Swarm Forge Is Testing the Right Capability. It's Missing the Right Doctrine.
What the Pentagon's drone swarm crucible reveals about the state of autonomous C2 — and what it still hasn't asked.
Swarm Forge Is Testing the Right Capability. It's Missing the Right Doctrine.
What the Pentagon's drone swarm crucible reveals about the state of autonomous C2 — and what it still hasn't asked
In January 2026, the Pentagon released footage of what it called the first kinetic drone swarm demonstration on American soil. A single operator. Three drones. Three targets. Simultaneous strikes, seconds apart, with no direct joystick control. The drones flew themselves to their targets, guided by onboard software rather than a human hand on a controller.
The demonstration was part of Swarm Forge — one of Secretary Hegseth's "pace-setting" projects, now being run by the Chief Digital and AI Office (CDAO) with quarterly Crucible events designed to validate and field autonomous swarm packages in 90 days or less.
It was an impressive demonstration. It was also three drones.
What Swarm Forge Is, and What It Requires
The Swarm Forge Crucible is not a research program. It is a validation and fielding pipeline. CDAO's solicitation is explicit: the goal is to deliver "validated swarm packages" — integrated platforms, mission software, coordination logic, interfaces, and tactics — ready for transition to operational units in 90 days or less.
The technical bar is significant. Participants must demonstrate:
- Autonomous, heterogeneous swarming across multi-vendor UAS platforms — not just different models from one vendor, but genuinely multi-vendor, multi-capability formations
- End-to-end autonomous completion of ISR and "Find, Fix, Finish" mission sets
- Operation in denied, degraded, intermittent, or limited communication (DDIL) environments
- A minimum of four UAS in simultaneous operation at the Crucible event this summer
The heterogeneous, multi-vendor requirement is the critical one. It rules out the easy path — building a tightly integrated single-vendor stack and calling it a swarm. The Pentagon is explicitly demanding what the commercial drone industry has consistently failed to deliver: genuine interoperability at the command-and-control layer.
The Doctrine Gap No Solicitation Has Named
Swarm Forge, DIU's Orchestrator Challenge, and DARPA's containerized constellation RFI are three parallel programs attacking the same underlying problem from different angles. The DoD is spending hundreds of millions of dollars to answer one question: how do you command a large, heterogeneous, autonomous force in a contested environment?
What is notable is what none of the solicitations have named directly: doctrine.
Every Crucible participant will bring a software stack. They will demonstrate impressive autonomous behaviors — collision avoidance, target handoff, formation maintenance. What they will struggle to demonstrate is coordinated mission execution under attrition — the ability for a formation that has lost nodes to continue executing coherently, without the surviving agents optimizing independently and fragmenting the mission.
That is not a software problem. It is a doctrine problem. And doctrine precedes software.
What the January Demonstration Actually Showed
The Swarm Forge demo was three drones striking three targets. Each drone found its target. Each drone executed its strike. The coordination required was minimal: don't hit the same target, don't collide.
At four drones — the Crucible minimum — the coordination problem is still tractable with relatively simple de-confliction logic. At forty drones, the problem changes qualitatively. At four hundred, it changes again.
The transition from "multiple coordinated assets" to "genuine swarm" is not a scaling problem. It is an architectural discontinuity. Systems built for four-drone de-confliction do not simply scale to four hundred. The coordination model breaks down. Emergent behaviors that were fine at small scale — agents optimizing locally, re-tasking independently, adapting their individual plans — become operationally destructive at swarm scale, where individual optimization without shared doctrine produces fragmented, uncounterable formations.
The Mongol army crossed this architectural discontinuity in the 13th century. At the scale of an Arban — ten soldiers — an experienced leader could command through direct observation and instruction. At the scale of a Zuun, a Minghan, a Tumen, direct instruction was impossible. The only solution was to encode intent deeply enough that each tier could operate as a coherent sub-formation without re-authorization from above.
The decimal hierarchy was that solution. Every Minghan commander received the Tumen's objective. Every Zuun commander received the Minghan's task. Every Arban received the Zuun's intent. When communications failed, the hierarchy continued. When a Minghan was lost to attrition, the adjacent Minghans adjusted. The formation was self-healing by design.
What the Crucible Should Be Testing
The current Swarm Forge Crucible specification tests autonomous capability. It should also test doctrinal coherence. Specifically:
Attrition resilience. Remove a node mid-mission. Does the formation adapt coherently, or do surviving agents optimize individually and fragment? A formation that loses one drone and continues executing its intended effect is qualitatively different from one that loses one drone and splits into three independent agents.
Intent fidelity under DDIL. Sever the uplink at mission start. The formation must complete its assigned mission set with no further human input. What is the degradation in mission coherence compared to a connected operation? How does this scale with formation size?
Re-synchronization after disconnection. When a disconnected sub-formation re-establishes contact with command, how cleanly does it re-integrate into the operational picture? Does it have a useful state model to offer, or does it simply await re-tasking?
Multi-tier coordination. Can the formation self-organize into nested sub-formations with distinct roles — some conducting ISR, some providing EW cover, some executing strike — coordinated under a single operator's intent without the operator micromanaging the tier structure?
These are the tests that distinguish a genuine autonomous formation from a collection of individually capable drones. They require doctrine-aware C2 architecture, not just capable edge agents.
KhanBMS and the Swarm Forge Requirement
KhanBMS is built for exactly the operational picture Swarm Forge is trying to validate. The decimal command hierarchy — Arban through Tumen — maps directly to the multi-tier coordination requirement. Intent-based tasking handles the DDIL requirement: operators command formations, not airframes, and every node carries embedded intent that survives disconnection. The hardware-agnostic modular asset layer addresses the multi-vendor heterogeneous requirement: new platforms onboard by manifest, not by software rebuild.
The Crucible will produce valuable data. It will demonstrate that autonomous multi-agent systems can perform impressive tactical behaviors. What it will not produce, without explicit doctrine testing, is evidence that those behaviors remain coherent at the scale and under the attrition conditions of actual peer conflict.
Swarm Forge is building the right capability. The next step is to define the right doctrine.
The Bigger Picture
Replicator promised thousands of autonomous systems by summer 2025. Hundreds materialized. The gap between prototype demonstrations and operationally deployable capability is not a hardware gap or even primarily a software gap.
It is a doctrine gap. The U.S. military does not yet have an operational doctrine for commanding large, autonomous, heterogeneous formations in DDIL environments. Every program — Swarm Forge, the Orchestrator Challenge, the DARPA constellation RFI — is circling the same absence.
The Mongol Empire deployed the largest coordinated ground forces of their era, across the largest operational theaters in history, with the most primitive communications infrastructure imaginable. They did it with a command doctrine so robust it survived the death of the Khan himself. The decimal hierarchy kept executing.
That doctrine scales. It is proven under the hardest possible conditions — not a test range in Florida, but the steppe of Central Asia, the mountain passes of Persia, the frozen rivers of Russia.
Swarm Forge is testing whether the hardware and software can keep up. The doctrine has been ready for eight hundred years.
KhanBMS is a modular battlefield management system for autonomous forces, architected on the Mongol decimal command hierarchy. Read the doctrine: The Mongolian Paradigm. Explore the system: khanbms.com.