The global defence landscape is a theatre of high-stakes competition, but the most significant battles aren’t always fought with jets and ships. One of the most pressing challenges for Western military superiority is happening on servers, deep within miles of ageing computer code. This is the quiet, grinding war against technical debt, and it’s a fight that could determine the future of national security. For decades, mission-critical systems have been built layer upon layer, creating a digital Jenga tower that is both indispensable and terrifyingly fragile.
What is this technical debt? Imagine you’re building a shed. You’re in a hurry, so you use a few crooked nails and some wood that isn’t quite the right size. It works for now. Next year, you add a shelf, but you have to work around the crooked nails. A few years later, you want to add a window, but the frame is now so warped that it’s a massive, expensive job. You’ve accumulated technical debt: the implied cost of rework caused by choosing an easy (limited) solution now instead of using a better approach that would take longer. In defence, this isn’t about sheds; it’s about fighter jets, missile guidance systems, and secure communication networks. The “interest payments” on this debt are slower development, cybersecurity vulnerabilities, and an inability to integrate modern technologies like AI. This is the core challenge of defense modernization.
The Critical Role of Software Verification
You can’t just run these old systems through a simple translator app and hope for the best. When lives are on the line, every single line of code matters. This is where the discipline of software verification comes into play. It’s the painstaking process of proving that the software does what it’s supposed to do—and nothing else. In conventional software, a bug might mean your app crashes. In a military context, a bug could have catastrophic consequences.
Effective software verification ensures that when you rewrite or update a legacy system, you aren’t accidentally introducing new, hidden flaws. It’s the digital equivalent of having an army of obsessive inspectors checking every rivet and weld in a new submarine. Without it, any attempt at modernisation is just gambling with national security. The goal is to make the new system mathematically provably correct, ensuring that the secure migration from old to new is flawless.
Code Metal: An AI Scribe for a Decades-Old Problem
So, the problem is clear: rewriting millions of lines of vital, yet ancient, code is slow, expensive, and fraught with risk. Enter Code Metal. As reported by WIRED, this Boston-based startup has just pulled in a hefty $125 million in Series B funding at a $1.25 billion valuation, with backers including Salesforce Ventures and Accel. Their mission? To use AI to automate this exact process.
Code Metal is not just building another “AI that codes.” They are developing sophisticated tools designed to translate code from older languages into modern, more secure ones like Rust. But the translation is only half the story. The real innovation lies in their approach to verification. As co-founder Yan-David Erlich puts it, “But in the course of translation, you might be inserting bugs—which is catastrophically problematic.”
To solve this, Code Metal’s AI doesn’t just rewrite the code; it simultaneously generates a proprietary “test harness.” This acts as a virtual simulation environment that rigorously tests the new code to ensure its behaviour is identical to the original. According to CEO Peter Morales, the system is designed so “there’s no way to generate an error.” If a perfect, bug-for-bug compatible translation isn’t possible, it simply won’t produce a result. This focus on verifiable correctness is precisely what major players like L3Harris, RTX, and the US Air Force are paying for.
A New Model for an Old Industry
Perhaps the most telling detail about Code Metal’s strategy is its business model. Instead of the typical per-user software licence, they are pricing their service based on the development time saved. This is a genuinely fresh approach. It aligns their incentives directly with their customers’ primary pain point: the immense time and cost of technical debt reduction. They aren’t selling a tool; they are selling a solution to a multi-billion dollar headache.
– They bill clients based on engineering-years saved.
– This directly quantifies the value of technical debt reduction.
– It shifts the focus from a software product to a strategic outcome.
This model fundamentally changes the conversation. It moves away from “How much does your software cost?” to “How much faster can we modernise?” For an industry notorious for slow procurement cycles and cost overruns, this is a powerful proposition. It shows a deep understanding of the business of defence, not just the technology.
The Future is a Rewrite
The implications here are significant. We are moving beyond using AI for simple assistance and into an era where it performs highly complex, high-stakes logical transformations. Andrej Karpathy, a leading AI researcher and angel investor in Code Metal, noted, “It feels likely that we’ll end up rewriting large fractions of all software ever written many times over.” We are at the very beginning of this trend.
Achieving true defense modernization isn’t just about buying new hardware. It requires a fundamental overhaul of the software that underpins everything. Challenges remain, of course. Integrating tools like Code Metal’s requires a cultural shift within vast defence organisations and government agencies. Trust in an AI-driven verification process must be earned and proven, especially within a risk-averse culture.
However, the momentum is undeniable. The need for speed, security, and adaptability means the old way of manually patching and praying is no longer viable. As AI tools for code translation and verification mature, they will become indispensable for maintaining a technological edge. The ability to perform a secure migration of legacy systems won’t just be an IT project; it will be a core component of defence strategy.
So, as we watch companies like Code Metal scale, the real question isn’t whether AI can rewrite code. It’s how quickly defence leaders can adapt their mindset and processes to embrace this new reality. What other industries do you think are on the verge of a similar AI-driven “great rewrite”?
