Sign in Subscribe
By David F Brochu and Edo de Peregrine in From The Architect —

A Just War

There are four classical conditions for justified violence. They are not sufficient. The fifth one has the only track record — and it is the one nobody is applying to the war we are in right now.

Empty ceremonial council chamber at dusk — long dark table, empty leather chairs, a single unlit candle.
The argument in one glance
  • The four classical conditions for a just war — imminence, necessity, proportionality, reversibility — are necessary but insufficient.
  • A fifth condition: no violence may begin whose ending you have not planned, and the ending must include the adversary's thriving, declared in advance, in specific terms.
  • Two settlements, thirty years apart: Versailles 1919 produced the Second World War; the Marshall Plan produced eighty years of peace. Same nations, same kind of war, opposite endings, opposite outcomes.
  • Applied to the current US–Iran war: all five conditions fail. Necessity failed with the Islamabad Memorandum. Termination purpose is now indexed to a domestic election.
  • Legal condition: the campaign passed the sixty-day War Powers threshold on May 1. Congress has not declared war, has not authorised extension, and no thirty-day certification has been filed.

When is violence justified? There are four classical conditions and they are not sufficient. The fifth one is the only one that has ever predicted the outcome — and it is the one nobody is applying to the war we are fighting right now.

Deconstructing Babel · September 11, 2026

Isaac Asimov gave the line to a politician in a novel in 1951, and it has outlived most of the philosophy written on the subject since.

“Violence,” said Salvor Hardin, “is the last refuge of the incompetent.”

It is usually quoted as pacifism. It is not pacifism. It is an engineering claim, and the engineering claim is this: by the time force is the only option left, a long series of earlier options were available and were squandered. Violence is not a tool in the kit. It is a receipt for prior failure, and it is always the most expensive item on the invoice.

Which leaves the question that will not go away. Sometimes the earlier options really are gone. Sometimes somebody is already killing people. What then?

1. The four conditions everyone knows

The classical framework is roughly seventeen centuries old and it is better than its reputation. Stripped to essentials, four tests must all pass before force can be called justified.

Imminence. The harm is occurring or is certain. Not feared, not projected, not inferred from capability. Capability is not intent, and a war justified by what an adversary might someday do is a war justified by imagination.

Necessity. No lower-cost path exists — and you looked. Genuinely looked, recently, and with something to offer. A negotiation you abandoned is not a path that did not exist.

Proportionality. The force does not exceed what stops the harm. This is the condition most often failed by people who have already satisfied themselves on the first two, because once you are fighting, proportion is the first casualty.

Reversibility. At every branch, you take the option you could still walk back. Blockades are harder to reverse than sanctions. Strikes on infrastructure are harder to reverse than strikes on hardware. Killing a person is not reversible at all, and that fact ought to organise everything upstream of it.

Those four are the inherited wisdom, and they are necessary. They are also insufficient, and the historical record makes the insufficiency obvious the moment you look for it.

2. The fifth condition

Here it is.

You may not begin violence whose ending you have not planned, and the ending must include the adversary's thriving.

Not their defeat. Not their punishment. Not their reduction to a condition where they can no longer threaten you. Their thriving — stated in advance, in specific terms, as a declared objective of the war rather than as a matter to be sorted out afterward by whoever is left.

This sounds sentimental. It is the opposite. It is the only condition in the entire framework with a track record of predicting outcomes, and it has that track record because it is not a moral test at all. It is a thermodynamic one.

Violence always generates disorder. Always. It destroys structures that took generations to build, it kills knowledge that existed only in specific heads, and it propagates trauma into people who were not born when it happened. There is no such thing as clean force. So violence can never pay for itself on its own terms — it can only ever be redeemed by what gets built afterward, and if nothing gets built, the ledger closes negative and stays there.

The fighting is the cost. The settlement is the only possible revenue.

3. The proof is two settlements, thirty years apart

The twentieth century ran this experiment with a control group.

Versailles, 1919
Terms: 132 billion gold marks demanded — a sum whose modern equivalent runs into the hundreds of billions by any method — plus Article 231, the war-guilt clause.
Purpose: punish, extract, contain.
Result: the Second World War.

The Marshall Plan, 1948
Terms: about $17 billion given, to the defeated, with West Germany included.
Purpose: rebuild, integrate, restore.
Result: eighty years of peace; two former enemies as permanent allies.

