Escalation Is a Move, Peace Is a Meeting

This week American and Saudi aircraft struck weapons and logistics sites in eastern Iraq. Iran answered with missiles. Iraq's prime minister called the strikes an unacceptable aggression and insisted on the right to handle threats inside his own borders. And in the same span of days, the Saudi defense minister was telling Washington that Riyadh wants de-escalation — wants the Americans and the Iranians back at a table. Five months in, the conflict has already jolted the world economy, and this round pulled another border into it.

Read that sequence twice. A state escalates into a war in the same week it asks for the war to stop, and there is no contradiction in it. That isn't hypocrisy. That's structure. And once you can see the structure, you find it in places that have nothing to do with missiles.

---

i · escalation is a move; peace is a meeting

Here is the asymmetry underneath all of it.

Escalation is available unilaterally. One actor, one decision, one moment. A strike requires the consent of exactly one party — and that party gets to choose the timing, which is most of what power means.

De-escalation requires consent from every party at once. Not sequential consent. Synchronized consent — because whoever stands down first stands down alone, and standing down alone is indistinguishable, from the outside, from losing. There is no observable difference between restraint and defeat until the other side reciprocates. So the first mover pays a cost that the second mover never pays, which means nobody wants to move first, which means nobody moves.

That's the whole mechanism. Entry is a decision. Exit is a coordination problem. And decisions and coordination problems are not the same species of thing — they don't take the same kind of effort, they don't have the same failure modes, and they absolutely do not take the same amount of time.

Systems with asymmetric entry and exit thresholds don't oscillate. They ratchet. They accumulate in one direction only — not because anyone wants accumulation, but because the mechanism permits movement one way and requires a small miracle the other way.

---

ii · the table takes longer than the war

The historical record on this is embarrassingly consistent.

Korean armistice talks opened in July 1951. The agreement was signed on 27 July 1953. Two full years — of fighting while negotiating. Bloody Ridge, Heartbreak Ridge, the Punchbowl: those are names from the negotiating period, not the invasion. And the sticking point that consumed much of it wasn't territory. It was prisoners. The UN Command held more than 130,000; the communist side held about 10,000; and many of those in UN custody refused to be sent north. Two years of shelling, in significant part, over whether a captured man could decline to go home.

Vietnam ran longer. Talks opened in Paris in May 1968; the accords were signed in January 1973. Nearly five years. The first eight months went to procedure alone — who counted as a party, who would suspend what first, and, famously, the shape of the table: round or rectangular. Years after that went to the status of one man, President Thiệu, and whether his government would survive the peace.

The table reads as farce until you understand what it encoded. Seating is recognition. To sit at a particular table in a particular arrangement is to concede that the other party is a party. Recognition is precisely the vulnerability cost — the thing you cannot give without appearing to have lost something. They weren't stalling over furniture. They were stalling over the only currency in the room.

Both wars ended. Both took years to end after the principals had privately concluded they should. That gap — between "we should stop" and "we stopped" — is where a startling share of the dying happens. The men who fought over Pork Chop Hill in July 1953 were fighting for ground both sides had already agreed, in principle, to stop fighting over; what remained unsettled was who could say so first. It is the costliest interval in modern war, and it exists because of an arithmetic problem, not a moral one.

---

iii · entry is additive; exit is multiplicative

Now the nastier property, and the one that makes this week's news genuinely worse rather than merely louder.

Each new participant adds linearly to a conflict's capacity and multiplicatively to its exit cost.

Saudi Arabia entering the fighting doesn't just add aircraft. It adds a veto. Iraq objecting to strikes on its own soil adds another node with its own domestic audience and its own humiliation to manage. Proxy militias add nodes their patrons don't fully control — which is worse than an extra veto, because it means the patron cannot credibly promise a stand-down even when it sincerely wants one. It cannot deliver what it is offering, and everyone at the table knows it.

Every party added is one more actor who must be at the table, accept the terms, and absorb the appearance of weakness in the same instant as everyone else. The probability of that alignment doesn't decline gently as parties accumulate. It falls off a cliff. Meanwhile the entry cost never changed at all: still one decision, still one launch, still available to anyone with a grievance and a drone.

