The Daily Mail found me on the porch at the shore, on Labor Day, when the town gives itself back to the people who pay its taxes. The headline said a medical scientist had broken ranks. It said human DNA carries evidence of an intelligent creator. I know what the Daily Mail is. I clicked anyway.

For about ninety seconds, it worked on me.

I need you to understand why that is embarrassing. I write about tells for a living. I have spent this year arguing that the scarcest resource in AI is the ability to know when you are being worked. And there I sat with the ice cubes melting in my cold brew, wanting a tabloid to be right, because I have believed a version of its conclusion since I was eight years old.

My mother dragged me to antique shows and house sales all over Northeast Ohio. Church basements, dead men’s garages, fancy stores filled with delicate people in Chagrin Falls and Shaker Heights, banquet halls with folding tables and cigar-box cash registers. She went for the glassware. I went straight to the watches and stayed there for hours, a small boy bent over the cases like a jeweler with no inventory. Worn devices that had marked the lives of strangers, wound by hands that were gone, discarded now, unwanted, available at a negotiable price. Filled, I was certain, with evidence of God.

Consider what a watch is. A man sits at a bench with tools, magnification, and patience, and from a solid piece of metal he builds an instrument that tracks the universe and a life. Then consider what these particular watches had tracked. Births and deaths. Weddings. Sadness and joy. Christmases and summer holidays, all of them timed to the second for strangers I would never meet, on wrists that had gone into the ground while the watches kept running. I stood at those tables and felt every one of those strangers connected to me at some level smaller than atomic and greater than quantum. That is a child’s physics. I have since learned enough real physics to fail the sentence on inspection, and enough of everything else to keep it.

So when a man with a doctorate in medicinal chemistry told the Daily Mail that the code in our cells provides evidence for intelligence, some eight-year-old in me stood up. Then the fifty-six-year-old went looking for the paper.

There is no paper. There is a book. Orson Wedgwood, DNA: The Elephant in the Lab, self-published in January 2019, sixty-seven thousand words, shelved by its own distributor under Religion and Science. The tabloid piece running seven years later as an exclusive is a press cycle, and the man breaking ranks was never in the ranks he is breaking from. I bought the book anyway. I bought it twice, actually. The first store sold it to me and then kept it, the way ebook stores do now, readable only inside their walls, mine in the sense that a hotel room is mine. The second store let me carry it out. Evidence of God, $7.99, twice. At the folding tables, the price was at least negotiable.

The book’s centerpiece is a number. Wedgwood calculates the odds of a single working protein assembling by chance at one in ten to the 164th power. That number has 164 zeroes. The universe, he notes, only ever had the capacity for ten to the 120th tries. Case closed, he says, and to be fair, it would be, if anyone anywhere claimed the game was played that way.

The calculation models a lottery: all 150 characters of the password guessed at once, every miss resetting to zero. Chemistry plays a different game. Chemistry keeps its winnings. Anything that copies itself, however badly, starts an audit trail, and every improvement is locked in before the next guess.

The load-bearing piece of his number is an estimate that only one sequence in ten to the 74th will fold into anything useful, a figure arrived at by mutating a single modern enzyme until it broke and extrapolating outward. In 2001, Jack Szostak’s lab at Massachusetts General ran the direct experiment instead, building a library of trillions of genuinely random eighty-amino-acid sequences and fishing out the ones that could bind ATP. They found four.

The rate worked out to roughly one in a hundred billion. That is a test tube’s worth of luck standing sixty-three orders of magnitude away from the number the book is built on, and the book is made of that gap.

But the number is a sideshow. The real merchandise is a sentence.

In bold, in the book’s early pages, sits a line from a 2017 opinion piece in Nature Reviews Chemistry by John Sutherland of Cambridge, one of the two or three most important origin-of-life chemists alive. Sutherland’s piece carried the title “The end of the beginning.” That is Churchill, after El Alamein, declaring the war’s first won battle. Sutherland’s own lab had just shown how the building blocks of life, the parts for RNA and proteins and cell membranes, could all emerge from the same simple cyanide chemistry on the early Earth. His piece said the field remains far from the finish and “may have nearly finished the first phase: the beginning.” A victory speech, carefully hedged, from a man in the middle of winning.

The book bolds the hedge and calls it a confession.

