> ## Content Index > Fetch the complete content index at: https://modalpathethics.com/llms.txt > Use this file to discover other available public pages before exploring further. # The United States Army Must Abandon the Skeleton - URL: https://modalpathethics.com/the-united-states-army-must-abandon-the-skeleton/ - Published: 2026-10-26T11:00:00.000Z - Updated: 2026-10-26T11:00:00.000Z - Description: Modal Path Ethics brings you the warfighters of the future. - Author: Aidan Lawson - Tags: Security Instruments, Chastening of the Controller ## The Recruit Has Fallen. In September, the United States Army held the finals of its xTechHumanoid competition at Marine Corps Base Quantico. The objective was to find out what emerging humanoid robots might eventually contribute to military operations. There were 103 white-paper submissions, ten finalists, nine live demonstrations, and five winners. The Army was looking at power systems, sensors, autonomy, and complete humanoid platforms. **Real engineers** had built **real things**, carried them into the outdoors, and subjected them to the indignity of physical reality. One of the winners was Alex, a robot from the Florida Institute for Human and Machine Cognition. Alex *walked through grass*. Alex *followed voice commands*. Alex *negotiated gravel and woods*, with some of the movement directed by a human operating a remote control. Then, somebody gave Alex a shovel. This was a major escalation. The robot gripped the shovel, attempted to dig, and even tried to transfer dirt into a bucket. It did not successfully complete this task. According to the October 19 Defense News account, Alex then collapsed and **had to be transported away on a gurney.** We should be fair. Alex had demonstrated meaningful progress in a difficult technical field. The Army was testing immature technology precisely because testing exposes where the technology is immature. IHMC was selected as a winner. A robot falling during a prototype demonstration does not mean its engineers are fools. The shovel has still won this engagement. For several minutes I tried to think about the deeper engineering problem rather than the image of a robot being wheeled off the parade ground after losing an argument with landscaping equipment. Then I read the specifications. --- ## Twenty-Eight Miles an Hour and the Body of a Man. Robert Griffin of IHMC described the mismatch between the Army's ambitions and current robotics. As an example, he pointed to a machine that would run at 28 miles per hour, carry 50 pounds, and operate for 24 hours. His conclusion was refreshingly concise: > “It's not a thing that exists in the world.” Important qualification: the publicly released Army competition document does not list 28 miles per hour as its baseline speed. Appendix B specifies tactical movement up to six miles per hour, faster cover-to-cover movement, a 50-pound bipedal load, and a 24-hour operating cycle combining active and passive behavior. Those are different statements. The interview describes a much more ambitious combined target; the actual written competition baseline is already demanding enough without pretending it requires a full day of 28-mile-per-hour man-bodied sprinting. That settled, there is an obvious question. > **Why does the robot have to be a dude?** The competition defines a *militarized humanoid* in terms of replicating a human warfighter's form, capabilities, and functions. The document specifies an upright biped as the primary posture. It even allows a multiped alternate posture carrying a larger load. They have discovered that additional legs are useful. They have assigned those legs a secondary role because the robot must first satisfy the traditional military requirement of looking like someone who can be issued the pants. This is an understandable technological research category. It may also be an extraordinarily expensive way to invent a worse mule. Consider what is being requested. The robot must carry weight. Traverse rough ground. Enter buildings. Climb over obstacles. Use tools. Observe its surroundings. Survive unpleasant conditions. Communicate with soldiers. Keep working after something goes horribly wrong. The human body does all of this stuff pretty well. It also falls over when its support is disturbed, puts most of its important organs in one tall target-shaped arrangement, and requires the entire creature to reorganize itself every time it needs to reach a shelf. --- ## In Defense of Knees, Briefly. There is a serious case for the humanoid, and I want to make it before abandoning all discipline. The built world is full of concessions to human bodies. Door handles are placed where hands can reach them. Ladders expect