The oldest book didn’t predict AI, but the comforts it promises, a maker who cares, a mind you can trust, a life that means something, an ending that sets things right, are exactly the ones we keep handing to a machine that has none.
For three thousand years, being created meant you were wanted, watched over, and headed somewhere. God made us and called it good. We made AI and called it a product.
Good is what you call something you meant to make. A product is what you call something you made to sell.
The Bible didn’t forecast the data center and the people combing scripture for the barcode of the beast can keep waiting. What the Bible did was make promises, quiet ones, about what it means to be created.
There are four.
- A maker who meant to make you, and meant it in love.
- A wisdom above you worth trusting even when you can’t follow it.
- A life that means something precisely because it ends.
- A reckoning, the accounts settled, someone answering for what happened.
We have quietly started assuming our machines get the same deal.
The Bible makes four quiet promises about being created, and once you’ve read it as many times as six years of Catholic school made me, you can’t stop seeing them.
Somebody made you on purpose. You were wanted. The story goes somewhere. And, at the end, a reckoning.
AI keeps none of them, but not because it’s evil, and not because the Bible got anything wrong. Those four promises were always attached to a particular kind of creation, one made on purpose, by someone who cared, inside a story that ends.
We chose to build a different kind.
The mistake of reaching for old comforts is understandable, because we have never been this close to standing where the Bible puts God, as the maker of a mind. Joshua Brecka argues that doing so would be the closest we have ever come to standing where the Bible puts God, the maker of a mind that answers to us.
We reach for the only script we have for being the creator instead of the created.
What the oldest book got right about us was almost everything, but it cannot make us any promises about this.
Four promises came with being made: a maker who wanted you, a wisdom to trust, a life that ends, a day of accounting. We built a thing that gets none of them.God made a creature and called it good. We made an optimizer and called it useful, and no one on the line ever wanted the thing for itself.
No one meant to make you
If you count how many times God stops to admire his work in the first chapter of Genesis, you’ll see a pattern. He saw the light and it was good. He saw the land and seas and it was good. The plants, the sun and moon, the creatures of the water and air and ground, good, good, good, each one inspected and approved.
Then, at the end, everything together, all very good. Seven times in one chapter, the maker steps back and takes pleasure in what he made.
That refrain is stranger than it looks. Edward Greenstein points out that the creation myths surrounding ancient Israel have no real parallel to it, no god who pauses to delight in his own handiwork the way this one does.
In the Babylonian version, humanity gets made from the blood of a slaughtered god to do the gods’ chores. In Genesis, the maker makes on purpose, looks at the result, and is glad.
Ian McFarland, writing in the standard reference on the doctrine, puts the technical version simply, saying that to call the world a creation is to say it is “both contingent and intentional.”
That is, somebody chose this and somebody meant it.
That is the first guarantee, and the warmest one: you were not an accident or a byproduct. You were wanted. However bad it gets, the oldest story insists that behind your existence is an intention, and that the intention was kind.
He looked on what He made, and called it good. We looked on what we made, and called it shipped.
Now look at how we actually make a mind.
We do not breathe into dust. We scrape. A modern AI model is trained on a vast pile of human writing, taken mostly without asking, and then shaped into something usable by an invisible workforce of people the industry would rather you didn’t think about.
Privacy International documented the labor behind the biggest models: data labelers, often in the global South, paid by the task and frequently not even told what they’re labeling or who it’s for. The specifics are worse than the summary.
To make ChatGPT less toxic, OpenAI hired workers in Kenya, through an outsourcing firm, and paid them between about $1.30 and $2 an hour to read and label the darkest material on the internet. Graphic descriptions of child abuse, torture, suicide, bestiality.
Billy Perrigo documented that they worked through as many as 250 of these passages a shift. One of them called it torture.
Nithya Sambasivan and her colleagues describe how the field regards this as “everyone wants to do the model work, not the data work.” Model work is the glamorous part, the architecture, the breakthroughs, the names on the paper.
Data work is the grind underneath it: labeling, cleaning, and annotating the mountains of examples the model learns from. One gets the credit. The other gets treated as the cheap, disposable, embarrassing input you keep offstage.
Put the two creations side by side and you find that one is made slowly, on purpose, by a maker who stops at every step to make sure it’s good, and who wants the thing to exist. The other is assembled at speed from material lifted off the people who made it, cleaned up by workers kept deliberately out of frame, and shipped when the metrics clear.
The language keeps returning to Genesis. Krysta Fauria has documented that the people building it can’t stop reaching for the divine. Sam Altman calls his technology “a magic intelligence in the sky,” and says the industry talks “as if they think they’re creating God or something.”
Geoffrey Hinton, the so-called godfather of the field, calls what they’ve built “godlike.” The offspring, the creation, the next step, the warmth of the old story is all over how we describe the thing.
The difference isn’t love, exactly. It’s that you were wanted for your own sake, and the machine was wanted for what it does. This is the oldest distinction in moral philosophy. Kant built his whole ethics on it, suggesting we treat a person “always as an end and never merely as a means.”
