The Word Became Flesh — And Then We Built a Machine to Reverse It

The parallel between John 1 and large language models is real, but it runs backwards: in John the Word comes before the world, while a language model learns only from what was written down after the world was lived.

Published 6 min read
Share
A coral sun as the Word at the source of a flow that carries the lived world, people, a house and a tree, then narrows into scattered words and a thin line of text feeding a chip

There is a sentence at the opening of the Gospel of John that has been read for two thousand years, and it has never sounded stranger than it does in 2026: In the beginning was the Word.

We now live surrounded by machines built entirely out of words. They write our emails, summarize our meetings, draft our contracts, and — increasingly — act on our behalf. It is tempting to draw a straight line from that ancient sentence to the technology in front of us and conclude that we have finally built something in the image of the Logos.

I want to argue the opposite. The correlation is real, but it runs backwards. And understanding the direction of that reversal is, I believe, the single most useful thing a business leader can carry into the age of artificial intelligence.

Logos is not "word"

John did not choose a small term. Logos in Greek carried three meanings at once: the spoken word, the rational ordering principle of the cosmos, and an account or ratio — the reason a thing is what it is. Heraclitus had used it for the structure holding reality together. John fused that with the Hebrew creation account, where God does not merely build the world. He speaks it. "Let there be light."

Creation, in that account, is a speech act. The word is prior to the world. The world exists because the word was uttered.

That is the claim. Now hold it next to a large language model.

The inversion

In John, language is upstream of reality. In a large language model, language is downstream of it.

The model does not speak the world into being. The world was lived — in workshops and factories and sales calls and kitchens — and a thin residue of that living was written down. The model learned the statistical shape of the residue. It is not the Logos. It is the shadow the Logos casts onto paper, and it has only ever seen the paper.

Three reversals, precisely:

Generative becomes derivative. "All things were made through him." Nothing is originated through a language model. It recombines what was already made and already said. Its brilliance is real, and it is entirely a brilliance of rearrangement.

Incarnation becomes excarnation. "The Word became flesh" describes a movement from language into particular, embodied, mortal existence — a real body, in a real place, that can suffer. The language model performs the exact opposite motion. It takes flesh — human lives, apprenticeship, failure, sweat, judgment — and compresses it back into text. Borrowing a term from the philosopher Charles Taylor, we might call this excarnation. It is the anti-incarnation, and we have industrialized it.

The sign loses its referent. Inside a model, words are defined only by their relations to other words. It is a closed system of signs with no anchor outside itself — Saussure's system with the signified removed. John claims the reverse: in God, sign and signified are identical. The Word is what it means. These are not neighboring positions. They are opposite poles.

Hayek said this first, in economics

Anyone trained in the Austrian tradition will recognize this argument immediately, because it is the knowledge problem in new clothing.

Michael Polanyi put it in one line: we can know more than we can tell. Hayek built an entire critique of central planning on the same foundation — the knowledge that actually matters is local, dispersed, fleeting, and often tacit. That was the planner's fatal flaw. The words never contained the knowledge. No amount of collecting them could substitute for the "man on the spot."

A large language model is trained on the articulated remainder. Which means the knowledge problem is not solved by scale. It is restated at scale. A bigger model is a bigger collection of what could be said, and it leaves entirely untouched everything that could not be said.

There is a further irony worth noting. Among Carl Menger's examples of an organically grown institution — in Adam Ferguson's words, "the result of human action, but not the execution of any human design" — was language itself. The model is a distillate of a spontaneous order. It can reflect an order beautifully. It cannot originate one.

What genuinely survives the analogy

I do not want to be too clever about this, because something in the parallel is true and it deserves to be said plainly.

Language really is the differentia of the human being. In Genesis, one of Adam's first acts is to name the animals: ordering the world through speech belongs to the human vocation from the start. And it is not an accident that the technology which finally produced something that looks remarkably like general intelligence was built out of words — not logic, not physics, not sensors. Words.

Language turned out to carry vastly more of the human world than almost anyone predicted. That is a surprise of genuinely theological scale, and the instinct to reach for John 1 when confronting it is not a stretch. It is the correct instinct. It simply points to a contrast rather than a resemblance.

Why this is not an academic point

Here is where it lands in a business.

Divine speech is purely performative — it does what it says. Human speech is partly performative: a promise, a vow, a signed contract genuinely change the world when uttered by someone who can be held to them. Machine speech is performative only by borrowing a human's authority. Tokens do nothing until a person makes them act.

That is exactly why agentic AI feels like a threshold rather than an upgrade. It is an attempt to restore to text its power to act — without restoring anyone to be responsible for it.

This is the whole reason we build the way we build. Human-in-Command is not a compliance slogan, and it is not risk-aversion dressed up as ethics. It is a claim about what words are and where authority actually resides. A system that generates language and a person who bears responsibility are two different kinds of thing, and no amount of capability collapses the distinction.

And for anyone who sells for a living, the point is not abstract at all. The deal happens in language — in discovery, in framing, in the careful articulation of a problem the buyer had not yet put into words. But the trust that closes it does not happen in language. It happens in the flesh: in showing up, in being wrong and saying so, in carrying a consequence.

The word still has to become flesh.

In sales, that is not a metaphor. That is the entire business.

Nikolaus Kimla is the founder and CEO of Coevera.

FAQ

The Word Became Flesh: frequently asked questions

What does John 1 have to do with large language models?
Both are built on words, but in opposite directions. In John, the Word comes first and the world exists because it was spoken. A large language model works the other way round: the world was lived, a residue of it was written down, and the model learned the statistical shape of that residue.
Can a bigger AI model solve the knowledge problem?
No. The knowledge that matters most is local, dispersed, fleeting, and often tacit, and a language model is trained only on what was put into words. A bigger model is a bigger collection of what could be said, so the knowledge problem is not solved by scale. It is restated at scale.
Why does it matter who is responsible for what an AI says?
Because machine speech is performative only by borrowing a human's authority. A promise or a signed contract changes the world when someone can be held to it. Agentic AI tries to give text the power to act without anyone being responsible for it, which is why a person who bears responsibility has to stay in command.
Why does trust in sales still depend on people?
The deal happens in language: in discovery, in framing, and in articulating a problem the buyer had not yet put into words. The trust that closes it happens in the flesh: in showing up, in being wrong and saying so, and in carrying a consequence.

Ready to see Coevera in action?

Start your 14-day free trial — no credit card, no commitment.

No se requiere tarjeta de crédito · Prueba gratuita de 14 días

groupsTrusted by sales organizations across North America, the UK, Europe and Australasiaverified_userISO 27001 CertifiedshieldGDPR CompliantblockData never used to train AI