feet and, again, hands. Vehicle seats, stairways, corridors, workbenches, tools, and maintenance access panels reflect a familiar range of human dimensions. A robot that can use existing equipment without reconstructing its environment is valuable. There are spaces where a wide tracked vehicle is useless, where four legs make a tight passage harder, and where the ability to operate an ordinary human tool matters more than raw carrying capacity. Bipedal research also teaches useful things about balance, perception, force control, and manipulation. The Army says humanoids are part of a broader robotics exploration, and that is a defensible program. The problem begins when compatibility becomes an aesthetic obligation. A machine can have hands without having hips. It can reach human-height controls without possessing a neck. It can fit right through a doorway while getting around on wheels, tracks, articulated limbs, or some kind of contraption that should *never* be shown to a horse. And the human shape brings its own engineering tax. Every step is a managed transfer of support. Heavy loads complicate balance. Falls threaten expensive mechanisms. A pair of human-looking hands can be useful; a full imitation human needs a great deal of machinery devoted to remaining an imitation human. No embodiment wins everywhere. A crawler can snag. Tracks struggle with some obstacles. Wheels are not magic. A centaur has just an awful time in a narrow staircase. An octopus robot may tie itself to a municipal handrail and have to be recovered by Parks and Recreation. **Fine.** Test those failures. But give the robot a body selected for what it must do, and let the body become increasingly offensive to the eye. I have prepared a series of replacement **Modal Path Ethical Warrior Bots**. The first few even contain actual engineering arguments. --- ## I. THE CENTAUR. ## We Have Discovered the Horse. Start conservatively. ![](https://storage.ghost.io/c/20/43/2043f11a-6ae3-404c-bb28-01fce8d9ac88/content/images/2026/10/84796-1.png) Keep two manipulator arms. Give them an adjustable working height and ordinary tool interfaces. Replace those legs with a long, low four-point body, wheeled legs, or a compact tracked chassis. Put batteries and compute inside the protected central mass. The head can remain, if necessary. Perhaps the Army will feel better if the robot has something to put a helmet on. It should be understood, however, that the computer has no metaphysical obligation to reside behind the eyes. A swiveling sensor mast is entirely capable of looking concerned. Now the machine can move equipment, brace itself while working, roll efficiently over decent ground, and distribute a payload across more than two supports. A low body can offer more stability for some jobs. The arms can still use a wrench or, only when absolutely necessary, attempt another shovel. I am not claiming this thing can sprint over a battlefield at 28 miles per hour for an entire day. Neither can the experimental humanoid. That was the point of the interview. I am, however, claiming that if the requirement is to carry fifty pounds over rough ground while accomplishing useful manipulation, *a mule with hands* deserves to enter the comparison before an artificial infantryman with exquisitely machined knees. This proposal is very reasonable. Unfortunately, I have been thinking about octopuses lately. --- ## II. THE OCTOPUS CORPORAL. ## The Controller Has Been Chastened. Remove the four legs. ![](https://storage.ghost.io/c/20/43/2043f11a-6ae3-404c-bb28-01fce8d9ac88/content/images/2026/10/84797.png) Install, oh, I don't know, perhaps thirty or forty slender, semi-rigid appendages beneath the waist. Give them contact sensing, local reflexes, variable stiffness where feasible, and different gripping ends. Some can hook, some can pinch, some can conform to rough surfaces. The robot has ordinary deployable wheels for moving across an Arby's parking lot, because I am unwilling to make it drag itself over asphalt like a wounded jellyfish. When it encounters a ladder, the appendages begin finding rungs. Some wrap around rails. Others support the body while fresh contacts are established. Near a ledge, several search for purchase while the rest keep it from falling. The body moves by redistributing tension and support. > Who coordinates forty limbs? The controller, naturally— **WRONG.** [**THE CONTROLLER IS CHASTENED.