The Pentagon Just Offered $100M for Something the Mongol Empire Solved in 1206
How the DIU Orchestrator Prize Challenge validates a command doctrine eight centuries in the making.
The Pentagon Just Offered $100M for Something the Mongol Empire Solved in 1206
How the DIU Orchestrator Prize Challenge validates a command doctrine eight centuries in the making
In January 2026, the Defense Innovation Unit announced the Autonomous Vehicle Orchestrator Prize Challenge — up to $100 million for software that can translate a battlefield commander's spoken intent into coordinated action across fleets of heterogeneous drones, robots, and autonomous surface vessels.
Lt. Gen. Frank Donovan, head of the Defense Autonomous Warfare Group (DAWG), framed the requirement plainly: "We want orchestrator technologies that allow humans to work the way they already command — through plain language that expresses desired effects, constraints, timing, and priorities — not by clicking through menus or programming behaviors."
That is, almost word for word, a description of intent-based command. And intent-based command is the central innovation of the Mongol decimal system.
What DIU Is Actually Asking For
Strip away the AI branding and the prize structure, and the DIU requirement has three core attributes:
1. Intent translation, not instruction execution. The operator expresses desired effects — destroy that convoy, hold that corridor, suppress that ridge — not explicit step-by-step commands. The system decomposes that intent into per-agent tasking automatically.
2. Disconnected, edge operation. The submission guidelines are explicit: "The Orchestrator must be designed to function effectively under intermittent connectivity. The system may also eventually need to run in a disconnected, edge environment without access to the cloud." In other words, agents must carry their intent forward even when the link to command is severed.
3. Heterogeneous, vendor-agnostic formation control. Air, ground, and maritime assets from different manufacturers, coordinated under a single operator's intent. Not a single-vendor stack — a composable formation.
These are not new problems. They are the three oldest problems in the command of large, distributed forces. And they were solved — decisively, at scale, and in combat — by Genghis Khan's decimal command doctrine.
The Decimal Solution
The Mongol military was organized in a strict decimal hierarchy: units of 10 (Arban), 100 (Zuun), 1,000 (Minghan), and 10,000 (Tumen), with the Khan commanding at the apex. But the innovation was not the numbers — it was the operating principle embedded in the structure.
Every tier commander received intent, not orders. The Tumen commander knew the objective. The Minghan commander knew their role within it. The Zuun commander knew their task within that. The Arban — the tactical edge node — knew exactly what effect they were expected to produce, and was trusted to produce it without waiting for re-authorization.
When communications were severed — and on a steppe battlefield, they often were — each tier continued executing. The formation did not freeze. It fought through.
This is precisely what DIU is asking for when it requires the Orchestrator to "function effectively under intermittent connectivity" and forbids "false assumptions of control." The Mongols didn't assume control. They delegated intent, and let the hierarchy execute.
Where Today's BMS Architectures Fall Short
The systems competing for the DIU prize face a structural problem: most modern battle management software is built on hub-and-spoke assumptions inherited from Cold War C2 doctrine. A central node issues commands. Subordinate nodes execute. The link is assumed.
In an EW-saturated environment — the environment every future peer conflict will feature — that assumption is a single point of failure. Jam the uplink, blind the formation. Destroy the lead node, halt the swarm.
The Mongol decimal model has no such vulnerability by design. Every tier carries the full intent of the mission. Attrition at any level does not cascade upward into paralysis — the surviving elements continue executing because they understand why, not just what.
The key architectural requirement this points to is not better AI inference at the edge. It is better intent encoding before the edge. The operator's spoken command must be decomposed into intent packets that are durable, portable, and self-sufficient — capable of executing without further instruction, but coherent enough to re-synchronize when connectivity is restored.
The KhanBMS Architecture
KhanBMS is built natively on this doctrine. The decimal command tiers — Arban through Tumen through Khan — are not a branding exercise. They are the actual operational abstraction:
- The Khan tier is where human intent originates. The operator commands desired effects.
- The Tumen tier (10,000-agent scale) manages divisional autonomous formation.
- The Minghan tier (1,000-agent scale) handles battalion-level kill-web synchronization.
- The Zuun tier (100-agent scale) coordinates drone company-level effects.
- The Arban tier (10-agent scale) is the tactical edge node — the fireteam, the CCA pair — executing with full embedded intent, no uplink required.
Every node in the hierarchy carries lawful intent. The kill-web fights through denial, not around it.
This is the architecture that the DIU Orchestrator specification is pointing toward, whether the authors know it or not.
What the $100M Challenge Gets Right — and What It Misses
The prize challenge is correctly framed around outcomes, not prescriptions. It does not specify an architecture. It specifies performance: translate intent, operate disconnected, coordinate heterogeneously. That is exactly right.
What the specification underweights is the doctrine layer beneath the software. Voice-to-command translation is a tractable AI problem. Ensuring that the translated command carries durable intent through attrition and denial is a structural architecture problem — and one that cannot be solved purely at the inference layer.
The teams competing for $100 million will build impressive prototypes. The question is whether those prototypes encode the right abstraction: formations commanded by intent, not by instruction.
Eight centuries of operational proof suggest the right abstraction is decimal.
KhanBMS is a modular battlefield management system for autonomous forces, architected on the Mongol decimal command hierarchy. Learn more at khanbms.com or read the full doctrine paper: The Mongolian Paradigm.