the disappearing interval

 A closing movement for “The Last Human Decision”

On August 12, 2026, in Huntsville, Alabama, the outgoing head of U.S. Space Command gave his last symposium speech and read out a list. Five capabilities the command says it will need between 2029 and 2033 to keep its edge over China in orbit: kinetic and non-kinetic space fires; means to counter an adversary’s low-orbit satellite constellations; refuelable satellites that can maneuver instead of husbanding the fuel they launched with; space domain awareness good enough for what the missile-defense world calls a “target-quality track”; and, third on the list, joint, integrated space command and control. Gen. Stephen Whiting was careful to name the genre. “This isn’t just a wishlist,” he said. “It’s a clear demand signal to several audiences.”

It is an interesting choice of words. Industry hears a requirement. Congress hears a justification. An adversary hears capability and intent. One transmission carries several messages simultaneously.

But one item deserves particular attention.

Whiting said the force must “enhance machine-to-machine connectivity to support decision-making and battle management of this global space fight on tactically relevant timelines.”

The important words may not be machine-to-machine. They may be tactically relevant timelines.

Because this essay has encountered that problem before.

In 1983, the Soviet Oko early-warning system told Stanislav Petrov that American missiles had been launched. The system was wrong. Petrov had something that increasingly sophisticated technology cannot guarantee:  time.

Time to look at the signal.

Time to compare it with what he knew.

Time to doubt the machine.

Time to decide.

More than two decades earlier, at NORAD, Air Marshal Roy Slemon confronted another extraordinary warning when the newly operational BMEWS system interpreted the rising moon as a massive incoming Soviet missile raid. He, too, occupied an interval between detection and action in which a human being could ask a profoundly important question:

Could the system be wrong?

The men mattered. Their judgment mattered. But something else mattered first.

The architecture gave them time to exercise judgment.

That distinction becomes important in orbit.

A contested space environment operates at velocities and distances that compress the interval between detection, interpretation and response. Sensors become faster. Networks become faster. Tracking becomes more precise. Machines exchange information directly because human communication increasingly becomes the slowest element in the chain.

There is nothing inherently sinister about this. Much of it is necessary. Machines can integrate sensor streams, maintain tracks and distribute information at speeds no human organization could match.

That is Licklider's promise still alive: machine and human doing different things well.

But there is another possibility.

As the system becomes faster, the machine may cease merely to provide information within the human decision window and begin determining the size of that window itself.

That is a different threshold.

The question is therefore not simply whether a human remains “in the loop.” That phrase may eventually tell us very little. A person can formally remain inside a decision architecture while possessing so little time, context or practical ability to intervene that the human role becomes largely procedural.

The better question is:  How much meaningful time remains between machine certainty and irreversible action?

Whiting does not tell us that humans are being removed from space battle management. His words actually preserve the opposite interpretation: machine-to-machine connectivity will support decision-making. That may prove to be exactly what is intended—machines managing the velocity of information while humans retain authority over consequential actions.

But the tension should not be dismissed.

Petrov's world contained minutes.

Slemon's world contained minutes.

The architecture Whiting describes is being designed for a domain in which minutes may be an eternity.

And that brings us back to the question underneath this entire essay.

For most of the history of automated warning, we have treated the human as the final safeguard against the machine's mistake. Sensors could fail. Algorithms could misclassify. Communications could distort. The human receiver remained the place where context, uncertainty and judgment could interrupt the chain.

But what happens when the speed required for the system to succeed becomes incompatible with the time required for that safeguard to function?

We should be careful here. The answer is not to romanticize slow human decision-making. Humans misinterpret signals too. They panic. They carry biases. They misunderstand adversaries. And in some circumstances hesitation can be as dangerous as haste.

The objective is not simply to preserve human delay.

It is to preserve meaningful human judgment.

That may prove to be one of the central design problems of machine-speed command and control: not merely making the system faster, but deliberately engineering places where speed stops.

A pause.

A threshold.

A requirement for corroboration.

A point at which uncertainty is allowed to remain uncertainty.

A place where the architecture itself says: not yet.

Petrov's pause was partly an artifact of the technological world in which he lived. We should not assume the next generation of systems will give it to us accidentally.

If we believe the pause matters, we may have to design it deliberately.

That is what makes the demand signal from Huntsville consequential.

Space Command is asking industry to build systems capable of operating on timelines increasingly inhospitable to unaided human cognition. That requirement may be unavoidable. It may even be essential.

But somewhere between the sensor and the weapon, between detection and decision, between machine speed and human responsibility, another requirement belongs on the list:  preserve the interval in which someone can still say no.

Petrov inherited that interval.

Slemon inherited it.

The next generation may have to build it.

 

— Samuel R. Lammie

 

A note on how this was made: This piece emerged through an iterative conversation among a human and two AI systems. I brought the underlying framework, questions, and professional perspective; Claude helped develop the initial argument and draft; ChatGPT challenged several of its stronger claims and helped reshape the piece around what I think is the more defensible question: not whether machines will replace human decision-makers, but whether the systems we design will preserve enough time for meaningful human judgment. I made the final choices. In a piece about human judgment in increasingly automated systems, that process seemed worth making visible.