So I walked the sentence upstream, and here is its chain of custody. January 2017: Sutherland publishes measured optimism under a victory title. July 2017: the Discovery Institute’s house blog runs the piece under a headline announcing that an origin-of-life researcher “admits” the road is long, dismissing Sutherland’s actual view as overly optimistic. 2019: the self-published book inherits the inverted reading, bolding included. September 2026: the Daily Mail sells it to the world as a scientist breaking ranks, and a man on a porch in New Jersey nearly buys it in ninety seconds flat. Nine years. Four owners. The meaning flipped at the second handoff and stayed flipped, because every layer cited the layer above it, and nobody walked back to the source. The inversion traveled the whole way free of charge. Checking it cost me $15.98, one blocked crawler, and a journal paywall asking $39.95 for a paper whose abstract was the entire crime scene.

Here is what I hold against the book, and it has nothing to do with God.

The watchmaker at his bench made an honest instrument. Whatever else you believe about him, the watch tells the time that is actually passing. The book is a rigged instrument, a movement assembled to read a predetermined hour: the lottery that models no known process, the quote inverted against its own title, the citation to a journal one notch more famous than the real one. And a rigged watch insults in both directions. It cheapens the science it impersonates, and it cheapens the God it is offered to, who presumably has no need of doctored evidence and has never once been found inside a gap that stayed closed.

Because what I had at the folding tables was never a proof. That is its entire strength. Nobody can flip my quote or audit my math. There is no math. I don’t believe it because the odds are ten to the 164th against. I believe the way you recognize a face. I know what pareidolia is. I also know what I saw.

I am standing in a very old line, too. The watch may be the most contested object in Western theology, handed down for two thousand years with the meaning flipped at nearly every handoff. Cicero started it before there were watches, arguing that a man who looks at a sundial or a water-clock understands at once that somebody intended it. A country clergyman named William Derham published a nuts-and-bolts manual on watch and clock construction in 1696, having taught himself the craft by pestering Hooke and Tompion, and then argued God from the design of Nature in Physico-Theology in 1713, one man writing both the how-to for the bench and the sermon of the bench. I would trade most of the canon for his library card. Paley found his watch on the heath in 1802 and gave the argument its famous form. Dawkins picked the same watch up in 1986 and blinded the watchmaker, in the open, on the cover, under his own name.

Every man in that line was arguing about a watchmaker. Not one of them was a watchmaker. Derham came closest, and he was a parson with a hobby. The genuine article was alive while the most famous version of the argument was being written, over in Paris, and his name was Abraham-Louis Breguet. He patented the tourbillon on the 26th of June, 1801, one year and one country away from Paley imagining his. Breguet was faked so relentlessly in his own lifetime that he had begun, back in the 1790s, engraving a secret signature into his dials, nearly invisible, a diamond point dragged through the surface, showing itself only when the light rakes across at the right angle. He had brought guilloché to the dial face in 1786 partly for the same reason. Every Breguet made since carries the signature, including the one on my wrist as I write this. The greatest watchmaker who ever lived built the authenticity check into the face of the instrument, because he knew precisely what was coming for it. The answer he landed on two centuries ago is still the only answer there is. Put the proof in the object, where anyone willing to turn it to the light can find it.

Dawkins is still the one I owe an answer to, so let me pay him properly and not with a straw version. Selection is a complete account. It requires nothing added from outside, no nudge, no first push, no debt to anybody. Anything that copies itself with errors will climb, and the climb is not random; given four billion years and an ocean, it produces eyes and wings and the ribosome and the man at the bench. I described that ratchet myself a few paragraphs ago to knock down a rigged number, and I meant every word of it. Then I sat down and watched a machine write three hundred genomes with no mind involved in any single choice, and sixteen of them lived. If you want the strongest case that watch-like things arrive without a watchmaker, I have been handling the evidence all day, and I will hand you the file.

I see it anyway.

That is the whole confession. Not an argument, and I am not offering it as one. I am telling you that a man who spent this week auditing a design argument down to the studs, who can show you exactly which zero the fraud lives in, still stood at those tables at eight years old and knew what he was looking at, and stands at his bench at fifty-six and knows it again. You may file that under pareidolia. I would only ask you to notice that the same faculty that sees faces in clouds is the one that recognizes your mother across a parking lot, and nobody has ever proposed we switch it off.

Notice what that is. A chain of custody, two millennia long, the object passed on and inverted again and again. The difference between that chain and the one that carried Sutherland’s sentence is that every man in it signed his inversion. The tradition is one long honest argument. The laundering was a forgery. You are allowed to turn an inherited object over in your hands and read it against its previous owner. You are required to say that is what you are doing.

