The Age of Adaptability
Why the next munitions revolution needs software, AI, and a new Iron Triangle.
The United States has exquisite weapons and exquisitely shallow stockpiles. That brutal fact, proved beyond a doubt by Ukraine and every Taiwan war game, should be our wake-up call. Our missile inventories weren’t designed for drawn-out conflict. We built exotic supercars, not mass-produced fleets. And when the enemy is coming fast, it doesn’t matter how good your last missile was. It matters whether you have any left to launch.
The problem isn’t just that we need more munitions. The problem is we’ve designed a system that can’t build them fast enough. Weapons like SM-6, JASSM, or Patriot were engineered in an era of strategic optimism, when war was short and production was someone else’s problem. They’re handcrafted, over-specified, and reliant on scattered, brittle supply chains. These are America’s gold-plated arrows: spectacular to behold, agonizingly slow to make.
Too many defense timelines talk about production and fielding in 2030–2035. That is a recipe for irrelevance. If you’re not solving the problem inside 12 to 18 months, go home. Kelly Johnson built the U-2 spy plane in nine months. That was nearly 70 years ago. The question for us today is simple: why can’t we move faster now, with AI, digital twins, and software-defined design?
Or consider another lesson from history. During the Second World War, Germany invested heavily in a small number of exquisite platforms—tanks, rockets, and fighters with better science and craftsmanship than anyone else. But exquisite didn’t equal decisive. In the end, they were outbuilt, outscaled, and overwhelmed. Their craftsmen were producing intricate machines while Detroit was cranking out functional machines in industrial quantities. America, the country that revolutionized mass production for war, cannot afford to repeat our old enemy’s mistake. But today we’re drifting into the same storm, with a force built around a small set of exquisite systems that may not arrive in time and are not adaptable to change.
So yes, we need more of what works today. But more than that, we need the capacity to build what we don’t yet know we’ll need tomorrow. Not just more output. Scalable output. Flexible output. Modular munitions that can be rapidly tailored, mass produced, and adapted as the threat evolves. The winner in tomorrow’s war won’t be the side with the best missile. It’ll be the one with the best system to build the next missile.
That’s the shift. It’s not just about output. It’s about adaptability. We need production lines and development pipelines that can flex. Factories that can shift from missile to drone to decoy. Platforms that can be updated via software. Architectures that let us plug in new payloads, sensors, and behaviors as fast as the threat evolves.
For decades, the Pentagon has lived by an Iron Triangle of cost, schedule, and performance. Problem is, this framework has hardwired the system to optimize for perfection at the expense of speed and scale. Programs chase exquisite performance against every conceivable threat scenario, costs balloon, and schedules stretch into decades. By the time a weapon finally arrives, the fight has already changed.
What we need now is a new Iron Triangle designed for an era of deep uncertainty. In this model, time to relevance is sacred. The first question isn’t whether a weapon can meet every requirement, but whether it can be in the field in 12 to 18 months. Second, capacity must be treated as strategic. It’s not enough to design one exquisite prototype—we need to be able to produce thousands, and to surge output as the fight evolves. Finally, performance must become the variable, not the constraint. Instead of demanding perfection up front, we must accept “good enough to fight now,” knowing that software updates, modular payloads, and AI-enabled iteration can rapidly improve capabilities in the field. This new Iron Triangle—speed, scale, and adaptability—breaks with the mindset of the 1990s. It recognizes that in modern warfare, relevance is perishable, mass matters, and perfection delivered too late is indistinguishable from failure.
We don’t just need cheaper weapons. We need weapons with metabolism—systems that can evolve, scale, and reconfigure at the speed of the fight. The future won’t reward exquisite systems delivered late. It will reward repeatable advantage: missiles that can be launched by the thousand, updated overnight, manufactured on demand, and launched again.
Look at the battlefield. A $50,000 Ukrainian drone took out a $2 million Russian tank. And hundreds more were ready to go. During the Red Sea attacks, the U.S. Navy burned through $2 million interceptors to shoot down drones that cost $2,000. That was a tactical asymmetry; scale it up, and it becomes a strategic industrial imbalance. Our adversaries are iterating. We’re inventorying.
Ultimately, the missile problem is a microcosm of the Pentagon’s industrial base dilemma. Whether it’s shipyards, sensors, or semiconductors, we’re optimized for predictability in an era that punishes it. This is where software and AI become the backbone of deterrence. Simulation can replace months of physical testing. AI can identify alternative suppliers before bottlenecks hit. Software-defined payloads can adapt in flight. Modular platforms can be recombined like Legos based on threat environment, not fixed configuration. AI agents can even recommend optimal loadouts for terrain, target, and enemy behavior, acting as codebases in motion, learning and improving every time they’re used.
Today’s acquisition system was built for certainty, but we now find ourselves in an era of deep uncertainty. The militaries that thrive will be the ones that adapt first—and adapt fastest. Some weapons are beginning to reflect this insight. The Navy’s Black Arrow cruise missile is affordable, modular, and flexible. Meanwhile, Anduril’s Barracuda family of modular, software-defined cruise missiles have a unit cost under $200,000 and are explicitly designed for hyper-scale production.
But we need policy, budget, and culture to catch up before we can realize these opportunities for adaptability at scale. Recent developments are encouraging. The bipartisan SPEED Act would blow up the Pentagon’s glacial requirements process to give program executives the authority—and responsibility—to deliver relevant capability, fast. The $14 billion in low-cost weapons funding in the reconciliation bill is a signal that Congress understands the stakes. And the FY26 budget shift away from aging airframes toward drones, expendables, and AI-enabled munitions is the start of a necessary rebalance.
But implementation is everything. It’s not enough to make new laws. We need new habits. Program managers need to work with operators to determine how fast is “fast enough” and how good is “good enough.” This is a major and disruptive change in the traditional program management model that many acquisition officials will be unable to embrace. They will need to make way for a new generation of program managers who are comfortable orchestrating a dynamic intelligence-development-operations cycle.
In WWII, we outbuilt and outmaneuvered the Axis, turning commercial factories into arsenals. That won’t work in the 21st century against the world’s manufacturing powerhouse. The U.S. military will have to out-adapt adversaries just as Ukraine has done against Russia. That means our factories and weapons will run on software, not raw manpower. And their edge won’t be their size but their speed of learning.
Our enemies are building for war. We’re still building for requirements. That has to change, because the next war won’t be won by the best missile. It will be won by the system that can build the right missile, in the right place, at the right time. Again. And again. And again.
The age of the exquisite alone is over. The age of adaptability has begun.
Greg Little is a senior counselor at Palantir.
Bryan Clark is a senior fellow at the Hudson Institute.
 | A guest post by
|