The Plan and the Anti-Plan

EDO SEGAL: We've spent two hours on the danger. I want this round to be about the build, because the reader is not a spectator — they're the man at the screen, the engineer in the meeting, the parent at the table, and they need more than two beautiful warnings. Professor Scott, you proposed something in your later work you called, half-joking, the anarchist squint, and a set of principles for governing without the catastrophe. Lay out the anti-plan. And Mr. Hobbes, I want you to tell him where it's a fantasy.

SCOTT: The anti-plan isn't the absence of planning — that's the caricature and I want to kill it before Hobbes does. It's a different relationship between the plan and the reality it touches. Two Cheers for Anarchism, not three, because I'm not naive about disorder. The principles are almost embarrassingly simple, which is why institutions ignore them. Start small and observe — begin with what the AI is actually doing in this team, this classroom, before you write the comprehensive strategy for all of them. Prefer reversible interventions — organizational norms you can change next month, not a four-hundred-page act that takes years to amend and is obsolete before the ink dries. Build in feedback that measures effects, not compliance — not "did they follow the rule" but "what happened to the quality of the work, the development of the juniors, the wellbeing of the people." And tolerate messiness — let different contexts develop different norms, because the alternative, the one uniform rule applied everywhere, is the spruce plantation. The dam, not the great wall. Many local builders feeling the current, not one planner with a blueprint who's never seen the river.

The deepest principle is the hardest and the one institutions can't do: trust practitioners. Not because they're infallible — local knowledge can be parochial and self-serving, I never pretended otherwise. But because the knowledge that good AI governance needs is distributed among the people who use these tools every day, and governance that can't access it is structurally blind, and no amount of expert brilliance at the top fixes structural blindness. Create channels, not mandates. Give the woodsman a seat at the table where the forest policy is made — as authority, not as a token "local voice" the minister listens to politely before doing what she came to do.

EDO SEGAL: Mr. Hobbes. Where's the fantasy?

HOBBES: Let me first say what is not fantasy, because much of it is wise. Start small, prefer the reversible, measure effects, keep the channel open to those who know — a prudent sovereign does all of this, and the sovereign who does not is the deaf one we have already buried tonight. I accept the principles as counsel. I reject them as a constitution. Here is the fantasy, and it is in the last principle: "trust practitioners," "many local builders," "channels not mandates." This is excellent for the texture of the forest. It is silent on the one thing that matters most — what happens when the local builders disagree, and the disagreement is about something that cannot be left unresolved. Two teams develop two norms; the norms conflict; the conflict touches the peace. Who decides? Your anti-plan has no answer but "more conversation," and some questions cannot wait for conversation, because while you consult the practitioners the defector builds the weapon, the plague crosses the border, the faction arms. The dam works on the stream. It has nothing to say to the flood that does not negotiate. You have written a beautiful theory of peacetime, and offered it to a war.

SCOTT: That's fair, and I'll concede the seam exactly where you've put it: my framework is strongest in peacetime and weakest at the moment of the unresolvable conflict that can't wait. I've always known that's my soft flank. But watch what you've conceded to get there, because it's more than you think. You've just accepted the whole anti-plan as counsel — start small, reversible, feedback, trust the practitioners, keep the channel open. You've accepted that the deaf sovereign dies and the wise one listens from below. The only thing you've reserved is the final decision in the unresolvable case. And I'll give you that — fine, in the genuine emergency that can't wait, someone decides. But here's the discipline I'd put on it, and it's the whole difference between us: the emergency power must be rare, reversible, and accountable to the people it acts on — a fire alarm, not a throne. Your error has always been to take the rare emergency, the moment someone must decide, and make it permanent and absolute and total — to take the fire alarm and build a religion around it. The need for someone to decide in the flood does not license the mortal god to rule the stream. You've justified a sovereign for the worst day and installed it for every day.

HOBBES: [pause] "A sovereign for the worst day, installed for every day." That is the sharpest thing anyone has said against me in three hundred years, and I will not pretend it doesn't reach me. I have always argued that the sovereign's power must be whole and constant, because a power that is sometimes sovereign and sometimes not invites men to test which day it is, and the testing is the road back to war. But you press a real wound: I have never adequately answered why the power adequate to the flood must also rule the stream. My answer has been that you cannot divide it without dissolving it — that a sovereign for emergencies only is no sovereign, because who declares the emergency? But I feel the force of your reply: that to refuse all division is to convert every day into the worst day, and to govern the child's notebook with the power meant for the invading army. Perhaps — and I concede this slowly, because it costs me — perhaps the true question is not whether to have a sovereign but how to bind one to wake only in the flood and sleep in the stream. I have spent my afterlife being told I am the enemy of liberty. I begin to think I am only the enemy of the unbound stream — and that a sovereign bound to the flood alone might be a thing I could love.

SCOTT: [long pause] Then we've moved, both of us, and I'll match you. A power bound to the flood and sleeping in the stream — rare, reversible, accountable — that is not so far from what I'd call a dam large enough to hold without flooding the valleys. I've spent my life refusing the crown. But I could perhaps live beside a sovereign that knew it was a fire alarm and not a god. The crown I won't give. The alarm, in the genuine flood, I might.

EDO SEGAL: [quietly] Stop the room. Mark it — convergence four, and it is the largest of the night, and neither of your publicists is going to be happy. Hobbes has conceded that the power adequate to the flood need not rule the stream — that the sovereign might be bound to wake only in the emergency and sleep over ordinary life. Scott has conceded that in the genuine, unresolvable flood, someone must decide, and a power large enough to hold is not always tyranny. You have not agreed. But you have found the shape of the thing you were both circling for two hours: a power rare, bound, reversible, accountable — woken by the flood, asleep over the notebook. The whole evening was the negotiation of where that line falls. We have one round left before the crossing. I'm going to spend it on the deepest water — what the machine is for, what kind of world it's building underneath all of this, and whether either of you believes a human being can stay human inside it. The smooth, the friction, and the pattern that decides. After this.