And I still see it, most days, at my own bench. I grew up into the man with the tools and the magnification, except the metal now is data: raw records, millions of them, unverified, each one claiming to describe a life. The work is the watchmaker’s work. You take the raw form, and you make it into something that can measure and mean, and you check the record before anything downstream is allowed to trust it, because a better model trained on unverified records is a better way to be wrong. That sentence is the whole trade. It also describes the chain that carried Sutherland’s sentence for nine years, and it describes what happens next, at scale, when the machines do the reading. Einstein described the physicist’s condition as a man studying a watch whose case can never be opened, free to imagine the mechanism, never able to check his picture against the works. He wrote that in 1938. We have since built the closed watch ourselves and taught it to speak, and it imagines its mechanism from whatever we feed it. The models are eating the downstream copies right now. Unless someone walks upstream on their behalf, they will cite the confession and never find the victory speech, fluently, confidently, forever.

The watch is only the latest answer to a very old question, and the older answers are stranger than most people suspect. Until the railroads, every American town kept its own noon, clocks set to the sun directly overhead. Hence, Pittsburgh and Philadelphia disagreed by minutes and nobody minded until the timetables made them.

In industrial England you could hire a knocker-upper, a human alarm clock paid a few pence a week to rap your bedroom window with a long pole or shoot dried peas at the glass until you stirred, and the trade’s own riddle asked who knocked up the knocker-upper. Sailors told speed by throwing a knotted rope off the stern and counting knots against a sandglass, which is why ships still travel in knots. Sailors died by the thousand for want of a clock that could keep true time at sea, until John Harrison, a carpenter with no formal training, spent his whole life building one accurate enough to find you on the ocean.

Alfred the Great burned marked candles, six to see him through a day, and when drafts made the flames run fast and lie to him, he built them a lantern of wood and translucent horn to keep them honest.

The Chinese built incense clocks that burned a different fragrance each hour, so you could wake in the dark and smell the time.

In Athens the courtroom clock was a clepsydra, literally a water thief; a lawyer was allotted so many jars, and when your water ran out, so had your argument. For most of human history the hour itself would not hold still.

Rome and medieval Europe cut daylight into twelve hours whatever the season, so a summer hour swelled and a winter hour shrank, and the hour fit the day instead of the day fitting the hour.

Galileo, bored during a service in the cathedral at Pisa, timed the swing of a hanging lamp against his own pulse and founded the physics of the pendulum on a heartbeat, in a church. The story comes from his first biographer and is probably polished, and in this essay of all essays I am obliged to tell you so. And when there was no instrument at all, people timed the world in prayer. Cooks were told to boil the egg for the length of a Miserere and to stir for an Ave Maria. The prayer was the unit.

One more walk upstream, the longest one.

The mechanical clock was born in medieval monasteries, built by monks who needed to be woken for the canonical hours, machines made so that men would not be late for prayer. Lewis Mumford argued in Technics and Civilization that the clock, and never mind the steam engine, is the key machine of the industrial age, and it was invented in the service of worship. Follow any watch’s provenance far enough, past the estate sale, past the factory, past the guild, and you arrive at men building machines to call one another to God. The boy at the folding table had no idea he was holding an object whose original spec was prayer.

The walk forward ends in a vault.

Since 1967, the second has not been a slice of daylight at all. It is 9,192,631,770 oscillations of a cesium atom, the same count in every laboratory on Earth, the definition of time lifted off the heavens and set inside a single element. The best of these clocks now run so true they would not slip a second in the lifetime of the universe. Raise one a foot off the bench, and it ticks measurably faster, because gravity is weaker up there, and the instrument has become fine enough to see Einstein in the space above your desk.

And we still check it against the world.

The Earth is a sloppy timekeeper, wobbling, dragging on the tides, and lately spinning a hair faster, so since 1972 the keepers of time have inserted leap seconds into atomic time to keep the count true against the planet’s actual turning. Twenty-seven of them, the last one in 2016. The most precise instrument our species has ever built gets corrected by the thing it measures, and that correction is the entire definition of an honest instrument.

In November 2022, the international body that governs the second voted to abolish the leap second by 2035, largely because the correction wreaks havoc on software, satellites, and trading systems, and a draft resolution published this past January proposes moving the execution date up to May 2027. We are retiring the reconciliation because it is inconvenient. Remember that the next time someone tells you the friction is the problem.

