Right now I'm deep in the weeds on large scale agent coordination. As AI agents take on more complex work, keeping them all on task becomes the hard part.

The example that got me going this week: OpenAI deploying 10,000 concurrent agents to propose a solution to the Navier-Stokes Millennium Prize Problem. That's a lot of agents. How do you keep them all pointed at the same goal?

So I've been writing about the techniques we built into Intent, our open-sourced developer tool. A few things make it manageable: focused context (every task runs on its own copy of your files so agents don't overwrite each other's work, with a living spec keeping them coordinated), agent roles (a coordinator breaks work into pieces, implementers write the code, verifiers check it against acceptance criteria), and intelligent handoffs.

That last one I've been especially into: agent succession. A finished agent briefs a fresh one, passing only what matters, then retires itself and its sub-agents to keep context focused and machines tidy.

And it all scales... not just many agents per workspace, but many parallel workspaces across many devices. Intent handles the coordination for you.

Developer workflows are the most mature example of this today, but the underlying approaches (focused context, agent roles, intelligent handoffs) apply to a lot more than writing code. You can try it all out at intentapp.dev.