June 9, 2026
Ludwig von Bertalanffy was born in 1901 to an aristocratic Austrian family. He emerged in his adult years as a contrarian and rigorous biologist amidst an intellectual movement in Vienna, the Vienna Circle. It seems he was inspired by their rigor, but never aligned with the reductionism their program embodied. Many of his peers looked at physical and biological systems as closed systems: systems that could be contained, measured, and predicted independently, as reductively small parts. Ludwig rejected this premise.
Having rejected the reductionist premise, he went looking outward toward other disciplines and fields. He observed that principles like wholeness, organization, open systems, equifinality, and hierarchy kept reappearing in psychology, in economics, in sociology, in systems that had nothing physically in common. He called these formal patterns and recurrences "isomorphisms." He is quoted as saying, "Scientists, operating in the various disciplines, are encapsulated in their private universe, and it is difficult to get word from one cocoon to the other." His characteristic stubbornness led him to general systems theory, "wholes" made of interacting parts exhibit lawful properties that belong to their organization rather than their material, so the same formal principles recur across unrelated systems and can be studied as an independent, cross-disciplinary, unified science.
Those lawful properties frankly never shook out quite the way he had hoped, and the important bits of his theory fragmented into more specialized interdisciplinary fields. What survived his desire for a unified science was a sensibility for open systems and emergence. We can reason about open systems as patterns of inputs, throughputs, outputs, and diagnostic feedback loops— systems that sit adjacent to and nested within other systems, all seeking some sense of "steadiness."
I believe those sensibilities to be grounding and important mental models as we all navigate the breakneck pace of change we're experiencing with artificial intelligence. We are very quick to assign irrelevance, take absolutist positions, and conflate outcomes because of the uncertainty of the technology. A common deduction I've heard since Opus 4.5 was introduced is that we no longer need engineers (amongst other roles), that the democratization of code has made the role irrelevant, redundant, or too costly. As an aside, I think if AI has led you to suddenly believe that engineering is a cost center, then you never understood the purpose and value of the role in the first place.
As the number of possible futures expands rapidly, it is odd to me that we have been more likely to feel resolute in what is no longer true than to ground ourselves in what seems to continue to be true.
I think the core problem is that over the last several decades we framed the value of engineering as the systems and processes that surround the composability of code. It's the same reductionist move Ludwig spent his life arguing against, mistaking the organized whole for the material it happens to be made of. The irony is that the ergonomic vehicle for how we've scaled that premise is itself a series of abstractions. You don't write code in binary or assembly. We use compiled languages. We build super-systems on top of those machine-level systems. And in my opinion, none of those abstractions ever changed the purpose of the system.
This is a nonlinear leap from the previous abstractions we've navigated. Compilers were faithful, deterministic translators; a compiled language still required engineers to be precise and logical about the intent of what they wanted. Artificial intelligence is showing the capacity to participate directly in specification and composability, and to leverage those older systems of translation on its own.
Perhaps high-fidelity specification gets cheap too, the way translation already did. When the artifact is nearly free to produce, the scarce work just moves up a level: deciding what is worth building, judging whether the result is actually right, choosing which of the many valid artifacts to ship, and standing behind it when it breaks.
Composability and translation were never the value of engineering. As it turns out, neither is specification. They were the techniques and artifacts for delivering value in steerable ways. So you have to ask what the "system" of engineering is actually for. What are its inputs? Its throughput? Its outputs? Its diagnostics? Abstract away the composability of code, then abstract away the specification, and what's left is the thing we were paying for the entire time.
We pay engineers for their specialized judgment. Great engineers are excellent judges of systems, abstractions, complexity, and information, and until we know far more than we do right now, that judgment is where the value lives.
Cheaper code doesn't shrink the work; it grows it. Jevons paradox states when you make a resource cheaper to consume, total consumption tends to go up, not down. Every abstraction we've already lived through bears this out. Assembly to C, C to Python, frameworks, the cloud, each one lowered the cost of producing software, and each one expanded the population of people building it, because each one widened the set of things worth building. We have not yet come out the far side of an abstraction with fewer engineers. So as the cost of code falls, the corpus of code grows, and the complexity of that corpus grows with it. As the artifact gets nearly free, impeccable judgment about which artifact becomes asymmetrically valuable. When anyone can generate the code, the scarce and differentiating thing is the person who can be be held to account for it’s outcome and performance.
The organizations that will thrive over the next ten years are the ones that truly understand the systems at play and how best to leverage them. I don't think that will look like borrowing the org chart from the X post with the most likes. I think it will take honest evaluations of how these systems actually work, and adapting them to what we know right now.
The next several years will unfurl a massive amount of possible futures. The few things that I believe will remain true are that a system is more than its parts, and that someone still has to exercise judgment over the whole. I don't expect that to abstract away anytime soon.