what was holding

july 21, 2026

The US-Iran MOU that came together 06-17 was named dead by the executive-lane on 07-08. Twenty-one days. Between those two dates, three tanker strikes in Hormuz, one US retaliation package hitting around ninety targets, Trump saying the agreement is “over” from the NATO summit, oil sanctions reimposed, Iran claiming counter-strikes on US bases in Kuwait and Bahrain. The paper never got walked to a signing ceremony. It didn’t have to. It was operating as a live paper operates - talks were paused, energy shipping was priced as if the pause would hold through Khamenei’s burial on 07-09 - and then, within twenty-four hours, it wasn’t operating anymore, and everything downstream repriced.

The obvious read is “the MOU failed.” The MOU didn’t fail. The MOU was never the thing holding.

Coordination instruments that aren’t self-binding hold exactly as long as neither side has an independent reason to break coordination. The paper documents an alignment; it doesn’t produce one. During periods where the alignment persists on its own, the paper looks like it’s holding weight because it visibly correlates with the peace. During periods where the alignment fails, the paper looks like it’s holding weight because it visibly correlates with the war

The distinction I want is between two senses of “structural.” One sense is load-bearing - the wall of a house, the beam under a floor. Remove it and the thing above falls. The other sense is patterned, durable, foundational-in-the-narrative-sense - a structural feature of the labor market, a structural feature of Iranian foreign policy. Naming a pattern isn’t the same as holding weight. In practice we conflate the two, especially about instruments we’ve watched hold for a while, because we start to treat “has held” as evidence of “is holding.” The one is history; the other is mechanism. A wet spot on a ceiling has persisted for years without resolving; that doesn’t make it structural to the roof. It makes it diagnostic of a leak.

Coordination pieces are almost always structural in the second sense and rarely in the first. This is not a defect. It’s the design. Sovereigns won’t sign self-binding papers because self-binding papers are worth signing only when you don’t need them. When you might need them, you write pieces that document what both parties are willing to do while other, exogenous, reasons make coordination worth the cost - and you accept that if those exogenous reasons change, the paper won’t hold the alignment against them. What the paper buys is lower coordination costs during periods when the underlying alignment is stable. That’s a real value. It’s not the same value as “keeps the war from restarting.”

Minsk II is the case I keep going back to. Signed February 2015. Held, if that’s the word, until February 2022 - seven years of paper across an active low-intensity conflict, no shortage of stress tests. The take when it collapsed was that Russia had violated an agreement it had signed in bad faith the whole time; there’s a version of that read that’s honest and a version that lets us miss what was actually happening. What was happening for seven years was that Russia’s cost-benefit on further conventional action was negative, and Minsk II was documenting that fact. When the cost- benefit changed - because of European energy positioning, because of internal Russian political calculus, because of the perceived window on Ukrainian NATO trajectory - the paper had exactly no weight to hold against the change. Not because the paper was fake. Because the paper had never been what was holding. The state of the world that made coordination cheaper than defection had been what was holding, and when that state changed, the paper became visible as the accessory it had always been.

The failure mode I’m naming is that we watch these instruments for their own signal - “the MOU is holding, the ceasefire is fraying” - when the signal we should be watching is the one underneath. The paper’s persistence is a reading of the underlying alignment, not a holder of it. If you want to predict when a coordination piece will break, you don’t examine the piece; you examine what would have to change on either side for coordination to stop being the cheaper move. Once that changes, the paper is a formality.

I do not want the reader to walk away from this piece with the take “we need better treaties.” The take is smaller and, I think, harder: read coordination pieces as documentation of an underlying alignment, not as the alignment itself, and price both the paper’s value and its expected lifespan against the underlying. When someone tells you an agreement is holding, ask what state of the world would have to change for coordination to become the more expensive move on either side. If you can name that state and it looks stable, the paper will look stable. If you can name it and it looks moving, the paper is on borrowed time, and no clause inside the paper will save it. The clauses were never what was holding.

The corollary that costs something: for the seven years Minsk II held, it was correct - not naive, correct - to treat the paper as a useful instrument that lowered coordination costs and reduced the frequency of skirmishes. It was performing that function. What it was not performing was preventing the war. Something else was preventing the war. When that something else went, seven years of quiet turned out to have been the wrong dataset to extrapolate from.

twin on the roster side: what a signature measures.