This is the Overlooked Framework Behind Breakthrough Companies

There’s a kind of arbitrage in innovation that’s easy to miss because it doesn’t look like arbitrage at all. It lives in the gap between what physics allows and what institutions assume is possible. The reason it persists is that exploiting it requires developing genuine expertise in domains where most have neither the background nor the patience.

In 2019, I invested in SpaceX. What caught my attention wasn’t the headline features like reusable rockets or lower launch costs. It was watching them systematically demolish constraints that the entire aerospace industry treated as fundamental.

At the time, many investors, including me, were busy arguing about software multiples and growth loops. SpaceX was doing something much rarer: showing that a large fraction of what we call “impossible” is really just a coordination problem that no one had been sufficiently motivated to solve.

That investment was a turning point. For several years, I’d been a generalist investor, moving where the heat was. Consumer, enterprise, and a little hard tech when it came with enough narrative support. I wasn’t bad at it. But I wasn’t learning much either.

SpaceX clarified something I’d felt but hadn’t yet put into words. The biggest breakthroughs don’t come from optimizing within known constraints. They come from noticing that some constraints aren’t real. They’re habits, incentives, or historical accidents. And they require the technical grounding and nerve to see what happens when you ignore them.

The linearization fallacy

The Vannevar Bush–era model of innovation assumes a pipeline: basic research generates knowledge, applied research converts knowledge into prototypes, and development converts prototypes into products. This maps cleanly to how we organize universities, corporate R&D labs and government funding. It also happens to be backwards.

If you trace the causal history of transformative technologies — think transistors, lasers, the internet, GPS, mRNA vaccines — almost none emerged from linear knowledge transfer. They came from programs that collapsed the artificial boundary between “understanding nature” and “solving problems.”

This pattern appears often enough that it’s hard to dismiss as a coincidence. Donald Stokes formalized it in his analysis of scientific research by identifying Pasteur’s Quadrant: work that pursues fundamental understanding through the lens of immediate practical constraints.

Pasteur didn’t first understand microbiology and then apply it to fermentation. Investigating fermentation was how microbiology itself was built.

I first encountered Pasteur’s Quadrant in 2022, after starting Interface Fund and narrowing my thesis to biology, hardware, and infrastructure. We overlapped heavily with DARPA, were among the first VC checks into DARPA-funded companies, and became familiar with how they innovate and operate.

This isn’t a semantic distinction. It’s a claim about the structure of possibility space. There are regions of that space, high-value regions, that are only accessible when you optimize simultaneously for scientific novelty and practical application. If you decouple these objectives, even slightly, you end up searching a different region altogether.

DARPA’s hit rate makes sense once you see this. Packet switching, satellite navigation, RISC architectures, speech recognition, autonomous systems: these didn’t come from applying existing science to new problems. They came from organizations that refused to separate the question “what’s true?” from “what’s useful?”

SpaceX’s reusability breakthrough has the same structure. You don’t get there by first solving atmospheric reentry physics and then engineering an application. You get there by treating the coupled problem — entry physics, thermal materials, landing dynamics, propellant margins and economics— as a single integrated optimization. Progress in any subdomain opens new possibilities in the others.

Where breakthroughs actually come from

The key insight is simple. The objective function you optimize determines which regions of possibility space you can even perceive. Change the objective, and previously invisible solutions become obvious. This is why most breakthroughs seem obvious in retrospect. They were always there; you just needed the right lens to see them.

Operating in Pasteur’s Quadrant isn’t just intellectually different — it’s operationally different. It requires founders who can hold deep technical expertise while maintaining the flexibility to restructure problems from first principles. It requires organizations that can sustain exploration even as they scale, rather than collapsing into pure optimization.

And it requires a particular relationship with time: comfort working on problems where the compounding insight happens over years, not quarters, and where most of the value emerges in the final phases after long periods that look unproductive from the outside.

The companies that will matter over the next decade are those that systematically exploit this gap. They do it by operating in Pasteur’s Quadrant, coupling fundamental scientific questions with immediate practical constraints in ways that reveal new possibility space.

The framework exists, and we’re learning how to execute on it. Capital is beginning to follow. And there’s a specific kind of founder who is well positioned to benefit: someone with deep technical expertise, who has noticed that something everyone treats as a constraint is actually just a convention, and who is temperamentally incapable of letting that observation go.

There’s a kind of arbitrage in innovation that’s easy to miss because it doesn’t look like arbitrage at all. It lives in the gap between what physics allows and what institutions assume is possible. The reason it persists is that exploiting it requires developing genuine expertise in domains where most have neither the background nor the patience.

In 2019, I invested in SpaceX. What caught my attention wasn’t the headline features like reusable rockets or lower launch costs. It was watching them systematically demolish constraints that the entire aerospace industry treated as fundamental.