**](https://modalpathethics.com/tag/chastening-of-the-controller/) Modal Path Ethics has already covered this in [*Shape Is Already Code*](https://modalpathethics.com/transition-action-shape-is-already-code/) and [*The Signal Never Becomes Data*](https://modalpathethics.com/transition-action-the-signal-never-becomes-data/). Physical structure can participate in sensing and response. Local contact can trigger bounded action without waiting for the supreme processor to issue a decree about each microscopic deformation. The central controller says: *up*. Or: *cross this gap*. Or: *oh god, no, stop; that grip is unsafe*. The limbs negotiate contact. They recruit support. They release when other contacts can carry the load. The central controller remains responsible for goals, conflicts, and correction, while the body handles as much of the immediate encounter as it can. This does not abolish the problems of load transfer, energy, entanglement, or the fact that a cable excellent in tension may buckle miserably when asked to push. Those are real engineering problems. They are also problems worth having if, in exchange, a robot can climb a ladder by grabbing the ladder. I would now like to move to the next proposal before someone asks me to budget the actuators. --- ## III. THE ROTARY HOOK-SPINE. ## The Wheel Has Evolved. The octopus is still too polite to gravity. ![](https://storage.ghost.io/c/20/43/2043f11a-6ae3-404c-bb28-01fce8d9ac88/content/images/2026/10/84798.png) Add several large rotating structures shaped like hooked crescents or scalloped wheels. Each curve has a recess capable, in principle, of finding purchase against an edge. The machine approaches a ledge on ordinary wheels, catches one of its hooked lobes, rotates upward, and lets the tentacles stabilize the movement. The body is now partly a climbing mechanism and partly something recovered from the prop department of a science-fiction movie after a workplace investigation. This is less deranged than it looks, I swear. Robots such as RHex already show that rotating curved legs can exploit body mechanics and terrain contact without perfectly scripting every footfall. My cool hooks are a proposed extension of that family, not a demonstrated universal solution to climbing. They could miss. They could jam. They could rip away weak material. They could catch the wrong side of a fence and leave the robot hanging upside down while the mission continues on without it. Yet the central premise remains attractive: let geometry perform some of the decision-making about contact. Also the resulting machine looks sorta like a mechanical shrimp carrying a police officer's torso. --- ## IV. THE SUITCASE CRAB. ## Packing a Solution For Every Obstacle. At this point, the top half of the humanoid has become an inconvenience. ![](https://storage.ghost.io/c/20/43/2043f11a-6ae3-404c-bb28-01fce8d9ac88/content/images/2026/10/84799.png) Fold the arms into the chassis. Flatten the body. Give it eight articulated supports around the perimeter and two hinged structural panels that unfold as ramps, working surfaces, or short bridging sections. **The Suitcase Crab** arrives at a difficult obstacle, lowers itself, extends the necessary parts of its body, and becomes part of the route through the obstacle. Smaller equipment can pass over it. It then folds itself back together and scuttles away. This is a very expensive way to carry a small bridge, but a bridge cannot generally decide to leave afterward. The Suitcase Crab has that choice. One may object that the machine is now performing infrastructure rather than infantry. Correct. The infantry has been relieved of a task. The Army is free to continue dressing robots in pants if that makes the procurement department comfortable. I would like the obstacle to be crossed. Please note that the official Suitcase Crab has no face. It has achieved its first moral victory by ending the expectation that a machine attempting a maintenance task should look me in the eyes first. --- ## V. THE WALKING LADDER. ## Solve the Problem by Becoming the Problem. The Army wants a robot that climbs ladders. Very well. ![](https://storage.ghost.io/c/20/43/2043f11a-6ae3-404c-bb28-01fce8d9ac88/content/images/2026/10/84800.png) The next robot **is a ladder**. It moves on retractable feet. Its central spine telescopes upward into a rigid rung-bearing structure, and its hooked upper section fastens against a suitable support. Other units may climb it. When the job is done, it retracts and marches off to another wall. This is not an answer to how the machine itself climbs a ladder. The Walking Ladder has instead brought its own ladder and abolished the original question. A critic might reasonably suggest issuing the soldiers a normal portable ladder. But this critic does not understand the future. A normal ladder cannot nod at the commander. A normal ladder cannot move to the next appointment. A normal ladder cannot be promoted after its service in an exercise. A normal ladder has never