Let's Talk About FOMAT: Fear of Missing Agent Time — Michael Richman, Cmd+Ctrl

Naming the feeling: Fear of Missing Agent Time

Michael Richmond opens with a joke that lands because it’s true: everyone knows FOMO, and now there’s FOMAT — fear of missing agent time. He describes the exact scenarios every heavy AI coder has felt: you’re on a walk and want to kick off an agent task, or you leave your desk for 30 minutes only to discover the agent got blocked after 2 and sat there waiting the whole time.

Why this gets worse as agents get better

Right now, he says, coding tasks often run for 5 to 45 minutes, so you can kind of guess when to check in. But if tasks soon run for 5 hours or 5 days, “check back in a bit” stops working entirely — you need to know the instant an agent needs you or finishes, wherever you are.

The mess on his screen that led him to build Cmd+Ctrl

Richmond, who leads engineering teams at Bitly and co-leads its AI coding tools strategy, says he built Command and Control because nothing else solved his own workflow problem yet. His setup is chaos in the familiar modern way: multiple terminal windows, multiple tabs, Claude Code here, Codex there, Gemini, Cursor — and beyond two or three sessions, he loses track of what’s happening.

The phone demo: agents that can actually reach you

He shows Cmd+Ctrl on an iPhone app, with sessions grouped by notifications, “on my radar,” and recents. In the demo, he sends Claude Code a simple prompt, leaves the app, gets a push notification when it finishes, taps back in, and resumes the same session — the point being not the prompt itself, but the fact that the terminal no longer owns the interaction.

Starting sessions from bed, not just from your IDE

The second demo shows him launching a brand-new Claude Code session from the mobile UI, choosing a directory and asking “what time is it?” He says, almost sheepishly, that he often starts his day by kicking off prompts from bed, then later resumes those same sessions in the CLI or keeps working from his phone.

Session overload and the stand-up dashboard

The third demo tackles the management problem directly: thousands of sessions, organized into push-notification subscriptions, “on my radar,” recent, and everything else. He highlights an overview dashboard that turns the latest messages into stand-up-style summaries, which is his way of making agent sprawl feel legible instead of exhausting.

Under the hood: one control plane for every agent

Richmond sketches the architecture quickly: each agent platform runs a Cmd+Ctrl daemon that tracks lifecycle events like “blocked” or “needs help,” sends them to a control plane, and exposes everything through a shared API and UI. The key idea is a true single pane of glass across tools and machines — Claude Code on a Mac, Codex CLI on a cloud VM, all visible together — and he notes the daemon layer is open source.

From coding flow to agent choreography

He ends on a broader claim: agentic coding has changed not just software development, but the texture of the work itself. The old flow state was one person locked into code; the new flow is “agent choreography,” moving between multiple agents, unblocking and redirecting them — and, crucially, building systems that let you step away, rest your brain, and still catch the moment when your agents need you.