Sensor-to-shooter is a decision, not a wireby Bharath Gopalaswamy and Daniel Dant
|
| Which interceptor takes which missile, which are left for later tiers, and which apparent launches are decoys or debris becomes the laser’s assignment problem scaled by a thousand, compressed into seconds, with no time for a human decision on each shot. |
Both statements were received as news about weapons. They were really news about decisions, marking the moment Golden Dome stopped being an argument about whether and became an argument about how. We have followed that argument through this series: in May, we argued that strategy is easy and logistics is hard; in June, that how the shield is financed will shape what gets built; in July, that the arsenal will have new, largely private owners; and most recently, that a glacier collapse in Nepal, watched by satellites from four countries and still unwarned, shows the layer between the camera and the person at risk is the layer Golden Dome most needs and has least examined. This article is about that layer, and about what it will take for the industry expected to build Golden Dome to build it.
Let us be clear. Golden Dome is the right program. The threat is real. An architecture with sensing and the first interceptor tier in orbit is the only one that can engage early enough to matter. And the three-year compression ordered in January 2025 is a feature, not a flaw.
What follows is offered by two people who want the program to succeed and think its hardest problem is being under-described.
What the recent past teaches
Three episodes show what happens when the decision layer fails. In December 2023, drones overflew Joint Base Langley-Eustis on 17 consecutive nights; the base detected them but could not establish intent, and no one held the authority to act. On January 28, 2024, a one-way attack drone struck Tower 22 in Jordan, killing three soldiers and wounding more than 40, after arriving as a friendly drone was returning; no interceptor was fired. In the Red Sea, warships defeated hundreds of Houthi drones with interceptors costing a hundred times more than the targets and lost the exchange while winning every engagement. Authority, discrimination, and assignment: three failures, all after detection.
| The companies expected to build Golden Dome are more ready than the public debate suggests, and more constrained than it recognizes. |
Directed energy can improve the cost exchange, but it makes the decision problem even sharper. A laser engages one target at a time, must dwell for seconds, and cannot be re-tasked until that engagement is complete. Consider FIREFLY, the palletized counter-drone laser KBR’s Kord Technologies unveiled at AUSA in 2024, which scales from 5 to 40 kilowatts and can be cued by FAAD, MEDUSA, and standalone radars. Its specification describes what it can detect, track, and defeat, but not what it should engage first. Which target it engages first, and which it leaves to another effector, is decided upstream, by whatever sits on the arrow between sensor and shooter. The effector is ready but it’s the decision architecture that makes it effective or not.
Now move that problem to orbit. A space-based interceptor is also a single-shot effector with a narrow window; it can reach a boosting missile only while the missile is under thrust and only when its orbit brings it within range. In a large-scale attack, hundreds of launches will be detectable and targetable by multiple interceptors. Which interceptor takes which missile, which are left for later tiers, and which apparent launches are decoys or debris becomes the laser’s assignment problem scaled by a thousand, compressed into seconds, with no time for a human decision on each shot. Langley, Tower 22, and the Red Sea were the rehearsal, and Golden Dome is the performance.
Machine speed is a policy, not a setting
General Guetlein’s description of scalable automation is correct as far as it goes: no human chain of command can act on thousands of tracks in the time a hypersonic attack allows. But “as the threat accelerates, we move toward automation” describes a slider without deciding who moves it. Who decides the threat has accelerated? Which engagements does the system then take on its own, and which does it refuse? What happens when a track looks like a launch but also like a debris event, or when an adversary has learned what moves the slider? These are questions of authority as much as engineering, questions nuclear command and control has confronted for 70 years and resolved through doctrine and law, not software. Israel’s Home Front Command offers a useful model: a radar detection can trigger a warning in seconds with no human in the chain because the rules were written in advance and rehearsed against thousands of real launches. Golden Dome’s space tier will need the same discipline: rules written in advance for a machine in orbit and exercised long before they are needed.
What industry is waiting for
Here the view from inside industry matters, because the companies expected to build Golden Dome are more ready than the public debate suggests, and more constrained than it recognizes.
They are ready. The Missile Defense Agency’s SHIELD contract, a ten-year vehicle with a $151 billion ceiling, has qualified more than 2,400 companies, ranging from the largest primes to startups. The Space Force has put 12 companies on agreements worth up to $3.2 billion to prototype space-based interceptors by 2028. The industrial base has been mobilized at a scale without precedent in missile defense, and industry has answered.
They are also constrained because of four reasons anyone who has sat in a capture meeting this year will recognize. First, a seat is not a task. SHIELD gives 2,400 companies the right to compete, but it does not tell them what will be competed or when. A prime cannot staff a team against a ceiling, only against a requirement. Second, the awards so far have bought things that can be photographed: interceptors, buses, launch. Search those awards for the rule connecting sensor to shooter, the model that rehearses it, or the test that proves it, and none appears. The government has bought the “eyes” and is buying the shooters. It has not yet bought the decision.
Third, the program has warned that a fiscal 2027 impasse could halt it. No company will build a decision architecture on its own money against a program that may stop in October. Fourth, the seam itself is the problem. The decision layer touches every contractor’s data and every effector’s envelope, and no single prime should own it. Left to one provider, it becomes a single point of failure and a source of leverage over the government. It gets built only if the government specifies the decision layer as an open-interface component and assigns integration to a vendor-neutral integrator whose job is to make everyone else’s hardware work together. That is precisely what the systems engineering and integration contractors on SHIELD exist to do. They are waiting for the requirement.
What the decision link would contain
Four things must exist before the first interceptor is fielded. None of them are hardware. All are capabilities the Missile Defense Agency and the Space Force already buy through existing contracting vehicles. What is missing is not a contractor, but a requirement.
| Congress could add one requirement at no additional cost: no funds may be obligated for interceptor procurement until the program delivers to the defense committees a description of the decision architecture. |
The first is pre-delegated engagement authority, written down for each tier in terms an operator can apply and a court can read. The second is rules of engagement settled and approved before the system is fielded, not improvised in the fight. The third is a rehearsal regime. The White Sands tests were designed, in Lieutenant General Brian Gibson’s words, to validate “sensor-to-shooter linkages.” That word matters. A linkage is a wire; a wire cannot decide. Two live tests a year prove only that the wire carries data. Proving the decision is right requires exercising it far more often than live tests allow, under independent scrutiny, before anyone trusts it with a live shot. Live tests prove the plumbing; simulation proves the judgment. The fourth is an open interface paired with a public account of who holds authority, what triggers its delegation, and what the system will never do on its own.
The line Congress could add
When the fiscal 2027 appropriation is written, Congress could add one requirement at no additional cost: no funds may be obligated for interceptor procurement until the program delivers to the defense committees a description of the decision architecture, meaning the engagement authorities at each tier, the conditions for their delegation, the limits placed on its behavior, and the exercise regime that validates its behavior. Every item is work the program must produce anyway and work the contractors already on its rosters can perform once asked. Specify the decision, then buy the shooters. It would also give industry what it has lacked since the SHIELD awards: a requirement to build against.
The satellites will be the visible part. The interceptors will be the expensive part. But the decision between them will determine whether either was worth buying. It is also the one part that can be designed, tested, and proven before the first hostile launch. Sensor-to-shooter is not a wire, on the ground or in orbit. It is a decision. Industry is ready to build it. The government must now define it.
Note: we are now moderating comments. There will be a delay in posting comments and no guarantee that all submitted comments will be posted.