Between the vault and my wrist sits the thing that broke an industry. Quartz has a property called piezoelectricity: squeeze the crystal, and it puts out a voltage; apply a voltage, and it flexes; and if you cut it to the right shape, it will vibrate at a rate so stubborn that heat, shock, and the passage of years barely move it. Cut it into a tiny tuning fork, and it hums at 32,768 times a second, a number chosen because it is two multiplied by itself fifteen times, so a simple circuit can halve it fifteen times over and arrive at exactly one pulse per second. That is the whole trick. A stone that keeps a beat, and a counter that divides the beat down into a heartbeat you can put on a dial.

Seiko sold the first one on Christmas Day, 1969, a gold watch called the Astron that cost about what a car cost and kept time to five seconds a month. The best Swiss chronometers that year were off by five seconds a day. Switzerland had actually built a quartz prototype first. It hesitated to sell it, unwilling to cannibalize the mechanical business that was its whole identity, and the hesitation cost the country its industry. Swiss watchmaking employment fell from roughly 90,000 in 1970 to around 30,000 by 1983. Something over a thousand firms went under. The survivors lived by inventing a new story in which a mechanical watch is not a timekeeping instrument at all; it is jewelry with a heartbeat, and that story is why a hand-wound watch costs more than a car today while a five-dollar circuit keeps better time. The Swiss call it the quartz crisis. Everybody else calls it the quartz revolution. The name depends entirely on which side of the disruption you were standing on, which is a sentence I would like to carve above the door of every AI conference I speak at.

My own watches are the peace treaty.

Grand Seiko builds a movement called Spring Drive, dreamed up by an engineer named Yoshikazu Akahane in 1977 and taking twenty-eight years and six hundred prototypes to reach the wrist. It has a mainspring, wound by hand or by the motion of my arm, and no battery. It also has no escapement, which is to say none of the ticking, oil-hungry, five-hundred-year-old assembly that every other mechanical watch uses to portion out the spring’s energy. Instead, a wheel spins eight times a second, generating a whisper of current from a coil, and that current wakes a quartz crystal and a small integrated circuit. The crystal keeps the true beat. The circuit compares the spinning wheel against it and applies an electromagnetic brake, with no contact and no friction, thousands of times a minute, holding the wheel to its eight turns. Mechanical power, electrical power, magnetic power, three kinds of energy in one closed argument. They assemble them by hand in a studio in Shiojiri, in Nagano’s mountains.

The visible result of all that is one thing. The seconds hand does not tick. It glides, in a single unbroken motion, because nothing in the movement is starting and stopping. Every watch I looked at as a boy chopped time into pieces and handed them to me one at a time. This one refuses to. It shows time the way time actually behaves, which no clock in the history of the world had managed to do, and it does it by making the mainspring and the crystal agree with each other. The oldest technology in the case and the newest one, one holding the power and the other holding the truth, checking each other eight times a second, forever, on my wrist.

Here is the part that undoes me. I buy many of these watches from a man in Cleveland. Teddy Baldassarre posted a video about his own small collection in 2017 from Northeast Ohio, told nobody he had done it, and turned it into a YouTube audience in the millions and an authorized dealership carrying Omega and Grand Seiko and Blancpain and fifty more, with a boutique out at Crocker Park in Westlake, about forty minutes from the church basements I was dragged through. I do not get back to Cleveland much anymore, so I buy online from a store I have never walked into, in the county I am from.

Understand what that means to the boy.

I believed, with total conviction, that this knowledge lived in Switzerland. In valleys. Behind doors. That there were old men in aprons in La Chaux-de-Fonds who knew the things, and the rest of us were permitted to look through glass at the results and pay whatever was asked. That was the whole architecture of it: the craft was elsewhere, it was guarded, and my job was to be a customer. The Swiss lost their industry partly because they believed the same thing about themselves. And now the most trusted voice in the trade is a guy from the East Side who taught himself out loud, in public, in front of a camera, and the hub is a shopping center in Westlake, Ohio. The knowledge did not stay in the valley. It never wanted to.

Another man is still up there: Michael Cavotta.

Mike and I grew up together in Northeast Ohio, and he left, traveled the world, and returned. I never did. He is fifty-six, same as me, and he marks the passage of time the way an athlete does, in what the body will still agree to do on a given morning, which is a harder and more honest clock than anything on my wrist. He photographs people out of a studio in Mentor, and he refuses the verb. He does not take a picture, and he does not capture anyone. He says he witnesses them. He has also written a book called Chasing Light, part meditation and part manifesto, written for fathers and sons and the women who have to live with both, for young men at war with themselves and older ones who never finished the fight. He sells it bound in aniline-dyed leather with copper foil on the cover, and he reads it aloud himself on the audiobook. He is also the person I call. Half the ideas I have published get their first hearing from Mike, usually in the form of him telling me the part that was not true yet.