had to endure a battery-life benchmark. The Walking Ladder might also one day crouch sideways and function as a bench. At this stage I am increasingly certain a government contract is coming. --- ## VI. THE TACTICAL TUMBLEWEED. ## Orientation Is a Civilian Luxury. Why should a robot remain upright? Why the hell should it have a front? Why, for that matter, should it ever have any idea which part of its body is currently the bottom? ![](https://storage.ghost.io/c/20/43/2043f11a-6ae3-404c-bb28-01fce8d9ac88/content/images/2026/10/84801.png) The Tactical Tumbleweed is a roughly spherical cage made from rotating curved limbs, hook-shaped struts, and retractable rolling elements. Cameras sit inside a stabilized central capsule. Whatever part of the cage happens to meet the earth becomes the temporary locomotion apparatus. It rolls. It catches. It rocks itself loose. Occasionally it spreads several struts and becomes a very nervous stool. I acknowledge that this would be fucking terrible at reaching a human-height control panel, that its gripping elements would love to collect wire, brush, and every loose scrap of fabric in Virginia, and that the engineering case is becoming very thin very fast. Still, this guy cannot fall over in the ordinary humanoid sense, because it has abolished the upright state. - Alex needed a gurney after falling. - The Tactical Tumbleweed could be transported by rolling it downhill. I am going to count that as a victory until a robotics professor contacts me. --- ## VII. THE COMBAT CHANDELIER. ## The Floor Has Lost Jurisdiction. The Tumbleweed's biggest weakness is being unnecessarily committed to touching the ground. ![](https://storage.ghost.io/c/20/43/2043f11a-6ae3-404c-bb28-01fce8d9ac88/content/images/2026/10/84802.png) The Combat Chandelier descends from overhead structures. A central capsule hangs beneath a forest of extensible limbs, each trying to find a beam, railing, pipe, or other secure attachment. Some shorten while others lengthen. The machine migrates through the ceiling structure by continuously exchanging grips. It has two little hands for operating equipment and a completely unjustified decorative brass ring around its middle. In a warehouse, industrial plant, or some kind of sufficiently convenient ancient ruin, the concept at least gestures toward an unusual form of traversal. In an open field it is just helpless. I have therefore added an umbrella. This umbrella does not resolve the lack of overhead anchoring. It does, however, make the machine look ready during rain. A responsible engineering review would probably terminate the design here. This article never had an engineering review. --- ## VIII. THE INFANTRYMAN IN TWELVE PIECES. ## Unit Cohesion Is Now Literal. Why insist that your robot remain assembled? ![](https://storage.ghost.io/c/20/43/2043f11a-6ae3-404c-bb28-01fce8d9ac88/content/images/2026/10/84803.png) Instead of a single large machine, give each support, sensor mast, manipulator, and battery module its own tiny independent vehicle. The pieces move separately, gather loosely around a task, and form a temporary working structure. When an obstacle changes, the arrangement changes. Need to lift something? Congregate underneath it. Need a view from above? Assemble a tower. Need to cross a narrow passage? Go through one at a time and reconvene on the other side. I am aware that modular robotics and multi-robot coordination already exist. I am also aware that forcing twelve self-propelled components to become one useful object may prove harder than building the original object. But, imagine the Army telling a soldier that his partner is late because its left shoulder is stuck in traffic. Imagine an officer conducting inspection and counting thirteen pieces when the manifest says twelve. Imagine a robot receiving a medal and nobody knowing which component to pin it on. This machine could salute by briefly arranging all twelve units into a large hand. No further questions at this time. --- ## IX. THE WALKING CHECKPOINT. ## It Is Now the Building. The program calls for checkpoint security. Fine. So build a checkpoint that walks. ![](https://storage.ghost.io/c/20/43/2043f11a-6ae3-404c-bb28-01fce8d9ac88/content/images/2026/10/84804-1.png) The Walking Checkpoint has four broad, arching legs. Its central body contains an inspection window, a folding desk, storage cabinets, lights, and a roof. Two ridiculous manipulators emerge from the front whenever somebody needs assistance. The entire establishment slowly moves to whichever road requires a checkpoint today. The human-shaped warfighter has been replaced by a mobile piece of municipal architecture. This may be a problem for transport, fuel economy, bridge