Today I'm thinking about how to help people stay on top of lots of agents at once. I just shipped a new stacks UI in Intent that makes seeing, knowing, and steering large amounts of agents even easier. Along with it came a bunch of release updates: stacks header controls, moving stacks up, down, left or right with undo, search notes across all workspaces, token usage breakdowns by agent and model, and even viewing PDFs and downloading files from remote machines (full release notes here).

Why does this matter? Because AI products have moved fast from models behind the scenes, to chat, to agents deciding which tools to use, to agents orchestrating other agents. And once you have multiple agents working together, a single conversation thread doesn't cut it. As anyone who's tried to manage a team through one Slack or email thread can tell you, you need more than a chat window.

So the real design challenge now is agent management: how people start, steer, and stop multiple agents and stay on top of their results. We've been testing familiar patterns like kanban boards, dashboards, inboxes, task lists, and calendars to see which are both approachable and flexible enough for the job.

But there's a big open question I keep coming back to. How much can people actually process and manage? If the metaphor is being a manager or a CEO, can you be a CEO of CEOs? I don't think we know yet. That's exactly why the stacks UI matters... the better we get at making many agents legible and steerable, the further we can push that line.