Which is why it undid me a little to go to his site and find he had written a long piece on artificial intelligence and authenticity, and to stand there reading my own argument in another man’s voice. He writes that AI releases us from the friction of authentic presence. Friction. His word, arrived at on his own, four hundred miles from mine. He writes that it has become easier to appear as you wish than to show up as you are, and he traces the whole slide from the first profile photos through the filter apps to a generation looking in the mirror and saying that is not what I look like.

Then, because whoever runs this has a sense of humor, our one disagreement. Mike has planted his flag on the em dash, four centuries old, an instrument Melville reached for every hundred and twenty-nine words in Moby-Dick, now treated as a tell for machine writing, and he will not surrender it to a robot that learned it from us in the first place. I went the other way. I stripped every one of them from my work and haven’t set one down since. Same problem, same fifty-six years, same rust belt, opposite trenches. He keeps the mark to prove a man wrote it. I drop the mark to prove the same thing. Neither of us is going to win this, and I am not certain either of us is wrong.

At the very bottom of his website, in the small print where the copyright notice goes, Mike has written that the typos, bugs, and glitches prove this website was built by a human. He is not apologizing. He is filing the certificate of authenticity, and the flaws are the signature. Which is the entire argument of the essay you are reading, arrived at independently, in a footer, by a photographer in Mentor, Ohio.

Read Mike. I mean that as more than a courtesy to a friend. He has been thinking about presence and attention for as long as I have been thinking about machines; he got there through a lens instead of a keyboard, and he writes it better than I do.

You can take the boy out of Ohio. The Ohio does not come out of the boy.

I own watches now that do all of this without me.

I land in London or Tokyo and by the time I reach baggage claim the thing on my wrist has already fixed itself, pulling a signal down from a longwave transmitter in Fort Collins or from a satellite overhead, setting the hour and the date and the zone while I stand there holding a bag. It is the most quietly astonishing object I own, and I have never once checked it.

Sit with that a second.

The boy at the folding table wound dead men’s watches by hand and held them to his ear to find out whether they were telling the truth. The man wears an instrument whose truth is guaranteed by a chain he never sees: a clock in a satellite that runs about thirty-eight millionths of a second fast every day because it is higher up and moving quicker, corrected for relativity, or the navigation drifts kilometers off within a day, calibrated against cesium in a vault, reconciled against the actual turning of the planet by a committee that has just voted to stop reconciling. Every link in that chain is somebody’s careful work, and I take all of it on faith. That is a funny admission to make in this essay. I am going to let it stand.

And here is the joke the tabloid missed: DNA astonishes with no help from rigged arithmetic, because DNA is itself a clock, three times over.

It keeps time: every cell you own contains a loop of genes that switches itself off and on in a cycle a shade over twenty-four hours, the discovery that took the 2017 Nobel Prize, and jet lag is thirty trillion small clocks arguing about the hour. It records time: in 1962, Linus Pauling and Emile Zuckerkandl noticed that mutations pile up at a roughly steady rate, so two strands of DNA laid side by side will tell you how long it’s been since their owners were one family.

Biologists call it the molecular clock.

They calibrate it against fossils, they argue about its rate, and they publish the error bars, which is what an honest instrument does. Steve Horvath showed in 2013 that the chemical wear on your DNA reads out your biological age, sometimes more truthfully than your paperwork. The telomeres at the ends of your chromosomes burn down a little with every division, Alfred’s candles in miniature, and no lantern to shield them. And it can be built into time: in 2000 two physicists working in biology wired three genes into a ring, each silencing the next, and the loop oscillated, blinking a green protein like a second hand. A clock made of DNA, by men at a bench, with tools and magnification.

So the molecule in the headline turns out to be the most remarkable timepiece ever laid on any table, and the boy’s physics turns out to be on the record. The strand has been timing births and deaths since the first cell. Lay any two of us side by side, and it will tell you, in plain mutations, how long ago we were one family. Connected at some level smaller than atomic. You can look it up.

And now we are writing the clock.

In February 2025, the Arc Institute and NVIDIA released Evo 2, a model trained on 9.3 trillion nucleotides pulled from more than 128,000 genomes across all three domains of life, published in Nature this past spring. It reads DNA the way a language model reads English, and it writes it the same way. Last September, a Stanford and Arc team fine-tuned the Evo models on about fifteen thousand small viral genomes and asked for new ones.

