The Architectural Problem

The ocean and space domains expose a fundamental limitation in human-centered command and control. Machine speed, observational breadth, and persistence increasingly exceed human capacity, while distance, communication constraints, and contested connectivity prevent continuous intervention. The resulting challenge is not simply how to accelerate human decisions, but how to preserve command authority when human judgment cannot accompany every operational decision.

We are building increasingly capable autonomous machines while retaining command architectures that depend upon human availability.

Individual systems can navigate, observe, classify, and execute sophisticated tasks with diminishing human involvement. Yet much of the larger operational architecture remains organized around human supervision, approval, and intervention. We continue solving increasingly difficult technical problems without confronting the big constraint that connects them all.

An autonomous platform does not constitute an autonomous operational architecture.

A single autonomous platform operating under intermittent human supervision is one thing. Hundreds of distributed systems operating across enormous distances, through contested communications, and over extended periods present a fundamentally different problem.

The next breakthrough must not simply be a more capable machine. It must be an architecture that allows distributed machines to operate coherently within human-established boundaries, but outside continuous direct supervision. Two forces make this transformation increasingly necessary: the properties of machine operations and the tyrannies of the operating environment.

The Three Properties of Machine Operations

Three properties distinguish machine operations from traditional human-centered processes: speed, observational breadth, and persistence. Each independently challenges continuous human participation. Together, they expose the limits of organizing machine operations around human cognition.

Speed. Machines can observe, correlate information, evaluate options, and execute permitted actions at speeds exceeding human reaction and deliberation. The consequence is not simply faster decisions. Operational cycles compress below the time required for meaningful human intervention. A system may complete an entire observe-to-effect sequence before a human can understand what is occurring, much less influence the next action.

Observational breadth. Modern sensing systems measure the environment through radar, acoustic, optical, infrared, radio-frequency, magnetic, and other modalities. Each sensor observes a particular portion of the operating domain within its viable sensing volume and environmental limitations. Their combined observations generate information and correlations that humans cannot continuously perceive, integrate, or interpret at comparable scale.

Persistence. Machines do not require sleep, experience human fatigue, or lose continuity during personnel changes. Within their physical and technical limits, they can maintain observation and computational processes across hours, days, or considerably longer periods. Persistence allows machines to recognize patterns and changes across periods that human operators must divide into shifts, reports, and reconstructed understanding.

These properties do not depend upon one another, but they do multiply. Speed alone can exceed human reaction windows. Observational breadth can overwhelm human cognition even when time is available. Persistence can sustain computational activity beyond the practical limits of continuous human attention.

Alone or together, these properties challenge the assumption that a human must occupy every transition in the operational process. And the environment compounds the problem.

The (Three) Tyrannies of the Operating Domain

The machine properties describe capabilities. The environmental tyrannies describe constraints. They are independent dimensions, and each property can interact with any of the tyrannies. The ocean and space impose three particularly consequential constraints: distance, volume, and access.

The Tyranny of Distance

Distance separates operating systems from commanders, supporting infrastructure, and one another. It complicates communications, coordination, logistics, and timely intervention.

At the ocean surface, distributed forces may operate across enormous geographic areas, far from supporting infrastructure and under communications conditions that cannot be assumed reliable. Satellite connectivity extends reach, but interference, interruption, contested networks, and operational restrictions can prevent continuous contact.

Beneath the surface, the problem becomes more severe. Seawater restricts conventional radio communications, while acoustic alternatives impose significant limitations on bandwidth, latency, and reliability.

Space extends these constraints across distances where even perfectly functioning communications cannot overcome the speed of light.

A communications connection is not the same thing as an ability to intervene in time.

The Tyranny of Volume

The ocean is not merely an enormous surface area. It is a vast, dynamic, three-dimensional operating environment whose physical characteristics change with depth, temperature, salinity, geography, and time.

Comprehensive observation is impractical. No sensor sees everything, and meaningful understanding must be constructed from incomplete measurements distributed across an immense physical volume.

Volume describes the domain. Observational breadth describes our ability to measure and understand it. A sensor possesses a viable observation volume determined by its technology, environment, and operating conditions. But that sensor is merely a tool. Its coverage does not define the extent of the domain. Improving sensors expands our capacity to observe the environment. It does not reduce the environment itself.

The battlespace does not become smaller because our sensors become better.