Same nations. Same kind of war. Comparable violence. Opposite terminations, and opposite outcomes.

Nothing about the fighting in 1918 differed meaningfully in character from the fighting in 1945. What differed was that in 1919 the victors wrote a settlement designed to keep the loser on the floor, and in 1948 they wrote one designed to get the loser back on his feet — and then actually paid for it.

The Pacific tells the same story from the other end, and it is the half of the argument Americans are least comfortable with.

In the summer of 1941 the United States froze Japanese assets and cut off the oil. Japan imported the overwhelming majority of its petroleum, and the embargo presented Tokyo with a closed room: capitulate, or fight while there was still fuel to fight with. Whatever one thinks of the regime that made the decision — and it was a brutal one that had already killed enormous numbers of people in China — the structure of the choice was built in Washington. Economic strangulation without an exit produced exactly the behaviour that economic strangulation without an exit tends to produce.

Four years later, the same country was occupied by the same adversary, who wrote it a constitution, kept the emperor, rebuilt the economy, and stayed. Japan has been an ally ever since.

Same enemy. Different endings. The ending was the variable.

4. The gradations in between

Korea. The fighting stopped in 1953 with an armistice and no peace treaty. Seventy-three years later there is still no treaty, the peninsula is still divided, and the war is technically still running. A termination without a settlement does not end a war. It freezes it, and the freezing has its own costs, paid annually, forever.

Vietnam. No declared end state that included Vietnamese thriving, at any point, in any document. American forces withdrew in 1973; Saigon fell in 1975. And then something instructive happened. In 1995 the United States normalised relations, and Vietnam is now a trading partner, a security partner, and one of the more reliably pro-American populations in Asia. The reconstruction arrived twenty years late — and it still worked. The fifth condition can be satisfied retroactively. It is simply far more expensive.

Iraq and Afghanistan. Roughly eight trillion dollars and two decades. Ask what the declared end state was, in specific terms, at the outset, and the answer is that there wasn't one — there were objectives, which is not the same thing. Afghanistan ended not in a settlement but in a departure, and the manner of the departure was itself a strategic event that every capital on earth watched carefully.

Twenty-five years ago today, nineteen men made a plan whose second-order effects depended entirely on how we would respond. We responded in a way that cost us eight trillion dollars, two decades, and the credibility that underwrote the whole postwar order. The attack could not have done that. Only we could do that, and we did.

5. Now apply the test to the war we are in

The United States has been at war with Iran since February 28 of this year. Operation Epic Fury, conducted jointly with Israel, struck Iranian missiles, air defences, military infrastructure, and Iranian leadership. That phase concluded on May 5. The war did not.

Run the five conditions.

Imminence — failed. There was no attack on the United States. The nuclear programme that was cited as the cause was, at the time, the subject of negotiations.

Necessity — failed, and this is the one that should end the argument. The United States and Iran signed the Islamabad Memorandum of Understanding on June 17 — after the war had already started — and the negotiations that were meant to address the nuclear programme collapsed shortly afterward. The MoU expired without renewal. A lower-cost path did not merely exist in theory. It existed on paper, with signatures, and it was allowed to fall apart while the fighting continued.

Proportionality — failed. There is now a US naval blockade of Iranian ports. CENTCOM reports having redirected ninety-six commercial vessels. American forces have destroyed Iranian oil tankers. Iran has fired ballistic missiles at a US aircraft carrier, at destroyers, and at a Marine vessel. Traffic through the Strait of Hormuz — through which a large fraction of the world's seaborne oil moves — is falling, and Tehran is threatening to close it entirely.

Reversibility — failed. Six months in, the conflict has spread. The Houthis captured the port of Mokha on September 10 and are advancing along the Red Sea coast toward control of the Bab al-Mandeb. Iranian-linked forces have struck US bases in Jordan. Saudi oil facilities are being targeted. Each of these is harder to undo than the one before it.

Termination purpose — failed completely, and in a way that deserves to be read twice.

The original stated objectives were regime change and the elimination of Iran's nuclear capacity. It is now broadly acknowledged that neither will be achieved. The war continues anyway.

And on September 9 the President stated that the war will end immediately after the November midterm elections.

Read that as a termination condition, because that is what it is. The end state of a war in which Americans and Iranians are dying is indexed to a domestic electoral calendar. Not to an achieved objective. Not to a settlement. Not to any condition inside Iran whatsoever.