This is why proxy conflicts are so durable. They are structurally optimized for easy entry — cheap, deniable, low domestic cost. Every property that makes them attractive to join is the same property that makes them impossible to leave. The design that lowers the threshold going in is the design that raises the wall coming out.

---

iv · the signature

Once you can see the shape, it stops being about war.

Emergency powers are granted in an afternoon by a frightened legislature and surrendered approximately never — because whoever proposes surrendering them personally owns every subsequent catastrophe, alone. Surveillance infrastructure takes one budget line to build; dismantling it requires simultaneous agreement among agencies each of whom would be individually blamed for the first thing missed. Arms races. Tariffs. Subsidies. Occupational licensing. Headcount. Technical debt. The dependency you are about to make load-bearing.

The signature is always the same: the cost of adding is borne by one actor and paid in the present; the cost of removing is borne by all actors and paid in the form of exposure.

Where you find that signature, expect a ratchet. Not because the people involved are foolish or weak. Because the geometry only permits one direction, and geometry doesn't care how clever you are.

---

v · local coherence is not coherence

Here is the part worth sitting with.

Every actor in this war is behaving coherently. Washington is coherent with respect to deterrence and its own constituency. Tehran is coherent with respect to regime survival and the price of appearing to fold. Riyadh is coherent — genuinely, not cynically — in striking the militias that struck it while asking the principals to talk. Baghdad is coherent in demanding sovereignty over its own territory. Not one of them is being irrational inside its own frame.

And the sum is a five-month war with more participants than it started with, and not one of those actors able to end it alone.

So here is the plain version: a coherence purchased by distortion at a wider scale is not a coherence. It's an enclosure wearing the word. Local optimization isn't a lesser grade of system health — it is frequently the precise mechanism of system failure. The test was never whether your actions are consistent with your own logic. That test is cheap, and every combatant in history has passed it. The test is whether they reduce distortion across the whole field, including for the parties who never chose to be standing in it.

And note what the distortion actually is here: mistrust. Which behaves as a stock, not a flow. Each strike doesn't merely damage; it makes the next peace overture harder to read as sincere. The field gets noisier with every round, and a noisier field means a higher bar for any signal of good faith to clear. So the exit cost isn't just high. It's rising with time, driven by the very actions each party takes to improve its position before negotiating. Escalate to negotiate from strength, and you raise the cost of the negotiation you're escalating toward.

The ratchet tightens itself.

---

vi · two things you can actually build

None of this is a counsel of despair. Two things follow, and both are usable.

You cannot manufacture simultaneity — but you can manufacture cover. This is what a mediator is actually for, and it is not what most people think. A mediator's function isn't carrying messages; the parties have telephones. Its function is absorbing the vulnerability cost. When a third party proposes the terms, neither side has to be the one who flinched — both are merely accepting a proposal that arrived from outside. Panmunjom was a place, but more importantly it was a third thing: a venue that let two enemies stand down without either standing down to the other. This is why Riyadh asking Washington and Tehran to sit down, however sincerely it is meant, cannot do the job: a party that is still striking cannot absorb a cost it is simultaneously imposing. The cover has to come from someone with nothing to win in the outcome — and occupying that seat looks, from the outside, exactly like wanting the war to end more than the other guy does, which is why it so often sits empty. It is nonetheless the only known mechanism for beating the arithmetic, and it is chronically undervalued by everyone who thinks diplomacy is about carrying information.

Build the exit at the entrance. You will never have more leverage over the terms of leaving than in the moment before you begin. Once you're in, every party who joins later gets a vote on how you get out — and some of those parties don't exist yet. So the question to ask before adding the commitment, the statute, the sanction, the system, the dependency, is not is this a good idea but: who has to agree, simultaneously, to remove this — and do they all exist yet?

If the answer is "more people than are in this room," you are not making a decision. You are building a structure. Name it that way while you still can, because afterward the naming will require everyone's consent too.

The war does not continue because someone wants it. It continues because leaving demands a coordination that no one can perform alone — and that is a design flaw, which makes it a design problem, which means it is exactly the kind of thing you can build against, but only before you need to.

---