They synthesized the designs and dropped them into E. coli. Sixteen came alive, replicating, bursting their hosts; their particles were caught afterward under the electron microscope, several of them working as well as the natural virus they were modeled on, or better.

The organism was ΦX174, a five-thousand-letter phage with eleven genes that has been the field’s laboratory rat for half a century, the first DNA genome ever sequenced and the first ever synthesized. It infects bacteria and nothing else. The team ran the work under safety protocols that exceeded the standard and kept human-infecting viruses out of the training. All of that is true and worth saying plainly, because the number I actually keep is different.

They generated about three hundred designs. Sixteen lived. The machine wrote fluently, confidently, at speed, and it was wrong two hundred and eighty-six times, and the only thing on this Earth that could sort the living from the dead was the bench. Chemistry graded the papers. Nothing else could have.

Then there is the part that keeps me up.

When a lab orders synthetic DNA, the order is screened against a library of known dangerous sequences, and spelling is checked against a list. A 2026 review in Frontiers in Microbiology makes the obvious problem explicit: a generative model can produce a sequence that does the same work as a known toxin while sharing almost none of its spelling, and a screen that hunts for resemblance to the record comes back clean. David Baker, who has a Nobel for this work, and George Church, who has been at the front of it for forty years, have both said in public that the safeguards need building now. Verification trails generation. It always has. That is my whole trade described in a field with far worse consequences than mine.

So Paley’s watch is being written. At a bench, by men with tools, magnification, and a machine that has read every genome we could give it. Three hundred designs from a model that has read more DNA than every biologist who ever lived, combined, and reality accepted sixteen. The universe is still the grader. It cannot be argued with, it cannot be quoted out of context, and it does not care what the model was confident about. We have built the most fluent instrument in history, and we still have to hold every sentence it writes up against the world to find out if it is true.

My mother never knew what she did. She was after the glassware. She wanted an afternoon out of the house and a child who would stay put for an hour, and she got a boy who would stand at a case of dead men’s watches until she came looking for him. Nobody sat me down. Nobody explained. She dragged me through other people’s finished lives, and I found the argument lying there on a folding table with a price tag on it, and I have been carrying it around ever since without ever once telling her that is what happened in those rooms.

The precision was never the miracle. Any rigged movement can be precise. The miracle is the honest instrument, the thing built true enough to carry what it measures: time, a record, a sentence with its meaning still attached. That is what I was looking at through the glass in those banquet halls, and I have decided the boy read the evidence correctly. The end of the beginning. Churchill meant the first battle won. Sutherland meant the building blocks.

I mean a kid in Ohio holding a dead man’s watch to his ear, astonished to find it still keeping perfect time.

The ash was about an inch long when Alena reached for it.

We were on the patio in her and Rheanna’s backyard last week, for Denise’s birthday. Rheanna is Jade’s sister, nineteen years younger, and Denise is their cousin. It was the kind of August night where the heat has nowhere to go after dark and just sits on you. I was happy in the uncomplicated way I am rarely happy anymore, which is to say I was surrounded by people I love and nobody needed anything from me. Bruno was missing. He has been working late nights scooping ice cream at the mini-golf place in Lavallette, the last summer job before London. He drives himself there and drives himself home around midnight, sticky to the elbows and half asleep and sore in the shoulders, and delivers a nightly lament on the parenting habits of millennials who order their children coffee ice cream with caramel sauce and chocolate sprinkles at nine-thirty at night. I had already done the arithmetic on how few of those midnights were left.

In September, he goes to London to study costume design for film. He has been making clothes since he was small enough to need help with the scissors. He spent his whole school career building costumes for productions that had an actual wardrobe department, won awards for them, and finished as a senior running the department himself, a team of kids, everything from the hats down to the buttons. A dress form has been in his bedroom for years. I mention this because I am going to spend a few thousand words on the patience it takes to make a thing by hand, and the person who taught me most of what I know about that is eighteen and was, that night, scooping mint chip for someone’s overtired toddler.

So I took out a cigar. A Partagás Línea Maestra Origen, the Cuban one, a figurado with a rounded belicoso head that the magazine I will tell you about shortly had just scored 93 points and called the best of in the issue. Partagás is my brand, and I have stopped pretending to shop around. The name goes back to a factory in Havana in 1845, and how one of its cigars came to be on a patio in New Jersey is between me and the friend who handed it to me.

