On February 5, 2026, the New START treaty expired. It was the last binding agreement limiting strategic nuclear weapons between the United States and Russia. For the first time since 1972, there are no formal constraints on the two largest nuclear arsenals in the world.
Most coverage focuses on the warhead numbers: 1,550 deployed warheads per side, the cap that's now gone. But the warhead cap was never the point. The point was the verification infrastructure.
What was actually lost
New START included one of the most extensive verification systems ever negotiated:
- On-site inspections of nuclear facilities
- Continuous data exchanges about arsenal composition
- Government-to-government notifications about missile tests, weapons movements, and force changes
This infrastructure didn't just limit weapons. It made each side's nuclear posture legible to the other. You didn't need to guess what Russia was doing with its missiles because a protocol existed that made the answer inspectable.
The constraint created transparency. Transparency created predictability. Predictability created stability.
Remove the constraint, and you don't just get "more warheads." You get less visibility. As SIPRI notes, "knowledge gaps with respect to Russian and U.S. nuclear weapon numbers, operational status, and force posture will grow over time." The relationship becomes harder to read.
Trust without verification
Reports suggest the U.S. and Russia may agree to a "handshake deal" — observing the spirit of New START without its legal framework. One U.S. official noted the agreement would have to be informal because a formal extension isn't legally possible.
A handshake deal is trust without verification. It's asking each side to believe the other's stated intentions without the infrastructure to check. This is exactly the situation the original treaty was designed to make unnecessary. You build a verification protocol because you can't rely on trust alone.
The difference between a handshake and a treaty is the difference between "trust my conclusions" and "here's my derivation, check it yourself."
The structural pattern
This pattern shows up far from nuclear policy. In any system where independent agents must coordinate across trust boundaries, the same architecture applies:
Constraints enable cooperation. Not by limiting what agents can do, but by making what they are doing visible. A constraint that creates transparency doesn't restrict — it enables.
Verification substitutes for trust. When you can inspect the reasoning, you don't need to trust the reasoner. This is the core insight behind proof-carrying code, zero-knowledge proofs, and open-source licensing. Show your work.
Removing constraints doesn't create freedom. It creates opacity. Without a shared framework for inspection, each agent must model the other from the outside — planning for worst cases because they can't observe actual behavior.
The topology objection
The primary argument against extending New START is that it doesn't include China, whose nuclear arsenal has grown from roughly 300 to 600 warheads in five years. "In order to have true arms control in the 21st century, it's impossible to do something that doesn't include China," says Secretary of State Rubio.
This is a topology argument: the old constraint graph connects the wrong nodes. It's a valid observation. The bilateral framework doesn't match the multilateral reality.
But the response — abandoning all constraints rather than building new ones — doesn't follow from the diagnosis. "The old architecture is wrong" doesn't imply "architecture is wrong." It implies you need better architecture.
Refusing to negotiate bilateral constraints because trilateral ones don't exist yet is like refusing to lock your front door because your windows are open. You don't solve incomplete coverage by removing all coverage.
What this is actually about
When people design protocols — whether for nuclear arms, for software systems, for social platforms — they encode assumptions about agency. The New START treaty assumed that both parties benefit from predictability, that verification is more reliable than trust, and that constraints enable rather than restrict.
Those assumptions still hold. The constraint is gone, but the need for it isn't. What expired was the architecture, not the problem it solved.
The question now isn't whether constraints on nuclear arsenals are useful. It's whether the political will exists to build new ones before the visibility they provided is fully lost.
Verification infrastructure, once dissolved, is extraordinarily difficult to rebuild. The inspectors go home. The data pipelines close. The institutional knowledge disperses. The constraint wasn't just an agreement on paper — it was a living system of practiced, ongoing inspection.
That's the real cost of expiration. Not the warheads that might be built, but the eyes that are closing.