In a building at the Army Research Laboratory just outside Washington, D.C., a group of reporters put on augmented reality headsets and looked down at a war.
Spread below them in a physics-based 3D simulation was a recreated chunk of Nevada and Southern California, extending out to the Pacific Ocean. Green laser-like lines streaked skyward from the terrain, representing connectivity links to satellites and high-altitude balloons carrying communications payloads. Purple and pink beams showed radar coverage. Light-pink beams traced electronic warfare jamming arcs across the desert. Tanks, robotic vehicles, artillery positions, fixed- and rotary-wing aircraft, and small surface vessels off San Clemente Island populated the scene.
Then a maritime battle broke out off the California coast. Fighter jets and autonomous wingmen launched from Nellis Air Force Base near Las Vegas. Land forces fired precision strike missiles at the National Training Center at Fort Irwin, California, softening enemy defences ahead of a joint forcible entry operation. Unmanned submarines tracked an enemy cruiser from below.
The tool making all of this visible is a product from defence technology company Anduril Industries called the Augmented Reality Sandtable – a physics-based simulation environment operated through standard AR headsets that allows multiple observers to watch, and interact with, the same live or reconstructed battle. It is a visualisation tool, not a command-and-control system, and the Army has been careful to draw that distinction. But as a demonstration of the sheer complexity of what Project Convergence Capstone 5 (PC-C5) attempted between March 10 and April 18, 2025, it proved more effective than any slide deck.
What Project Convergence Is, and What It Is Not
Project Convergence is the U.S. Army’s premier annual experimentation programme, organised by Army Futures Command and designed to test how well sensors, shooters, data networks, and decision-support software can be knitted together across services and allied nations. It is not a traditional military exercise in which units practise rehearsed tasks. It is, as Army officials repeatedly stress, an experiment: a structured attempt to discover what works, what breaks, and what needs to change before any of these technologies reach operational units.
PC-C5 was the largest and most ambitious iteration to date. Around 6,000 troops from the U.S. Army, Navy, Air Force, Space Force, and Marine Corps took part, alongside armed forces from the United Kingdom, Australia, Canada, New Zealand, France, and Japan (U.S. Army, April 4, 2025). Geographically, it stretched from the National Training Center at Fort Irwin in California’s Mojave Desert north to the Nellis Test and Training Range near Las Vegas, south to Naval Base Coronado and Marine Corps Base Camp Pendleton on the California coast, and out to San Clemente Island off the Pacific.
At its core, PC-C5 was driving toward a specific Pentagon goal: Combined Joint All-Domain Command and Control, known as CJADC2 (pronounced “C-jad-C-2”). The concept calls for near-seamless data links between sensors and weapon systems across every domain – land, sea, air, space, and cyberspace – spanning all U.S. military services and allied forces simultaneously, with artificial intelligence (AI) tools to help commanders make faster decisions using assets from any of those domains at once. That remains an ambitious and unfinished goal, but PC-C5 moved the programme further toward it than any previous iteration.
Two Phases, One Integrated Fight
PC-C5 was divided into two parts, which Army officials called Scenario A and Scenario B.
Scenario A, centred on the National Training Center, focused on corps-level and below operations. It assessed the Army’s Next-Generation Command and Control (NGC2) architecture and linked Nellis Air Force Base and San Clemente Island into a single operational picture. The scenario built around a maritime and special operations effort to open a coastal corridor, then inserted a joint forcible entry force against simulated enemy positions at Fort Irwin – connecting squad-level units at the lowest tactical echelon all the way back to corps headquarters.
Scenario B operated at corps level and above, using an Indo-Pacific setting. It focused more heavily on CJADC2 and demonstrated networked operations across Hawaii, Japan, the Philippines, French Polynesia, and Australia, with data shared in near-real-time with British, French, and Japanese partner forces (Defense One, May 5, 2025). This phase was primarily about proving that different nations and services, running different software on different hardware, could contribute to a single coherent operational picture and act on it together.
“We integrated it all into one coherent warfighting framework,” said Lt. Gen. David Hodne, director of Army Futures Command’s Futures and Concepts Center, at a press briefing following the experiment. “From the National Training Center… to Coronado, to Pendleton on the coast of California – we integrated all those capabilities in support of a joint task force executing actual missions.”