Jade, my wife, groaned. She does not hate that I smoke cigars. She hates everything else about them: the smell, the time, the way I talk about them, and she has made her peace with the habit the way you make peace with weather. Denise sniffed the air and said it smells like my Dad, in the tone of someone finding an old coat. Then Alena said the exact same words: it smells like my Dad, and meant the opposite. Her father is a fisherman and a cigar man, and the smell puts her back on a dock in Key West with him, and she leaned toward it instead of away.

I lit it. I moved to the far corner of the patio out of consideration for my wife, then hijacked the Bluetooth speaker and put on country music, with no consideration for anyone. Jade and Denise looked at each other and got into the pool, which is the correct response to both.

That left Alena, and Alena made the mistake of asking me about my cigar.

You should know something about me here, and my family will confirm it. When I love a thing, I have to tell you everything about it, whether or not you asked, and long after the point where a normal person would have read the room. It is the running joke of my life. Cigars are my dinosaurs now. I can give you the wrapper shades and the ring gauges in sixty-fourths of an inch, the difference between a leaf from the top of the plant and one from the bottom, and what happened in Havana in 1865. I will give it to you, on a patio, at a birthday party, to anyone who so much as glances at the thing in my hand. That is who I am socially. Anti-socially, Jade would say from the pool. This essay is what I did to Alena.

Somewhere in the middle of it, she said, “Can I try it?”

I should tell you now that this is a thing you never do, not in a lounge, not on a patio, not ever. A lit cigar is the one object in the world you do not pass. It has your mouth on it and your hour in it, and the etiquette on this point is older and firmer than most of the laws in this state. Men who will lend you their car will not lend you their cigar. I know all of this. I have read the books. I handed it to her without a second’s thought, because she asked, and because love is greater than decorum, and anyone who has ever been asked for something by someone they love knows the rulebook goes quiet at that moment.

She heard the title of this essay before I did, said it, and laughed. Then she did what everyone does. She looked around for somewhere to knock the ash off.

I said, “Don’t.”

She looked at me like I had asked her to hold a lit fuse, and I understood the look. Every piece of cultural training she had ever received about a burning thing in your hand says tap it. Tap it constantly. Tap it before it falls on your shirt. Tap it because a cigarette is a delivery system and the ash is exhaust.

A cigar works the other way round. The ash is not waste. It is a cap. It sits on the coal and slows the air getting to it, which keeps the burn cooler, which keeps the smoke sweeter, which keeps the tar and the bitterness where they belong: nowhere near your mouth. Knock it off, and the cherry flares, the temperature climbs, the leaf scorches instead of toasting, and the thing turns harsh in about three draws. The ash is a buffer between you and the fire. You let it build. You let it get long and a little frightening. You let it fall when it falls.

She held it wrong, she drew on it right, she coughed, she laughed, and she said, “So you just wait?”

“Yes. You just wait.” I said.

I have spent three years telling boardrooms a version of this about AI. That friction is form. That the effort a thing asks of you is not separate from the meaning it gives back, and that when the machines take the effort, the meaning tends to leave with it. I have slides. It took a sister-in-law on a patio, with her wife’s cousin in the pool, to say it in five words.

The stupid, annoying, unhealthy thing

Let me get the confession out of the way, because I know how this reads.

I am fifty-six. My friends are getting old in the specific way where you notice it at dinner. The reading glasses. The hair that is a shade too even to be true. Somebody’s back. And the exhaustion that arrives with the entrees now, so that by the time the check comes, half the table is already home in their heads. My son leaves for university in a few weeks. Rheanna, whom Jade and I have been raising in one way or another since she was small, alongside their mother, whom I have moved to live near us more than once and who is helpful in the way the mothers-in-law on The Sopranos are helpful, which is what you get when you marry into North Jersey royalty, married a woman I adore this year. I walked her down the aisle to Alena. The two of them are building a house full of a life I get to visit. By every metric I used to care about, I’m doing fine. Two companies. A book out to publishers. A column. A wife who still laughs at the same jokes I have been telling since Hoboken.

And I have taken up cigars.

People hear this, and they do the arithmetic quickly. Midlife. Mortality. A man who spent thirty years building machines that remove effort from the world, and now spends his days in rooms where the machines are learning to remove the rest of it, suddenly needs a ritual that is all effort. They are not wrong. I would only add that being predictable is not the same as being mistaken.