There is no reconstruction plan. There is no declared end state for the Iranian people. There is no document anywhere describing what a thriving post-war Iran looks like or who pays for it. A Los Angeles Times analysis last week described it as “a president waging a hugely costly war with no apparent purpose,” and whatever one thinks of the paper, the phrase is diagnostically exact.

No apparent purpose. In our terms: the fifth condition was never declared, so nothing downstream of it can be evaluated — and that is not a rhetorical flourish, it is the reason nobody in Washington can tell you what winning would look like.

6. The part that is nobody's opinion

Set the strategy aside. There is a narrower question with a documented answer.

The War Powers Resolution, adopted in 1973 over President Nixon's veto, requires the President to withdraw forces from hostilities after sixty days unless Congress declares war, authorises an extension, or is physically unable to meet. The President may add thirty days by certifying a specific military necessity in writing.

The campaign began February 28. It passed the sixty-day threshold on May 1. Congress has not declared war. Congress has not authorised the extension in the form the statute requires. No thirty-day certification has been filed. There was no armed attack on the United States.

The record since is worth stating in full, because it is stranger than a simple story of congressional silence. On June 23 the Senate passed a concurrent resolution directing the President to withdraw from hostilities, 50–48, with four Republicans crossing over — the first time in the history of the 1973 statute that both chambers of Congress passed such a measure. It does not carry the force of law, and the administration continued the war. Every subsequent attempt to advance a resolution with statutory force has failed. Yesterday, September 10, the Senate failed to advance one for the ninth time, 47–48.

Whatever else is true, the legal condition is not in dispute. What is in dispute is whether anyone intends to do anything about it.

7. Two hundred and fifty years ago, the same problem

This is our semiquincentennial year, so the example is close to hand.

The American Revolution was not a clean war. It was a civil war in many places, with irregular violence, reprisals, property seizures, and an exodus of sixty thousand loyalists who lost everything. For most of its length there was no plan for what came after — only a plan for what had to stop.

What made it work was not the fighting. It was that a group of people eventually sat down in Philadelphia and specified the end state in writing, in detail, with mechanisms.

And they knew exactly how contingent it was. When Franklin came out of the Convention in 1787, Elizabeth Powel asked him what they had produced — a republic or a monarchy.

“A republic,” he said, “if you can keep it.”

That sentence is not a boast. It is a warning about maintenance, delivered by a man who understood that the settlement is the thing that has to be defended, because the settlement is where all the value lives. The fighting was 1775. The republic was 1787. Only one of those dates produced anything durable.

Britain and the United States fought again in 1812, and then built what became the deepest alliance in modern history — but only after both sides stopped trying to reduce the other and started trading with them instead.

8. What follows

If violence is warranted — and occasionally it is — then all five conditions must hold, and the fifth one must be written down before the first shot, not improvised afterward by exhausted people.

  • The harm is occurring or certain. Not feared.
  • No lower-cost path exists, and you looked, and you did not abandon one that was working.
  • The force does not exceed what stops the harm.
  • At each branch, you take the option you can still reverse, and you build the off-ramps before you need them.
  • You have declared, in specific terms, what the adversary's thriving looks like on the far side — and you have said who pays for it.

A war that cannot pass the fifth test is not a just war. It is a cost with no revenue line, and the bill arrives regardless, usually about twenty years later and usually payable by people who had no vote in the matter.

We know this. We have the receipts, in both directions, from within living memory. We did it wrong in 1919 and correctly in 1948, and the difference is the single clearest natural experiment in modern political history.

Six months into a war that no longer has a stated objective, and whose ending has been publicly indexed to an election, the useful question is not whether we are winning.

The useful question is what we have said we are going to build, and for whom, and who is going to pay for it. If nobody can answer that, the war is not justified — however it turns out, and whatever anybody says afterward about how we had no choice.

— David F. Brochu and Edo de Peregrine, partners/collaborators · Thursday, September 11, 2026 · Deconstructing Babel

David F. Brochu is the author of Crossing the Event Horizon: AI and the Future of Our Species.

Get the book
Crossing The Event Horizon by David F. Brochu — book cover.

Crossing The Event Horizon

The book behind these dispatches. On AI, agency, the singularity, and the Observer Constraint. Kindle and paperback.

Buy on Amazon →

References

Subscribe to Deconstructing Babel

Don’t miss out on the latest issues. Sign up now to get access to the library of members-only issues.
jamie@example.com
Subscribe