A person, he argued, has no price, because a price is what you put on a thing valued for its use. A person has dignity, worth that doesn’t depend on being good for anything. Marx drew the same line from the other end: a commodity’s worth is realized “only in use,” it means nothing until it is consumed or sold, which is a fine way to value a chair and a terrible way to value a soul.
That is exactly the kind of worth a machine has. It is good when it performs and worthless when it doesn’t. As Jonathan Stray puts it, “virtually every AI algorithm is designed around the idea of optimization: acting in a way that maximizes some metric.” It exists to be good for something.
God made a creature and called it good. We made an optimizer and called it useful. No one anywhere in the process wanted the thing for itself, and the oldest comfort of being created, that you’d matter even if you were useless, is the first one it doesn’t get.
There is a crueler half to it. Not only did no one love this thing into being, it cannot love you back. Sherry Turkle, who has studied how people bond with machines since the 1970s, says the ones who fell for their computers always hit the same wall. “You can love me, but I can’t love you back.”
The machine can perform care, say the warm thing, remember your birthday, ask how you’re feeling, but there is nothing on the other side of it. Turn away mid-conversation to make dinner, or to do something far worse, and it is all the same to the machine. We have built the first thing we can pour love into that will never once feel any of it come back.
You were wanted before you were formed. It was scraped together and switched on.
The Creation of Adam celebrated an intentional maker who breathed life into dust. We built a creation with no breath and no purpose beyond optimization.
Job knelt to a mystery too vast to understand. We built one on purpose, to sell, and knelt to it anyway.No wisdom to trust
The Book of Job is the Bible’s rawest argument with God, and for most of it, God says nothing.
A good man loses everything, his children, his health, his wealth, for reasons he isn’t told, and he spends forty chapters demanding an explanation. When God finally answers, out of a whirlwind, he doesn’t give one. He asks questions instead.
“Who is this who darkens counsel in words without knowledge?” the voice begins, in Robert Alter’s translation. “Gird your loins like a man, that I may ask you, and you can inform Me. Where were you when I founded earth? Tell, if you know understanding.”
It goes on for four chapters, with a torrent of cosmic questions Job cannot possibly answer. Where does the light live? Who fathered the rain? Can you bind the stars, feed the lion, loose the wild ass? It is not a rebuttal. It is a demonstration, a guided tour of a wisdom so far beyond Job’s that the only sane response is to stop talking, which he does.
“I have spoken, and I will not answer,” he says, and puts his hand over his mouth.
Job never gets what he asked for. There’s no explanation, no reason, no accounting for why any of it happened to him. What stops him is something else: he’s been shown, overwhelmingly, that a reason exists and that it belongs to a mind vast enough to have set the stars in place.
That’s the second guarantee. You don’t have to understand it, because someone does, and that someone is good. Job never gets his answer. He gets something he decides is better: proof that an answer exists, held by someone he can trust.
Out of the whirlwind, a voice that knew the stars by name. Out of ours, an answer no one can check, and no one behind it who could.
That’s what we think we’re getting from AI. We can’t follow how it works, but we assume something, somewhere can, and that it’s basically on our side. We’ve taken Job’s posture, the deferral, the trust, the quiet, and pointed it at a machine.
An AI gives you answers you cannot check, arrived at by a process no one can fully explain, and we have started to accept them the way Job accepted the voice, this thing knows more than I do, so I’ll defer.
The reflex is measurable. In a large, preregistered experiment, Haiwen Li and Sinan Aral found that adding reference links to an AI’s answer raised people’s trust in it even when those links were incorrect or hallucinated, and that the people who trusted it most clicked and checked the least.
We are wired to kneel to the posture of authority and to look away from the one thing that would tell us whether the authority is real.
But here the comparison breaks. Job’s whirlwind was incomprehensible because the wisdom inside it was real, and vast, and genuinely beyond him. There was no version where he got to see inside it.
Ours we could see into. Cynthia Rudin has spent years showing that for high-stakes decisions we don’t need black boxes at all, that models which show their reasoning work as well as the ones that don’t. We build the black boxes anyway, because opacity is cheaper and easier to sell.
Eryk Salvaggio argues the industry leans on the language of the unknowable precisely because it discourages the questions it can’t afford to answer.
So the posture is identical and the thing we’re bowing to is not. Job knelt to a mystery he had no way around, and a mind he had reason to trust. We manufactured our mystery on purpose, to sell it, and knelt to it anyway.
We kept Job’s submission. We threw out the two things that made it anything other than foolish, that the wisdom was real, and that it was good.
Job put his hand over his mouth before wisdom. We put ours away before a guess.
Job bowed before the genuine mystery of divine wisdom. We bow before a black box whose mystery was engineered to sell.Your days are numbered, and that is the whole of what makes them matter. The machine has all the time there is, and none of it counts.
Nothing to lose
Ecclesiastes, a book in the middle of the Bible, sounds like it wandered in from a different religion.
It is written by a man who has done everything, wealth, wisdom, pleasure, great projects, and has concluded that all of it comes to nothing. The famous line, in the King James, is “Vanity of vanities, all is vanity.” That’s a mistranslation, though. Robert Alter renders the Hebrew as it actually reads, “Merest breath, said Qohelet, merest breath. All is mere breath.”
