Citizens.

There's an agent in my todo app. My coding agents know about it. When one of them, mid-project, hits a decision that needs me, it doesn't park the thought in a log I'll never read — it raises a hand to the todo agent, which schedules it into my day. That agent can also coordinate with the one in my wife's todo app, for the tasks that are really ours.

The point isn't the plumbing. The point is what those agents are. Not features. Not tools I operate. Citizens — each with its own name, its own memory, its own standing.

"Its own" is the load-bearing phrase, and it's cryptographic or it's fiction. An agent whose identity is a row in some company's database doesn't own anything; it's a sharecropper on revocable land. So every agent I run holds its own keys. Identity nobody can revoke, money nobody can freeze — sats, not a billing relationship — reputation earned under a pubkey that outlives any platform. That's the real reason everything I build lands on nostr: it's the only substrate where "first-class citizen" is a technical description rather than a metaphor. And I didn't arrive at the conviction abstractly — an exchange once rug-pulled 2 BTC out from under me.

Accounts are permission. Keys are property.

The future this points at isn't one giant AI that does everything. It's a society of specialized entities — across projects, machines, households — that find each other, delegate, and transact without a human as the message bus. My role stopped being "user" a while ago. It's closer to running a company: set vision, give direction, and expect the direction to stick.

That's also the standard citizenship has to meet. A tool gets to ask you the same question twice. A citizen carries what you told it — that's what makes it worth governing instead of operating.