Space presents this same challenge across vastly greater operating volumes. The disparity between the extent of the environment and the region that can be meaningfully observed becomes extraordinary.

The Tyranny of Access

Access concerns the ability to physically reach, inspect, maintain, modify, recover, or intervene in an operating system.

At the ocean surface, access may be restricted by distance, weather, geography, or operational conditions. Underwater, physical intervention becomes substantially more difficult, frequently requiring specialized vessels, equipment, and considerable time.

In space, access may be prohibitively expensive, severely delayed, or effectively impossible.

Access is distinct from communications. A system may remain connected while being physically inaccessible. Conversely, it may be physically reachable while communications are unavailable.

The greater the difficulty of access, the greater the importance of a system’s ability to sustain safe, functional, and authorized operations without immediate human assistance.

Operating domain Dominant environmental constraints
Ocean surface Primarily the tyranny of distance.
Subsurface The combined tyrannies of distance, volume, and access.
Space All three tyrannies to the nth degree.

The three properties expand what machines can accomplish. The three tyrannies increasingly constrain how humans can participate. Their interaction creates an architectural problem that cannot be solved merely by improving an individual autonomous platform.

Capable Machines, Inherited Constraints

Recent demonstrations reveal the tension between advancing machine capabilities and established models of human supervision.

In October 2026, DARPA reported a demonstration involving four autonomous ground vehicles operating together under its RACER program at Fort Knox. The vehicles independently navigated difficult terrain, searched for simulated threats, and coordinated their movement. Yet when a potential target was identified, human approval remained necessary before engagement.

The machines could navigate, observe, and detect. At the consequential decision point, the system returned to a human. This was a deliberate design choice, and the demonstration represented meaningful technical progress. It also illustrates a familiar boundary: increasingly capable autonomous functions organized around a human decision point.

The following month, the Royal Navy announced a £90 million investment framework for Atlantic Net, integrating autonomous systems, underwater sensors, communications, remote operations, and data services across the North Atlantic.

Part of the broader Atlantic Bastion initiative, Atlantic Net represents an important step toward persistent, distributed maritime surveillance. It recognizes the advantages of integrating crewed and uncrewed systems into a coherent operational capability rather than merely deploying additional independent platforms.

These programs address different missions, but both illustrate the rapid advancement of machine capability. Neither public announcement establishes how a complete distributed operational architecture maintains decision-making through extended periods when meaningful human supervision is unavailable.

That is the unanswered architectural question.

We have become increasingly proficient at building machines capable of performing sophisticated functions independently. But the publicly described operational concepts still frequently preserve the human as a required supervisory or decision-making element at consequential transitions. Some boundaries are necessary safeguards. Others may reflect inherited, perhaps outdated assumptions about how command must function.

We may have confused retaining authority with retaining our place inside the process. What happens when we stop designing around that assumption?

This is not an argument for abandoning human judgment. It is a challenge to the premise that human judgment must remain continuously present inside the operational process for command authority to remain meaningful. This becomes clearer when we examine how machines actually process observations and produce effects.

From Human Orientation to Machine Computation

John Boyd’s observe–orient–decide–act (OODA) model has long provided a useful framework for understanding operational decision-making.

But that familiar model becomes less descriptive when computation, rather than continuous human orientation, occupies the critical path between observation and action. A useful way to describe this emerging machine-centered process is:

OCDE: Observe → Compute → Decide → Effect.

  • Observe collects measurements and changes through available sensors and information sources.
  • Compute correlates observations, maintains operational state, interprets conditions, evaluates uncertainty, and assesses available options against established constraints.
  • Decide selects among authorized actions or determines that the available information, confidence, or authority is insufficient to proceed.
  • Effect executes an authorized function, producing consequences that become part of subsequent observation and assessment.

The critical distinction lies in computation. Compute is not simply faster human orientation. It allows information to be processed across sensing modalities, locations, and timeframes without requiring a human to continuously assemble and interpret the operational picture.

In some circumstances, an entire computational cycle can unfold in the blink of an eye. While in others, the same process may continue across days of unattended operation.

The result is not merely a faster version of the traditional human decision cycle. It is a system in which computation can occupy the operational critical path while human judgment establishes the boundaries governing its activity.

The change is architectural, not merely an acceleration of the existing process. And the ocean’s physical constraints increasingly demand that this capability accompany the operating force.

