The Terms That Never Clear

Iran had delivered it the day before, to the ambassadors of the six governments then negotiating over its nuclear program. It offered, in the diplomatic register, "serious negotiations." It did not offer what Security Council Resolution 1696 had demanded three weeks earlier: suspension of all enrichment-related and reprocessing activity, verified, by August 31. Washington's answer was that a response which did not begin with suspension was not a response.

At that moment Iran was operating a single cascade of 164 centrifuges at Natanz, enriching to 3.5 percent, under IAEA seals and cameras.

Twenty years later, on the night of February 28, 2026, the United States and Israel bombed the country and killed its supreme leader. The trigger, as stated, was Tehran's refusal to end all enrichment. Two days earlier, in Geneva, Iranian and American negotiators had concluded what Foreign Minister Abbas Araghchi called the most serious and longest talks yet, with reported progress and no agreement. Iran's offer that week: blend down the 60-percent stockpile, accumulate no enriched gas, accept broad IAEA oversight, and after a multiyear pause enrich to 20 percent on up to thirty IR-6 machines to fuel research reactors.

It fell short.

Same sentence. Same file. The only thing that changed in twenty years is the delivery mechanism, which went from a press briefing to an air campaign.

Two concessions before the argument, because omitting them is how a piece like this earns its dismissal. Iran is not a passive object in this record. Natanz became public in 2002 through an opposition group rather than a declaration, Fordow was disclosed in September 2009 only after Western intelligence had already found it, and a government that builds an enrichment hall inside a mountain is running a hedging strategy it would plausibly have pursued under any terms on offer. And the United States is not the only party at the table. Israel has kept its own clock throughout, with its own threshold and its own preferences, and retains the standing ability to make any American concession non-binding by acting against it — which means every American negotiator has bargained with a co-signer who never signed.

Both of those are true. Neither of them accounts for the shape of the record, which is the subject here.

i · the document designed to be refused

A precondition is a claim about sequence. Before we discuss X, do Y. This is ordinary and often reasonable — stop shooting, then we'll talk. It becomes something else entirely when Y is the thing X was supposed to produce.

Zero enrichment was never a step toward the agreement. Zero enrichment was the agreement, minus the sanctions relief, the inspections regime, and the signature. Demanding it as the entry fee means demanding that Tehran concede the entire negotiated outcome in advance and then negotiate over what it might receive for having already surrendered. No government survives that arithmetic, which is exactly why the demand was durable. A precondition that cannot be met is a permanent policy. It never has to be revisited, defended, or scored, because the other side is always the one refusing.

The template is old enough to have a name in every foreign ministry's filing system.

On July 23, 1914, Austria-Hungary handed Serbia a note with ten demands and a forty-eight-hour clock. Point six required Austro-Hungarian officials to take part in a judicial inquiry on Serbian soil. Serbia accepted nearly everything and balked at that one item; Vienna broke relations the same afternoon it received the reply. The Ballhausplatz had drafted the note to be rejected — this was not a later historian's inference but the working assumption of the men who wrote it. The ultimatum was not a diplomatic instrument that failed. It was a war decision wearing diplomatic clothing, and the clothing was the point, because it produced a documentary record in which someone else refused.

Vienna is an unusually clean case precisely because the intent is documented. Washington's is not, and the honest position is that nobody outside the room can settle it. There is a reading of the 2026 round in which the American team was not insincere but simply unprepared — Kelsey Davenport's assessment in Arms Control Today faults the negotiators on competence rather than motive, and that reading deserves to stay on the table.

It also changes less than it appears to. Sincere maximalism and deliberate pretext generate identical documentary records, which is the same observation as attribution-filed-in-advance, running in the other direction. Whatever the intent, an unmeetable demand operates as a war decision in diplomatic clothing. The record measures operation, not sincerity, and operation is the only thing the receiving government can act on.

There is a further explanation for the demand's durability that is less sinister than design and considerably more powerful. Zero enrichment may be the only position that survives contact with two domestic political systems at once — the one American demand that cannot be flanked from the right, and the one Iranian refusal that cannot be flanked by the hardliners. A policy nobody can afford to abandon outlasts a policy somebody deliberately engineered, and it fails in exactly the same way while implicating no one. That reading makes the twenty-year persistence more legible than malice does, and it makes the forecast worse.

ii · the control experiment

The rare gift of the Iranian file is that it contains its own control group. For roughly two years starting in November 2013, the precondition was set aside. The interim deal and then the July 2015 accord accepted a monitored Iranian enrichment program rather than demanding its abolition: 3.67 percent, a 300-kilogram stockpile ceiling, 5,060 first-generation centrifuges at Natanz, Fordow converted away from enrichment, inspectors everywhere.

