What follows is a discussion of the human-machine intelligence and its radically changing relationship to the medium of war, that is, objects in spacetime. Thanks to advancements and scaling in technology like drones and AI, we are seeing exponential growth in (1) a given actor’s range, the distance across space and time at which the actor may causally touch an object. Further, we are seeing growth in (2) the categories of possible relations and the types of objects which may be affected by that actor. And we are seeing that (3) a given action is carried out at greater speed and force and lower cost than before.

We may visualize these factors together as growth in the actor’s causal “neighborhood”. Or, we may say that the space between thought and effective action is collapsing. The relations between actors and objects in 4D space are becoming, in a word, extensible. The exponential growth in that extensibility means that the aforementioned relations will in several years time become nearly arbitrary, and the battlespace will be live-editable rather than merely traversable. From the commander’s perspective, war will feel like a global realtime strategy game, with fog and friction playing a diminished role in his perception, even if they lurk in new spaces like the AI models themselves. And once brain-machine interface, digital twinning, self-improving technologies, and self-assembling factories take hold alongside AI and sensor fusion, there will be almost no limit to what a single actor can accomplish.

To elucidate all this, first we need to agree on a series of definitions and explanations, some of which are admittedly of my own formulation and I intend to use freely in future articles.

Principal: A biological, rational will.

Agent: An autonomous machine intelligence. In theory, capable of everything a principal is capable of aside from being assigned philosophical concepts of free will and personhood. That is, machine intelligence is not to be regarded as a friend or an enemy or having any rights, since the assignment of these properties to an arbitrarily replicable/scalable digital entity would have suicidal implications for humanity.

Actor: A principal or agent. All agents and principals are actors.

Objects: In a military context, includes actors, drones, equipment, infrastructure, or anything else that may be usable or targetable. Distinct from “the object of war”, which in that singular form refers to ultimate rather than intermediate uses and aims of war.

[Drone] fleet: A group of drones/weapons potentially controlled by agents but ultimately commanded by principals. Includes all domains: air, land, sea, subsurface, and space drones. Intended to be distinct from “swarm”, which necessitates simultaneous or localized use.

[Drone] squadron: A military unit encompassing a drone fleet, the principal(s) in command, and supporting elements.

Tactician: A principal with command over a drone fleet.

War: The organized use of violence by a group of principals intended to impose political will or destruction upon another such group.

Spacetime: Space and time treated as a unified medium.

Range: Distance across spacetime at which action may be effected.

Resolution: The fineness of distinction at which an actor can perceive, decide, and act within spacetime. Resolution is in some sense opposed but also complementary to range and is deepening as range is extending. In time specifically, resolution covers latency, [OODA] looping, and micro-scales. At present, AI is greatly improving temporal resolution, and sensor fusion is greatly improving spatial resolution.

Lens: (Bear with me on this multiparagraph explanation…) The coming unified or federated digital system of systems which will include many or all of the following: human- and machine-readable and -usable digital twin, common operating picture (COP), sensor fusion, simulation, analysis, course of action suggestion, mission planning, mission control, drone programming, drone flight, meshing, secure messaging, training, wiki / lessons learned / after action reports, payments, R&D, design, 3D printing, local offline use, cloud use, and more. The point isn’t that all of these features need to be actually instantiated on, say, a drone pilot’s field laptop, which would be prohibitively resource-intensive.

Lens (continued): Rather, the skeleton only needs to be in place. The disparate parts simply need to be commonly accessible, seamlessly integrated, and so forth, which is why the creation of common open standards for hardware, software, and communications is so vital now. The idea here is that commanders, soldiers, analysts, political leaders, and military industrialists (both engineers and management) are all using the same system to do their jobs and all be part of one huge feedback loop (with whatever minimal compartmentalizing of information is needed); and also that AI is infused throughout the whole system locally and at scale, analyzing enormous amounts of data and suggesting improvements.

Lens (continued): At present, there are existing platforms which each take on multiple features from the list I gave, and which if plugged in together with a common protocol could already encompass most of the listed features. Ukraine’s Delta platform is one, which focuses heavily on the COP side and is already a sort of software suite of suites. I’m aware of efforts to plug 3D military drone simulators into Mission Planner, an open source drone programming and drone control software suite that natively only renders in 2D. As these platforms continue to unify, we will approach the creation of the Lens. For the human user, 2D will give way to 3D, which will give way to VR, which will give way to AI VR- / world-generation-on-demand, which will eventually even give way to direct brain-machine interface like Neuralink.

