Thesis

Every Hard Problem Left Is a Systems Problem

AI is compressing the digital layer into a commodity. The engineers who hold complex physical systems in their heads are about to become the most important people in the room.

Greg DogumApril 20268 min read

Here's what nobody in tech is talking about: the software revolution is eating itself. AI agents are writing code. LLMs are building applications from natural language. The entire digital layer — the thing that made Silicon Valley the center of the universe for three decades — is being compressed into something approaching a commodity. The marginal cost of producing software is collapsing toward zero.

So what happens when code is no longer the bottleneck? The world shifts from bits to atoms. And the hardest problems left — the ones that actually matter, the ones that determine whether civilizations advance or stall — aren't software problems. They're systems problems.

Autonomous vehicles. Clean energy grids. Space infrastructure. Defense platforms. Medical devices. Every one of these requires someone who can hold the entire picture: hardware talking to software talking to physics talking to humans talking to constraints. Not one layer. All of them, simultaneously.

That's a systems engineer. And right now, they're the most undervalued people in technology.

The ability to decompose a complex system into layers of causality — to see not just what exists, but what's missing — is the one skill that gets more valuable as AI gets more powerful.
Skill Demand Trajectory — 2025 → 2030
Falling
Saturated
Plateau
Rising
Exponential
Prompt Eng.
Vibe Coding
Full-Stack
Hardware
Systems Thinking

The Tooling Crisis Nobody Talks About

Here's what makes this worse: the tools systems engineers use today were designed before the smartphone existed. We're talking about software that costs thousands per seat, takes months to learn, and fights you at every step. Enterprise platforms built for the procurement era, not the engineering era.

Every other engineering discipline got its modern moment. Developers got GitHub, Figma, VS Code — tools that feel like extensions of how they think. Designers got a revolution in collaborative, intuitive software. Systems engineers got nothing. They're still drowning in document-centric processes, static diagrams that go stale the moment they're created, and tools that treat the model as an afterthought to the paperwork.

The global model-based systems engineering market is approaching $7 billion and growing at 15%+ year over year. The tooling serving it hasn't had a meaningful innovation in over a decade. The incumbents charge $600–$2,500 per user per year for software that looks and feels like it was designed by a committee in 2009 — because it was.

The Widening Gap
System complexity is accelerating. Tooling hasn't moved.
MBSE tooling innovation vs. system complexity, 2010–2026
The Gap20102014201820222026
System complexity
MBSE tooling innovation
Think about it: a $7 billion market growing at 15% per year, served by tooling that hasn't had a meaningful innovation in over a decade. In software, that mismatch would be called an emergency. In systems engineering, it's called Tuesday.

Why This Matters Right Now

This isn't an abstract observation about the distant future. Three things are converging right now that make this the inflection point:

01
AI as Thinking Partner
AI is ready to reason about structure — not just generate text
The technology finally exists to have AI reason about system decomposition, dependencies, constraints, and tradeoffs — not just summarize documents or generate boilerplate. But nobody is building this for the people who need it most. The entire AI tooling ecosystem is aimed at software developers and knowledge workers. Systems engineers are an afterthought.
02
The Atoms Economy
Hardware investment is surging across every sector simultaneously
Defense spending is accelerating globally. Space is commercializing at scale. Energy infrastructure is being rebuilt from scratch. Hardware startups are raising at record pace. Every one of these domains needs rigorous systems thinking, and the demand for engineers who can do it is outpacing supply. The atoms economy isn't coming — it's already here.
03
The Knowledge Cliff
Experienced systems engineers are retiring faster than they're being replaced
The tacit knowledge — the hard-won intuition about how complex systems fail, where interfaces break down, what the requirements document doesn't say — is walking out the door. If we don't build tools that capture and amplify that thinking, we lose a generation of institutional knowledge. This isn't a talent problem. It's a tooling problem.

The Skill AI Can't Replace

Here's the counterintuitive truth: AI makes systems thinking more valuable, not less. AI can generate code, write documents, even suggest architectures. But it cannot hold the intent of a system. It cannot feel the tension between competing constraints and make the judgment call. It cannot look at a complex decomposition and sense that something is missing.

That intuition — the ability to see the shape of a system before it exists, to reason about what should be there — is irreducibly human. It is what separates a list of components from a coherent design. And it's exactly what systems engineers do every day, in domains where getting it wrong doesn't mean a 404 page. It means a failed launch, a grounded fleet, a safety incident.

Software engineers had their decade. The next one belongs to the people who build things you can't ctrl-Z.

The people who can think in systems are about to become the most important people in the room. The question is whether we'll give them tools worthy of the work.

Cairn is the AI engineering workbench for systems that matter.

Sign up and start building for free.