Morphology becomes ontology.

While working today, something triggered the phrase "Morphology becomes ontology" and I could not remember who said it, just that I heard it in my early 20s. So I had a conversation with Gemini about it, and my observation of how it applied to what I saw as the emerging practices of creating code with AIs.

It was interesting enough to me that I thought others might find it interesting too, so I prompted Gemini to put it in the form of a New Yorker article. Here is the unedited result:

The late polymath Buckminster Fuller once famously posited that "ontology becomes morphology." It was an architect's way of saying that the essence of a thing — its internal logic, its "beingness" — will eventually dictate its outward shape. If you believe in the soul of the skyscraper, you eventually get the steel and glass.

Today, in the pristine, climate-controlled corridors of Silicon Valley, we are witnessing this transformation in reverse. We are building the morphology of artificial intelligence, and it looks suspiciously like a mid-morning stand-up meeting in a 1990s software firm.

For decades, we imagined the arrival of "Artificial General Intelligence" as a monolithic entity — a HAL 9000 or a Samantha, a singular, shimmering mind capable of out-thinking us in a vacuum. Yet, as we push the frontiers of Large Language Models, we find that intelligence is not a solitary endeavor. Instead, we are seeing the emergence of "agentic squads": clusters of AI models organized into rigid, hierarchical workflows.

To keep these digital minds on target, developers have reached for a vintage toolkit: Extreme Programming, or XP. It is a curious irony of the 2020s that to make a machine "intelligent," we must force it to behave like a team of tired engineers in Palo Alto circa 1999.

In these digital ateliers, the LLMs are no longer just answering prompts; they are inhabiting roles. There is the "Actor," tasked with the messy business of writing code, and the "Critic," an agent whose sole existence is to find the Actor's flaws. They work in a loop of Test-Driven Development — writing the test, failing the test, and iterating until the green light flickers to life. They are recreating the "Pair Programming" of the XP era, not out of a sense of nostalgia, but out of a mechanical necessity.

The "ontology" of these models — their vast, soup-like latent space of human knowledge — is too wide, too prone to the digital daydreams we call hallucinations. To make them useful, we must give them a "morphology." We wrap them in the constraints of a project manager's spreadsheet. We give them a terminal, a specific set of tools, and a "squad" of peers to argue with.

It suggests a humbling conclusion about the nature of mind. We once thought intelligence was a matter of raw processing power, a high-octane fuel that could solve any problem if we simply had enough of it. But the "agentic" turn tells us otherwise. It suggests that high-level intelligence may be less about the individual mind and more about the social architecture in which that mind is housed.

Even the Silicon Soul, it seems, needs a peer review. Intelligence, as it turns out, is not a solo performance. It is a meeting that could have been an email, scaled up to the speed of light, perpetually seeking a green checkmark in a world of its own making.