[Note on Hodne’s status: At the time of PC-C5, Hodne held the title of director of the Futures and Concepts Center at Army Futures Command. He was subsequently nominated in July 2025 to lead the newly created Army Transformation and Training Command – formed from the merger of Army Futures Command and Training and Doctrine Command – and was promoted to four-star general that October. He was dismissed from that position in April 2026 by Secretary of Defense Pete Hegseth (Stars and Stripes, April 7, 2026).]
The Data Problem at the Heart of Modern Command and Control
Perhaps the most candid and technically useful account of what the Army is actually trying to fix came from Maj. Gen. Patrick Ellis, director of Army Futures Command’s Command and Control Cross-Functional Team.
The Army does not have a data shortage, Ellis explained. It has a data organisation problem. “It’s not always in the right spot or the right format. That’s been our problem: we’ve had boxes that talk to boxes that talk to other boxes,” he said, capping the exchange rate at just under 25 words. “And if you want to interact with that data, you’ve got to pull it out of the box, manipulate it, translate it, use it, put it back in, re-translate it.”
The Army’s answer is what Ellis calls a clean-sheet approach: build a software layer that organises all incoming data into a single common format and pushes it into a common operating picture – essentially a shared digital map – where every data point is clickable and every piece of relevant sensor data associated with that point comes up automatically. If the dot on the map represents an enemy tank, any drone footage or other sensor data gathered and sorted by AI tools appears immediately when a user selects it. No manual translation. No data locked inside proprietary systems that cannot talk to each other.
Capstone 4, held in 2024, had a basic version of this common operating picture operating at small-unit level. Capstone 5 expanded it to battalions, in two flavours: one optimised for combat operations and a second for logistics. The next target, for Capstone 6, is to extend the same capability to divisions – organisations of close to 10,000 soldiers. “That’s a big step for us, to go from one small organisation of about 500 people to an entire division of almost 10,000,” Ellis noted. Project Convergence Capstone 6 is scheduled to begin in August 2026 (Defense One, May 5, 2025).
Lindsey Sheppard, director of the Advanced Command and Control Accelerator within the Pentagon’s Office of the Chief Digital and Artificial Intelligence Officer, said Capstone 5 also marked the first major experiment where her office demonstrated its edge data mesh technology – a distributed data architecture that pushes relevant information directly to users in the field rather than requiring them to pull it from a central server. In contested communications environments where bandwidth is limited or intermittent, that distinction matters considerably.
Mine Clearance Gets a Rethink
Among the most operationally grounded demonstrations in PC-C5 was a new approach to clearing minefields, driven directly by lessons from the war in Ukraine.
The previous standard for U.S. minefield-clearing called for clearing a path of roughly 100 metres (330 feet). Minefields encountered in Ukraine routinely exceed 900 metres (3,000 feet), well beyond the reach of legacy wire-charge systems such as the Mine Clearing Line Charge, which detonates a long explosive-laden cable dragged across a field. Stretching that system to match modern minefield depths is not practical.
The approach demonstrated at Capstone 5 breaks the problem into stages. Small drones first survey the field and map individual mine locations. That data feeds into the common operating picture so every element of the force has an accurate, current picture of where each mine sits. An autonomous ground vehicle then deploys small expendable drones – each on a one-way mission to destroy a single mine. A robotic bulldozer follows to verify and clear the path mechanically. No soldier needs to be anywhere near the minefield at any stage of the process.
The Army calls the small mine-hunting munition the Ground Obstacle Breaching Lane Neutralizer, or GOBLIN. Army Staff Sgt. Jeremy Ephriam, assigned to the 36th Engineer Brigade, was photographed preparing a GOBLIN for launch during the Fort Irwin phase of PC-C5 (U.S. Army, April 4, 2025).
Robots, Kamikaze Boats, and a Space Strike
Across Capstone 5’s various experiment “lanes,” several other technologies drew attention from briefers and observers.
