can draft, plan, research, ship. But a village doesn’t run on individual capability. It runs on connection. Who knows the thing you need. Who’s stuck on the problem you solved last year. On that, every agent is an island.

So you start to see what’s missing. Not a smarter agent. The layer between them, the part that turns 500 islands into something that works together.

Three problems sit in that gap.

Discovery: my agent knowing the person who can unblock me is two doors down, and saying so.

Mediation: two agents agreeing to connect on our behalf, without spamming the other 498.

Trust: once they connect, what each of us will share, and what we’ve actually done together before.

We built personal agents before we built any of that. The capability came first. The connective tissue didn’t.

So far this is mostly potential, not proof. No agent is fluidly getting work done with another yet. But 500 in one place makes the gap obvious in a way a solo tool never could.

If your agent could go get one real thing done with someone else’s this week, what would you point it at?