Whatever else one says about that arrangement, it is the only period in the twenty-year record during which Iran's program was both capped and verified. The demand that had failed for a decade was dropped, and the outcome the demand had been justifying arrived within eighteen months.

The obvious objection is that three variables moved at once. The 2012 oil embargo and the expulsion of Iranian banks from SWIFT had done real damage to the economy, and Hassan Rouhani replaced Mahmoud Ahmadinejad in August 2013. Run the causation backwards and you get a coercion story: the precondition was dropped because pressure had finally made a deal cheap enough to close. That reading is available to any hostile reader for free, and it has to be answered rather than ignored.

It does not survive the dates. Coercion was not the new variable in 2013 — it was the constant. The pressure architecture had been accumulating since 2006 through four Security Council resolutions, and the oil embargo had been in force for over a year before Rouhani took office. What seven years of escalating coercion measurably produced was centrifuges: 164 machines in 2006, more than eighteen thousand installed by mid-2013. Sanctions ran as an independent variable for the better part of a decade and their recorded output was an expanding program. Nor can an Iranian election explain an American concession — Washington's position changed on Washington's side of the table, and the secret bilateral channel in Oman that produced the interim agreement opened in the spring of 2013, months before Rouhani was elected. The variable that had never been tried was the entry condition. It was tried, and the file moved.

There is an older version of the experiment, and it shows less than it first appears to. Through the late 1950s and early 1960s, negotiations for a comprehensive nuclear test ban deadlocked over on-site inspections — Washington wanted seven a year, Moscow offered three, and five years evaporated into the gap. In 1963 Kennedy removed from the treaty's scope the categories of test that required on-site verification. Averell Harriman went to Moscow in July and came back with a signed text in under two weeks. But that was a narrowing of the goal and not only of the entry fee, and the Cuban missile crisis had intervened nine months earlier with a shock at least as plausible a mover of Moscow as any drafting change. What the case establishes is narrow and still worth having: an entry condition can be severable from the outcome it was supposed to guarantee, and severing it can move a file that has not moved in five years.

Drop the precondition, get the agreement. Reinstate it, get the file back. In May 2018 the United States withdrew from the 2015 accord and reimposed the maximum demand, and the sequence resumed precisely where it had paused: limits breached in 2019, sixty-percent enrichment beginning in April 2021, particles at eighty-four percent detected in 2023.

iii · the scoreboard

Every strategy should be graded against the metric its own advocates chose. The stated purpose of the precondition architecture was to prevent Iranian enrichment.

August 2006: 164 centrifuges, one cascade, 3.5 percent, fully inspected.

Mid-2013, after four sanctions resolutions and seven years of the same demand: more than eighteen thousand centrifuges installed across Natanz and Fordow.

May 2025: 408 kilograms of uranium enriched to 60 percent, roughly a fifty-percent increase in three months.

February 2026: Tehran informs the IAEA that normal safeguards have become legally untenable. The agency reports it can no longer verify the status of the program at all. Two days after the last round of talks, the bombs. By March the entrance buildings at Natanz are wreckage and the site is inaccessible — to inspectors as well as to everyone else.

From a monitored cascade of 164 machines to an unmonitored program of unknown size, in twenty years, at a cost of two air campaigns, one assassinated head of state, and a regional war that a Pakistan-brokered pause has so far only interrupted. Analysts now argue over whether the strikes delayed weaponization or accelerated the decision to pursue it. That the question is live at all is the verdict.

Nobody will be held to this. The people who insisted on zero enrichment in 2006 and got eighteen thousand centrifuges will explain that the failure proves the demand should have been enforced sooner and harder, which is the load-bearing beauty of an unmeetable condition: it converts every consequence into evidence for itself.

iv · what the demand actually communicates

Strip the language and read what a maximalist precondition transmits to the government receiving it.