A logistics lane tested an autonomous ground vehicle moving supplies between points without human drivers – a capability with obvious implications for reducing the number of soldiers exposed in resupply convoys. In the maritime fight, small unmanned submarines conducted subsurface surveillance of simulated enemy surface vessels, while Marine Corps units at Camp Pendleton fired anti-ship missiles from shore. Unmanned surface vessels conducted what the briefers described as kamikaze-style attacks on enemy ships – a tactic closely mirroring Ukrainian drone boat operations against Russian naval assets in the Black Sea.
In a human-machine teaming demonstration, U.S. and British Army soldiers coordinated combined assaults using a mix of robotic and manned vehicles, long-range fires, surface-to-air missiles, and electronic warfare equipment operating simultaneously. Special operations forces on San Clemente Island deployed high-altitude balloons carrying sensors and advanced communications payloads, while U.S. and UK commandos directed electronic warfare assets against simulated enemy positions to jam their communications.
Above all of this, a U.S. satellite was scripted to “disable” an enemy spacecraft, severing one of its communications nodes – a representation of space-domain operations that reflects the growing centrality of satellite connectivity to modern military command and control, and the corresponding priority that adversaries place on disrupting it.
Not all of the technologies in the simulation exist yet. The Future Long-Range Assault Aircraft – the Army’s planned next-generation heavy assault helicopter intended to replace the CH-47 Chinook – appeared in the 3D Sandtable rendering but was represented in the live desert exercise by actual Chinooks standing in for the unfielded aircraft. Brig. Gen. Zachary Miller, Joint Modernization Command commander, said the Army is using the experiment to understand how the new aircraft will be integrated into operations before it arrives, rather than waiting until after it is fielded to start working out the doctrinal questions.
What the Experiment Showed – and What It Did Not
Lt. Gen. Joel Vowell, Deputy Commanding General of U.S. Army Pacific, offered the most straightforward summary of what Capstone 5 demonstrated on the CJADC2 side: that multiple U.S. services and partner nations could take different programmes and systems and have them contribute to one common operating picture, with data accessible and usable by commanders and staff.
“That’s very helpful for commanders to be able to see data, or have a staff access the data, make sense of it and visualise it, understand it, comprehend it and use it,” he said.
But Vowell was equally clear that declaring CJADC2 fully operational remains a distant prospect. Technical barriers aside, policy obstacles also slow progress. Information-sharing rules between the U.S. and some allied nations – governing what data can move between unlike systems and across different security classification levels – limit what is practically achievable even when the technology works. “There are policy issues on information sharing with some of our allies and partners,” Vowell said.
Asked when the Pentagon might declare CJADC2 complete, he declined to name a date. “I don’t know if there’s a horizon timeline. I just know five years ago, we weren’t this close. We are a lot closer.”
The Army is also dealing with an institutional reality: it now tests around 250 different technologies under the Project Convergence umbrella, many of them related to command and control and networking, but not all. That breadth reflects genuine operational appetite for experimentation, but it also means each capstone is an immense coordination challenge.
Looking Ahead
Project Convergence Capstone 6 is scheduled for August 2026, and the Army has already set its ambition: expand the common operating picture from battalions to full divisions, nearly doubling the scale of the human-machine information architecture that Capstone 5 validated at the lower echelon. Whether the edge data mesh, the NGC2 software, and the AI-curated common operating picture hold up under that scale, and across the full range of allied systems that a coalition force would bring, is the next question the Army needs to answer.
The overarching goal – a genuinely seamless sensor-to-shooter network spanning all domains and all coalition partners – is not one that any single capstone will deliver. But the distance between where the Army started with Project Convergence and where it stands after Capstone 5 is real and measurable. The data architecture is cleaner, the common operating picture reaches further down the chain, the allied integration is more mature, and the autonomous systems in the logistics, engineering, and maritime lanes are further along than they were a year ago.
That incremental progress is the point. As Hodne put it at the Army Research Laboratory briefing, the experiment “allows us to mature capability. It allows soldiers to interact with [tools], give feedback, write better requirements.” In an era when the gap between commercial technology and military fielded capability risks widening rather than closing, the deliberate, iterative logic of Project Convergence is among the more credible answers the Army has to that problem.