ratings, electrical reliability, and the citizens who wake up to discover hostile government offices approaching their front lawn. Those are administrative difficulties. This is a tremendous machine. If another unit is damaged, the Walking Checkpoint can open a door and admit it for repairs. If Alex falls again, the building already has a place to store the gurney. The circle is closing. --- ## X. THE WAR GOOSE. ## No Remaining Justification. Fuck locomotion benchmarks. ![](https://storage.ghost.io/c/20/43/2043f11a-6ae3-404c-bb28-01fce8d9ac88/content/images/2026/10/84805-2.png) The War Goose has a long telescoping neck, an entirely unnecessary beak-shaped manipulator, four asymmetrical legs, two rotating hook-wheels, and a rear section full of an unknown quantity of very small retractable arms. The War Goose cannot run 28 miles per hour. It will not carry fifty pounds without serious modification. It definitely will not operate continuously for twenty-four hours, because I have included a mechanical procedure during which it sits in a puddle and refuses instructions. > What is its military purpose? We do not know. When ordered forward, it honks. When ordered to retreat, it honks. When a lieutenant attempts to explain the doctrine, it stretches its neck over his shoulder and pretends to read the map. This machine has no demonstrable operational advantage from being a goose. I can not defend its physical architecture, its power budget, its communications, or the manner in which it behaves around vehicles. It does, however, get up after encountering the shovel. I say promote it. --- ## The Army May Keep Its Robots. The real Army competition was a legitimate experiment. It exposed limits, rewarded promising teams, and produced useful information about batteries, sensing, manipulation, autonomy, and humanoid mobility. A big public falldown is not proof that the entire field has failed. The people who brought Alex to Quantico arguably did considerably more for robotics than I have done by inventing the War Goose. But the event also reveals a question that becomes harder to ignore once the human shape stops looking inevitable. If the task is to carry something, why are there only two supports? If the task is to climb, why is the robot allowed only the same grasping strategy as a human? If the task is to move through unknown geometry, why must one distant controller narrate every contact into a plan before the body can respond? If the task is to replace human exposure to dangerous work, why assume that the best replacement inherits the silhouette of the exposed human? Those questions do not necessarily require a tactical chandelier to be worth asking. There will be reasons to build humanoids. There will be reasons to build mules, crabs, crawling machines, distributed tool systems, and very complicated furniture. Some will fail spectacularly. Others will grip the shovel and discover the dangers of dirt. The only rule I propose is that the machine should have to earn its body. The human form is an excellent solution to being human. The Army is trying to build something else. Give it the chance it deserves to look fucking ridiculous. When the next one falls over, keep the gurney available. Or release the Tactical Tumbleweed. --- ### Sources - Daniel Terrill, [“US Army competition explores military uses for humanoid robots,”](https://www.defensenews.com/industry/techwatch/2026/10/19/us-army-competition-explores-military-uses-for-humanoid-robots/?ref=modalpathethics.com) *Defense News*, October 19, 2026\. The reporting on Alex's field demonstration and Robert Griffin's remarks. - U.S. Army, [xTechHumanoid Competition Announcement](https://xtech.army.mil/wp-content/uploads/2025/07/xTechHumanoid-Competition-RFI.pdf?ref=modalpathethics.com), Appendix B, especially pp. 11 and 15 for mobility, payload, posture, and power expectations. - U.S. Army FUZE, [“xTechHumanoid Winners Advance Military Exploration of Emerging Humanoid Capabilities,”](https://xtech.army.mil/announcement/xtechhumanoid-winners-advance-military-exploration-of-emerging-humanoid-capabilities/?ref=modalpathethics.com) September 2026, competition results and engineering aims. - [“Transition Action: Shape Is Already Code,”](https://modalpathethics.com/transition-action-shape-is-already-code/) June 8, 2026, and [“Transition Action: The Signal Never Becomes Data,”](https://modalpathethics.com/transition-action-the-signal-never-becomes-data/) July 20, 2026. - Saranli, Buehler, and Koditschek, [“RHex: A Simple and Highly Mobile Hexapod Robot,”](https://publications.ri.cmu.edu/rhex-a-simple-and-highly-mobile-hexapod-robot?ref=modalpathethics.com) 2001, for the existing curved rotating-leg precedent.