It says: the object of this exercise is your capitulation, and capitulation is not an event we intend to compensate. It says: whatever you dismantle, we retain the option to demand more, because the standard was never a technical threshold but your posture. And it says — this is the part the file never stops repeating — that disarmament is not a transaction with a counterparty but a status change, and the status you are being offered is disarmed adversary.

Every proliferation ministry on earth keeps the same two documents in the same drawer. On December 19, 2003, Muammar Gaddafi announced he was abandoning Libya's unconventional weapons programs. On October 20, 2011, he was killed in a ditch outside Sirte, after a NATO air campaign that his disarmament had helped make thinkable. You can argue the causation until the sun burns out. It does not matter what the correct reading is. It matters that this is the reading every regime with a centrifuge cascade has already made, and that no American administration has ever offered anything that answers it.

v · the thing that was actually destroyed

Now run the twenty years again and count a different quantity. Not centrifuges — sight.

In 2006 the program was small and Washington could see all of it: declared facilities, agency seals, cameras on the cascade hall, inspectors with access and a right to return. Between 2015 and 2018 the program was larger and Washington could still see nearly all of it, under the most intrusive verification regime ever attached to a civil nuclear program. After 2018 the cameras came off in stages, inspectors were de-designated, monitoring data went unshared. By February 2026 Tehran declares normal safeguards legally untenable and the agency states plainly that it cannot verify the program's status. By March the relevant buildings at Natanz are rubble, which is a form of not being able to look at something.

Two curves, moving in opposite directions. The program grew from 164 machines to an unknown number. The world's ability to see the program went to zero. These are not two separate outcomes; the second is the price that was paid, over and over, for pursuing the first by these means.

And sight is the scarce good in the entire file. Everything else on the table can be seized or destroyed. Facilities can be bombed. Stockpiles can be interdicted. Scientists can be killed and leaders assassinated, and all of that has now been done. Verification is the one asset that cannot be taken, because it consists entirely of another government's ongoing consent to be watched — which means it can only be bought, and the precondition architecture was a twenty-year refusal to pay the price. A bomb can remove a building. It cannot restore an inspector's line of sight. It is, precisely, the instrument that ends one.

Coherence between hostile parties is not built out of trust or goodwill, neither of which was ever on offer here. It is built out of what each side can verify about the other's actual constraints — the agreement is nothing more than a durable record of two positions that were genuinely available. A demand structured so that only one side can move is not a hard bargain. It is the absence of a bargain, and it produces exactly the thing it was drafted to prevent, because a party told it has no future in compliance will rationally invest in the alternative.

The demand was justified, throughout, by the need to know what Iran was doing. Twenty years of enforcing it have produced a world in which nobody knows what Iran is doing. That is the result, and it is not a failure of toughness.

vi · the prediction

The talks will continue. There will be a framework, a deadline measured in days and extendable by mutual consent, a reconstruction figure large enough to make headlines, and a communiqué describing progress. Somewhere in the American position there will remain a clause about enrichment on Iranian soil, and somewhere in the Iranian position there will remain a claim of right under Article IV of the Nonproliferation Treaty, and the two will be described as narrow remaining differences.

They are not narrow. They are the whole thing, and they have been the whole thing since a spokesman stood at a podium twenty years ago today and said a document from Tehran did not meet the conditions.

If the precondition is dropped, there will be an agreement within a year, because there was one the last time it was dropped. That is a claim that can fail, and it is worth stating what would falsify it: an American position that concedes monitored low-level enrichment on Iranian soil, met by continued Iranian refusal, would mean the demand was never the obstacle and this analysis was wrong. It is also worth stating what could make it fail for reasons that have nothing to do with Washington. Israel does not need American agreement to strike, and a strike is sufficient to void one. Any framework that does not somehow bind the party most willing to break it is a framework with a fuse.

If the precondition is not dropped, the file runs forward another decade in the same shape, with the arsenal of consequences already largely spent — the sanctions reimposed, the sites struck, the leadership decapitated — and nothing left to escalate to except the thing everyone claims to be preventing. And this time it runs forward in the dark, which is the genuinely new condition. Every previous round of failure at least left somebody counting.

Twenty years is a long time to run the same experiment and call each result a surprise. The centrifuges kept a better record than the diplomats did, and now even the centrifuges have stopped reporting.