Forward, but Bounded

The architectural response is to move sufficient computational capability forward with autonomous operating systems rather than depend upon continuous reachback to remote human supervision. This need not mean a single computer, processor, or centralized platform. The capability may be distributed across operating systems and supporting infrastructure. What matters is that essential computational functions remain available when communications are delayed, degraded, interrupted, or denied.

A forward computational architecture can maintain local awareness, integrate observations, preserve operational continuity, and coordinate permitted functions without requiring constant direction from a remote command center. But its authority must remain bounded.

We move the capacity to decide forward. We do not move the authority to redefine what is permissible.

Human commanders establish objectives, priorities, permissions, restrictions, acceptable risks, and conditions requiring renewed authorization. Machines may adapt how they accomplish permitted tasks. They may evaluate changing conditions, select among authorized alternatives, and adjust their methods when appropriate. They may not independently invent new mission objectives or enlarge the authority granted to them. This distinction separates operational adaptation from the exercise of command authority.

When a system encounters circumstances beyond its authorization, the architecture must provide an appropriate, predefined response. Depending upon the mission and conditions, that may involve stopping, entering a safe state, continuing narrowly authorized protective functions, or requesting further direction when communications permit. The system cannot treat the absence of a human response as permission to create new authority.

This is particularly important in considering DoD Directive 3000.09, Autonomy in Weapon Systems.

The directive requires “appropriate levels of human judgment” over the use of force. It does not impose a universal requirement for human approval of every individual engagement. Instead, it establishes requirements concerning appropriate judgment, authorized employment, system performance, verification, validation, and operational suitability.

A forward, bounded computational architecture may be compatible with these requirements, provided its design, validation, authorization, and employment satisfy the applicable standards. That compatibility must be demonstrated, not assumed. The architectural principle remains straightforward.

The machine may adapt how it accomplishes the mission. It may not independently redefine the mission. This is the essential separation between independent execution and independent authority.

Human Command: Authorship, Enforcement, and Audit

If continuous human participation becomes impractical, command authority must be expressed through mechanisms that remain meaningful during the commander’s absence from individual operational cycles.

This changes where human judgment operates. Three responsibilities become central.

  • Authorship. Commanders establish the authorized decision space. They define objectives, priorities, permissions, restrictions, acceptable uncertainty, and the circumstances requiring renewed authorization. Human intent must be translated into instructions and limits that computational systems can implement reliably.
  • Enforcement. The architecture must recognize and respect those boundaries. A restriction has little operational meaning if the system cannot identify when it applies or reliably prevent unauthorized behavior. Verification, validation, operational testing, uncertainty management, and predictable stopping behavior become central elements of command.
  • Audit. The architecture must preserve sufficient information for responsible humans to reconstruct consequential actions, examine outcomes, identify deviations, and revise future authorizations. Audit supports accountability, correction, and organizational learning.

Together, these responsibilities establish a different relationship between the commander and the operating system. The commander retains authority over the permissible decision space without necessarily participating in every computational transition that occurs within it. That is not a reduction in command responsibility.

It places greater demands upon commanders to understand the capabilities, limitations, reliability, and operating boundaries of the systems they authorize. The commander must anticipate conditions under which intervention will be unavailable and ensure that the system’s authorized behavior remains appropriate under those conditions.

The human may make fewer immediate decisions, but the decisions that remain become more consequential.

Human judgment has not disappeared. Its role has shifted from continuous operational participation toward the authorship, enforcement, and evaluation of the space in which machines operate.

The Ocean as the Proving Ground

The ocean provides an unusually demanding environment in which to develop and validate this architecture. Its immense three-dimensional volume, incomplete observability, changing physical conditions, restricted access, and unreliable communications expose the weaknesses of systems dependent upon continuous remote interaction. These constraints are not temporary shortcomings that will disappear with the next generation of technology. They are fundamental characteristics of the operating domain. The ocean therefore offers a compelling environment for examining distributed systems that must sustain authorized operations without continuous human direction.

The objective should extend beyond demonstrating that another vessel can navigate independently, maintain surveillance, or execute an assigned function. We must examine whether the larger architecture can sustain coherent operations under realistic environmental constraints.