I must here also elaborate on the numbered items in my intro paragraph:

(1) Growth in a given actor’s range: a competent “midrange strike” (mid strike) drone team in Ukraine can now hit targets at somewhere between 80 and 300 km, depending on the employment of the current “hot” technologies of the day, such as Starlink, autonomous computer vision, balloon-dropped drones, etc. This is a doubling of range year-over-year, and we see similar growth in the max effective range of close and deep strike teams. If we extend this trend out, the mid strike team will be able to strike anywhere on Earth in six to eight years. The trend is virtually guaranteed, since no new technology is needed to enable it; all that’s needed is the scaling and integration of existing technology and systems. Transoceanic, repeating, and mothership drones are already here.

Further, teams will soon be able to preplan strikes as far into the future as they like and watch them unfold autonomously. Why do I focus here on mid strike teams? Because unlike the deep strike teams, which are centrally directed, the mid strike teams often operate with relative independence, and can achieve serious effects for tens of thousands of dollars. Ballistic missiles, too, could already strike anywhere on earth with little warning; this is something new altogether.

(2) Growth in the categories of possible relations and the types of objects which may be affected by that actor: a given actor may control any drones to achieve any effect on any objects. He may control drones of arbitrary magnitude (size and capabilities), multitude (single or millions), domain (sea, subsea, air, land, space, multidomain); he may accomplish any mission (strike, ISR, mine laying, sabotage, combat engineering tasks, information/psychological warfare, assassinations, etc.); he may target single or many objects at any depth in any country.

As long as the actor has the right robots and a means of controlling them directly or indirectly (as by giving orders in advance, executed autonomously), he may wake up any morning and choose violence against anything he wishes. And he may do so against “future” objects in the sense that he can preplan autonomous attacks, or against “past” objects insofar as the enemy cannot keep up with him (more on that below).

(3) Actions at greater speed and force and lower cost than before: As mentioned above, within several years the “mid range” strike team’s spatial range will eclipse the globe. The tactician emerging from these teams is also seeing his drones steadily increasing in speed (for example, jet-powered drones are now moving at hundreds instead of tens of kph); data is interpreted and analyzed for him at machine speed via AI, aiding in both mission planning and lessons learned; his defenses autonomously move to intercept the enemy’s own rapid strikes faster than a human could process what’s happening.

He effectively lives in the future of a foe who doesn’t keep up with his increase in both speed and temporal resolution. And all of this at scale–yes, individual drones may not carry enormous payloads, but he can synchronously control as many of them as he likes via “swarming” protocols. All of it at no more cost per effect than we already see today, as robots are only getting cheaper and better (AI in particular, at least 100x cheaper year-over-year).

The principal and the agents he controls together are actors which wage war as part of a drone squadron, directing drone fleets against enemy objects at ranges which will soon eclipse the entire globe. For now, we can imagine this principal as a tactician whose role will emerge from the elite pilots of small, independent drone teams, the other members of which are glorified helpers who physically move and emplace equipment and help with ancillary functions like communications or navigation. With the advent of robotics, autonomy, and various communications technologies, we are seeing both that the pilot-become-tactician need not be physically colocated with the rest of the team, and indeed also that the rest of the team itself may soon become obsolete as humans are replaced by a chain of robots delivering, repairing, and resupplying other robots. Therefore the tactician is the fundamental human unit of war, and his agents and individual drones are the fundamental units of machine combat power.

The tactician could be an enlisted soldier now, but must likely at some point emerge as an officer class, eventually possibly being akin to an age-of-sail ship captain at sea, entrusted with broad authority by the political and military apparatus, since he must make rapid decisions even in communications blackouts, potentially regarding the employment of the destructive potential of a hundred thousand or more drones. As I’ve described in prior articles, we can infer an immense selection pressure here for intellect. A dumb grunt could never handle the tasks of such a role, much less be entrusted to wield power wisely. And if you have two tacticians, one smarter and one dumber, both in charge of ten drones, would you not take all the drones from the dumber and give them to the smarter, so long as he has the technical means and cognitive capacity to employ them well enough?

We see an actual parallel to this happening in Ukraine, where effective units are awarded points based on combat results, allowing them to procure more drones and be even more effective, strangling the less effective units of resources. Effective units rapidly rise up the echelons, transforming from battalions to brigades and so on. Commanders are similarly promoted, often left in command of the same growing unit. The same may happen to individual drone tacticians who prove themselves, and the emergent structure will begin to intentionally select and train to account for this.