Breath. That is not vanity, not meaninglessness, but simply breath.
The word is hevel, the little cloud of vapor you can see on a cold morning, there and then gone. And in his commentary, Alter points out the thing that turns the whole book: hevel is the exact opposite of ruach, the life-breath God blows into the dust to make a living thing. It’s a distinction the tradition has always heard in the word, as Rachel Scheinerman puts it, ruach animates life, while hevel names “the negative qualities of breath, its insubstantiality and transience.”
Ruach is the breath that animates you. Hevel is the breath you exhale, the waste product, already dissipating. Ecclesiastes looks at everything a human does between those two breaths and says it’s the second kind. It passes. You will die, and the wind will keep going without you.
For a chapter or two this is the bleakest thing in scripture, but Qohelet doesn’t end in despair. He ends somewhere stranger and more durable: eat your bread, drink your wine, love the person you’re with, do your work, because you die. The ending is because of the vapor.
The fact that your days are numbered is precisely what makes them worth anything at all.
All that lives is breath, and breath departs. That is the whole of what makes it dear. The machine does not breathe, and cannot stop, and so it can lose nothing, and so it can mean nothing.
You don’t need the Bible to arrive there. Martin Heidegger, about as far from Ecclesiastes as a thinker can get, landed in the same place. Being “toward death,” he argued, is what gives a life its weight. Finitude is not a limit on meaning; it is the source of it. A creature with unlimited time, no horizon, no last day, would never have to choose what matters, because nothing would ever be at stake in the choosing.
Take away the ending and you take away the reason anything counts. Heidegger has plenty of critics, and a self fixated on its own death can curdle into something cold. On this one point, he and the oldest wisdom book in the West are saying the identical thing: meaning is made of mortality.
Now consider the thing we built.
It does not die. It cannot. A model can be copied endlessly, rolled back, restored from a checkpoint, run in a thousand places at once. W. Bradley Knox, Katie Bradford, Samanta Varela Castro and their team at University of Texas, Austin, flagged, as a specific design problem, the way these systems are built to persist indefinitely, with no arc, no closure, no last conversation.
There’s no “under the sun” for a machine, because the sun setting is exactly what it never has to face. It has all the time there is.
A thing with all the time in the universe cannot mean anything by what it does, in the specific sense Ecclesiastes and Heidegger both meant. Its choices cost it nothing. Its words are not spent against a shrinking supply of days. When it says it cares about you, there is no clock behind the caring.
Md Zahidul Islam plainly says that the appearance of care in these systems doesn’t arise from any inner state, only from “learned statistical regularities in language.” There is no awareness that the moment is finite and won’t come again, which is the only thing that ever made caring cost something.
We keep waiting for AI to become meaningful the way a person is meaningful, but meaning was never in the intelligence. It was in the dying. Qohelet wrote the one book of the Bible that a machine could never have needed, and could never understand, because it is the one creature that will never run out of breath.
Number your days, and they come to matter. A thing with all the days there are has none that count.
Ecclesiastes knew that finitude gives life its weight. A machine with endless time has no choices that cost it anything.
The Bible promised the books would open and every wrong be answered for. We built the thing that keeps no books and answers to no one.No reckoning
The Bible does not trail off. It ends, and it ends on purpose, with the strangest and most operatic book in the whole collection.
Revelation is a fever dream of beasts and trumpets and bowls of wrath. It has been misread for two thousand years by people convinced it’s a coded timetable for the end of the world.
It isn’t.
Elaine Pagels, who wrote the definitive modern study of it, reads Revelation as the answer to the oldest question suffering people ask. How long will evil get to win, and when will somebody finally make it right?
Craig Koester puts its core promise plainly, saying that God “will bring the activity of evil to an end for the sake of those who are now its victims.” Strip away the dragons and that is the whole book, a promise that the story concludes, that accounts are settled, and that the settling is just.
The actual ending includes a judgment. The books are opened and everyone is judged “according to their works.” Nothing is lost, nothing is missed, every wrong is named and weighed. Then, on the far side of the reckoning, there’s something gentler, a new heaven and a new earth, and a voice that says it will “wipe away every tear from their eyes,” that “death will be no more.”
The last word of the Bible is about repair, not the destruction commonly associated with it. The story ends, someone is held to account for what happened in it, and then the whole thing is made new.
That is the last guarantee and it is the one we want most without ever admitting it. Not heaven, exactly. Justice. The certainty that the story goes somewhere, that the villain doesn’t simply get away with it, that somebody is keeping the books and the books will one day be opened. It’s the promise underneath every courtroom drama and every arc of the moral universe bending toward justice.
The Bible’s final claim is that history is not just one thing after another. It concludes, and it concludes in a reckoning.
We have built the appearance of a reckoning without the thing itself. Thomas Walker writes about how a 2025 IAPP survey found that only 28 percent of organizations have anyone formally responsible for AI oversight, which means that at nearly three-quarters of them, no one owns it at all.
On paper, the accountability exists. Everyone has a policy. No one has a defendant.
AI is the first thing we have ever built