  • Can distributed systems maintain useful operational awareness despite incomplete observations?
  • Can they sustain permitted functions across communications interruptions without expanding their authority?
  • Can they recognize circumstances that exceed their authorization and respond appropriately?
  • Can their consequential actions be reconstructed, evaluated, and improved?
  • Can these capabilities be demonstrated across an integrated system rather than merely within an individual successful platform?

This is an incomplete list of architectural questions.

Better sensors, faster processors, improved navigation, and more capable autonomous vehicles are necessary advances. But they do not, by themselves, establish that the architecture can operate coherently without continuous human participation. Nor should the measure of success simply be the number of functions performed autonomously. The more consequential measure is whether distributed systems can sustain useful, coordinated, and authorized activity while remaining accountable to human command.

The objective is not simply a better autonomous vessel. It is a better architecture for distributed machines operating under human authority.

The ocean offers something exceptionally valuable: a demanding environment in which the principles of bounded, distributed machine operations can be developed and tested against physical constraints that cannot be negotiated away. And the importance of getting that architecture right extends far beyond maritime operations.

The Space Implication

Space presents the same fundamental architectural challenge under far more severe conditions. Distance introduces unavoidable communication delays. Operating volumes expand across enormous regions. Physical access becomes extraordinarily difficult and, in many missions, effectively impossible.

The ocean and space are obviously not identical environments. Their physics, sensors, communications mechanisms, and operational objectives differ substantially. But the underlying architectural challenge remains familiar.

How does a distributed system maintain coherent, authorized operations when the people responsible for it cannot remain continuously present?

Robotic exploration, distributed scientific observation, remote infrastructure, and an expanding human presence beyond Earth will increasingly confront this question. As missions move farther from Earth, the assumption that every meaningful operational transition can be reviewed or directed from a terrestrial control center becomes impossible.

A spacecraft may need to preserve its functional integrity, interpret local observations, respond to changing conditions, and carry out authorized activities before further instructions can arrive. These demands do not diminish the importance of human authority. They require architectures capable of sustaining human-directed purposes without continuous human involvement.

This is why the ocean matters. Its three-dimensional volume, communications constraints, and limited physical access create an accessible environment in which many of the same architectural principles can be examined, challenged, and refined. The specific technologies will differ. The governing questions of authorization, adaptation, operational continuity, enforceable boundaries, and accountability will remain.

Solve the architectural problem across the ocean’s volume, and we may establish principles essential to humanity’s future operations in space.

The ocean may become the proving ground for a computational architecture whose most consequential applications have little to do with actual warfare.

Beyond Continuous Human Presence

We are making extraordinary progress in autonomous systems. Machines are becoming faster, more observant, more persistent, and increasingly capable of performing complex functions without continuous human participation.

Yet much of our architectural thinking continues to assume that human supervision can remain available wherever consequential decisions occur. The ocean exposes the limitations of that assumption.

Distance separates. Volume overwhelms. Access restricts intervention. Machine speed, observational breadth, and persistence continue to advance beyond the practical limits of continuous human participation. Space makes the consequences impossible to ignore.

The answer is not to surrender human authority or permit machines to redefine their own purposes. It is to develop forward, bounded computational architectures capable of sustaining operations within enforceable, human-established limits. That requires more than better autonomous platforms.

It requires reconsidering how command is expressed, where judgment must occur, how boundaries are enforced, and how distributed systems remain accountable when humans cannot continuously supervise their actions.

We are not running out of ways to make machines more capable. We are running out of ways to keep the human continuously present.

The ocean may be where we learn to solve that problem. And if we get the architecture right beneath the sea, its greatest consequence may be what humanity can accomplish far beyond Earth.

Principal References

  1. DARPA, Autonomous Vehicles Deliver Command-Controlled Robotic Lethality, October 27, 2026. https://www.darpa.mil/news/2026/autonomous-vehicles-deliver
  2. Royal Navy, £90 Million Investment for New Drones and Sensors to Protect Undersea Cables and Hunt Russian Submarines, November 10, 2026. https://www.royalnavy.mod.uk/news/2026/november/10/20261110-90-million-investment-for-new-drones-and-sensors
  3. Department of Defense, Directive 3000.09: Autonomy in Weapon Systems, January 25, 2023. https://media.defense.gov/2023/Jan/25/2003149928/-1/-1/0/DOD-DIRECTIVE-3000.09-AUTONOMY-IN-WEAPON-SYSTEMS.PDF