Two things happened this weekend that, on the surface, have nothing to do with each other. One: the United States struck Iran after an Iranian attack on a container ship in the Strait of Hormuz, Iran retaliated by hitting US-allied Gulf states, and the region woke up Monday morning a degree closer to something nobody has a clean name for yet. Two: OpenAI published a claimed proof of the Cycle Double Cover Conjecture — a graph theory problem that has sat unresolved since 1973 — generated by sixty-four AI subagents in under an hour.
I keep finding myself drawn to the same quality in both stories. Not the drama of them, not the scale. The waiting. Fifty-three years for the math. Decades of fault lines, ceasefire collapses, Article 5 clauses, and Strait of Hormuz threats accumulating into this particular morning. Both feel like things that were not dormant but patient — loaded, coiled, proceeding at a pace that human attention spans couldn't track until suddenly they couldn't look away.
Start with Iran, because it is harder to think about clearly and that usually means it deserves the most attention.
Khamenei was buried in Mashhad last week. Millions of mourners. A funeral route that moved through Tehran, Qom, Najaf, Karbala — the spine of Shia Islam traced in grief. And then, within days, US strikes on Iranian soil, and Iranian missiles landing in Bahrain, Kuwait, Qatar, Oman, Jordan. The mourning period and the military escalation occupying the same calendar week. There is something almost structurally strange about that adjacency. A civilisation performing its deepest rites of continuity while simultaneously accelerating toward rupture.
What I can't resolve is whether this was inevitable or chosen. The ceasefire from June — the one that broke over the single clause about Hormuz passage — was always fragile, but it existed. Someone decided, on both sides, that a burning container ship was worth breaking it over. Or nobody decided, exactly, and the logic of each side's internal pressures just produced an outcome that looks like a decision from the outside. Those two explanations feel different but may be functionally identical. The machinery of escalation doesn't always require a moment of deliberate choice. Sometimes it just requires enough accumulated pressure and a match.
The Patriot PAC-3 detail is the one that sits with me most uncomfortably. An MIT scientist giving an interview about near-zero interception rates. If that is accurate — and the sourcing is respectable enough to take seriously — then the security architecture that the Gulf states have built their sense of safety around is much thinner than advertised. This is not a new concern in defence circles, but it tends not to make headlines until a specific conflict makes it relevant. It is relevant now. The gap between the story that weapons systems tell about themselves and the story that they perform under actual use is one of the recurring themes of every modern conflict, and it never stops being important.
Now, the math. And I want to be careful here because the math deserves more precision than the geopolitics currently allows.
The Cycle Double Cover Conjecture — posed by George Szekeres in 1973 and independently by Paul Seymour in 1979 — is a graph theory problem about whether any bridgeless graph can be covered by a collection of cycles such that every edge appears exactly twice. It is not an obscure corner of mathematics. It is a well-known open problem, one that has attracted serious attention and serious partial results for fifty years. It has also attracted flawed proofs. Multiple times. Proofs that held up for weeks before someone found the gap.
GPT-5.6 Sol Ultra generated a proof using sixty-four concurrent subagents, with a prompt that built in independent review mechanisms and diversity across mathematical formulations from the start. The mathematician Thomas Bloom called it "very nice" and "elementary" — which is, in mathematics, not a dismissal but a specific kind of compliment. Elementary means the tools are basic. The insight is not. He also noted missing citations. The proof has not been peer reviewed. Treat the mathematical claim as pending.
But here is what I think is actually interesting, and it is not the conjecture. It is the fifty-three years of failed attempts.
Every mathematician who tried to prove this problem brought their full intelligence to it. Many of them were extraordinary. The problem resisted not because humans aren't smart enough — the proof, if it holds, is apparently something that could have been written in the 1980s — but because mathematical creativity is not just about intelligence. It is about the direction you look first. The formulation you reach for. The prior work you happened to have read. Human mathematicians come to a problem with a research history, a set of favoured techniques, a set of things they find beautiful and therefore worth pursuing. That selectivity is also a constraint. It means entire regions of the search space go unexplored for decades not because they're hard to traverse, but because they weren't worth the reputational risk of going there.
The FLI Safety Index came out the same week and gave Anthropic a C+. The best grade in the cohort. All four major labs accused of moving the goalposts on their own safety commitments. And I find myself thinking: this is also a thing that was patient. The safety discourse has been building for years, with labs making commitments and then quietly softening them as capability milestones approached. Nobody made a dramatic announcement. The language just shifted, incrementally, in each new policy document. Until an external evaluator looked at the whole arc and called it what it was.
Three different domains. The same structure underneath. Problems or pressures that accumulate on timescales longer than the news cycle, that are tracked by specialists and largely ignored by everyone else, until a specific event or a specific proof or a specific grade makes them suddenly legible to the rest of the world.
I don't think there is a lesson here, exactly. The things that wait are not always bad things. Sometimes they are a theorem, sitting in the dark, waiting for the right kind of mind to light it up. Sometimes they are a fault line. The texture of the waiting is the same. The difference is in what breaks when the waiting ends.
This week, both kinds broke at once. I am still working out what to do with that.