The emergence of the tactician, taken together with the numbered items above and the fact that all this will be experienced remotely via the Lens I described, means that this new class of soldier will comprise a set of disembodied wills/intellects “editing” the battlefield, projecting their consciousness across the world to kill remotely. Perhaps you’ve heard it said that “The map is not the territory.” The digital twin nature of the Lens means that the map is become the territory, and actions taken on it are the same as taking actions on the territory, so long as sufficient robots and edge compute are assigned to execute on the principal’s will. Actually, from the principal’s perspective, the map will be larger than life, since the sensor fusion and AI analysis will exceed anything his own senses and intellect could have accomplished without it.

A sufficiently developed Lens could also simulate future states, for example gaming out entire wars in advance with much less hand-waving and notionality than is presently required for such purposes. With such capability, one can virtually operate in future battlespaces as-yet unperceived by an enemy of lesser capability in the way an online bullet chess player may have a flash of understanding and stage multiple premoves, watching as his opponents reality slowly catches up move by move. This applies to noncombat usage, too, of course. Everything from equipment (R&D) to factories to entire supply chains may soon be modeled accurately, tweaked, and self-assembled into reality.

Some implications of all this for the tactician are worth exploring. First, he will need to have the creativity, discipline, initiative, and fieldcraft of a special operations soldier and an intelligence officer combined. Since the weapon ranges at play are so great, and since the enemy has similar weapons, and since the tactician need not be colocated with his weaponry, it follows that various methods will be employed by him to reduce his chance of being targeted, such as driving civilian vehicles or using faraday bags to obfuscate his digital/comms trail (again, already seen in Ukraine). The enemy, knowing this, will hunt him accordingly, relying on intel to drive their targeting. Tacticians will engage in massive machine-on-machine duels, but always against the background of the underlying hunt for the principal, which follows from the definition of war itself.

On another note, friction and fog arose because the world resisted the actor’s will at human speed. Orders traveled at the pace of horses or ships, and a wide gap existed between abstract plans, physical execution, and the update loop between them. At machine speed and in the Lens, the character of friction changes, since the world’s connectivity, not merely its contents, becomes the medium of action. Further, the plan therein is the execution from the actor’s perspective, so long as he wants it to be and isn’t just running a training simulation.

That being said, there is another friction, an internal one, which emerges as a result of design choice in equipment, AI models, organization, the development of doctrine and requirements, etc. I won’t belabor the point too much since I’ve made it in my previous articles, but I’ll just say here that I hope for the triumph of decentralization, distribution, open source, and open architecture, since they can massively reduce this internal friction if they are (ironically) nurtured by the centralized powers that be. I bring it up again to apply this idea to the construction of the Lens - if Anduril or Palantir alone gets to oversee it, progress will be stymied and combat effectiveness will drop.

If, on the other hand, the platform is agnostic to which AI models or drones are used, we will be much better off. Such a Lens could be dual use, even beyond the military-industrial functions I previously elaborated on. It could be used for agriculture, retail, search and rescue, firefighting, and more. We need a Lens as open and far-reaching in its effects as Linux.

Because reach is becoming effectively unbounded, war is becoming accessible enough to wage everywhere, all the time. And combined with the current strength of the defense and the difficulty of achieving total victory, we can expect proxies and terrorists and hybrid and deniable operations to continue to proliferate. As a result, I expect our special operations forces will drift ever further toward something like a global FBI, preempting and preventing and responding rather than storming and seizing. On the home front, with threats multiplying and the attack surface of our complex society ever-growing, I again call for decentralization as a more effective answer than a police state, as stated in previous articles.

Clausewitz said “Like an obelisk toward which the principal streets of a place converge, the strong will of a proud spirit stands prominent and commanding in the middle of the art of war,” and that remains true across the totality of the war. The will of nations, leaders, and soldiers will still decide the issue. But within any given machine-speed micro-engagement (like an interception of an incoming drone), the human will is becoming locally irrelevant, and the non-cognitive functions of humanity will only be useful in war in narrower and narrower bands of action; will is moving upstream in the kinetic chain.

I’ll close with a couple tangential asides, since this isn’t a formal paper– The only eventual foil to the actor’s imminent ability to reach everything everywhere may be the eventual expansion of humanity to the stars, growing the potential action space in tandem. Although, if we can deliver humans to a location, we can probably deliver munitions too. Also, pretty funny that anyone thinks riflemen or human-crewed tanks have any relevance in the doctrine one can see emerging from everything I’ve described above.