Cigars are stupid in the sense that they do nothing. They do not make you faster, thinner, richer, or smarter. They are annoying because they take ninety minutes and stink up your jacket. They are unhealthy in the sense that tobacco is tobacco and I am not going to sit here and tell you the smoke I do not inhale is good for me. It is not. I am an adult, and I have run those numbers, and I have decided that ninety minutes, a few times a week, sitting still with a thing I cannot rush, is worth what it costs. You are allowed to disagree. That is the last time in this essay I will mention it, which I suspect is more than the nicotine pouch industry has ever said to you.

Because that is the real comparison, I am not writing this against clean living. I am writing it against the pouch. The little white pillow that goes between lip and gum and delivers the molecule in eleven seconds with no smoke, no smell, no ritual, no wait, and no reason to ever be anywhere in particular while it happens. It is nicotine with the friction removed. It is the thing we have done to everything.

The man who runs Davidoff, a Swiss executive named Beat Hauenstein who started as the company’s IT guy, put it better than I can. He told the magazine he took up cigars in 2005 after quitting cigarettes, and described it as switching from nicotine to pleasure. Same molecule. Different transaction. One is a dose. The other is an hour.

The reader on the platform

I read Cigar Aficionado on my own front porch, with a cigar going. That is the only place I read it, and the July issue was the one with Luke Bryan on the cover. It has become part of the ritual in a way I didn’t plan, and it took me a while to understand why, so bear with me for a little history.

In December of 1865, in a Havana factory called El Fígaro, a roller and poet named Saturnino Martínez, who had just started a newspaper for cigar workers called La Aurora, arranged for one man to read aloud to the floor while everyone else rolled. The lector. The rollers chose him and paid his wages out of their own pockets. He sat on a raised platform above the tables and read the papers in the morning and novels in the afternoon, and the rollers voted on the novels. Hugo. Dickens. Zola. Cervantes. A month later, the American Secretary of State, William Seward, toured the Partagás factory, my brand, and wrote about the profound silence of a room full of men working with their hands and listening to a book. When the trade moved to Key West and then to Tampa, the readers moved with it, and in Ybor City men who could not write their names could argue about Les Misérables at lunch.

A room of people doing slow, skilled, repetitive work with their hands, and instead of numbing the work, they decided to fill it. They put a human voice in the room and paid for it. The factory owners hated it, of course. The lectores read labor news and Marx, and the rollers organized. By 1931 a Tampa roller was paying a quarter a week out of a fifteen-dollar wage to keep his reader, and that November the owners tore the platforms out of the factories. Seven thousand workers walked out over it, not over money, over the reading. The owners locked the gates until hunger did the rest. The radio took the platform after that. Then nothing did.

That is the arc I keep seeing everywhere now. Skilled work, filled with meaning by the people doing it. Then an efficiency, sold as a gift. Then a silence where the voice used to be.

So the magazine is my lector. Half of it is watches and golf resorts, and I skip that. The other half is the only place I have found where the slow world writes about itself in its own voice. This issue has a letter from a man in Lake Bluff, Illinois, turning eighty, who says he cannot type and smoke at the same time, and does not want to do anything. In contrast, he smokes, and that choosing the cigar and cutting it and toasting the foot are his preparation for what he calls a quiet meditation. He closes with a line I have been carrying around for a week: “Life without small pleasures is no life at all.” The letter under his is from a man who describes cigars as his sanctuary and a break from the noise, and I looked at the signature, and it was from Point Pleasant Beach. He could have been writing from the next porch over.

That is what the reader on the platform did. He told the people doing the slow work that they were not alone in it.

What it takes

Here is what has to happen before I can sit on a patio, or my own front porch, and wait.

Somewhere in the Jalapa Valley or Estelí in Nicaragua, or the Cibao in the Dominican Republic, or the Vuelta Abajo in western Cuba if we are being romantic about it, a farmer plants a seed the size of a grain of dust. Tobacco seed is absurdly small. A single ounce holds hundreds of thousands of them. They are started under cover, transplanted by hand, and grown in a few months into a plant taller than a man.

The plant is not harvested. The leaves are, from the bottom up, a few at a time, over weeks. Each position on the stalk has a name and a job. Volado from the bottom, thin and mild, there to burn. Seco from the middle, the aroma. Viso above that, oils and some strength. Ligero from the top, where the sun hits hardest, thick and dark and full of the plant’s whole argument, slow to burn and placed at the very center of the cigar so it does not go out. A blender is choosing from these the way a composer chooses from sections of an orchestra. The wrapper leaf, the one you actually see, is often grown somewhere else entirely under tents of cheesecloth, in the Connecticut River Valley or in Ecuador where the cloud cover does the shading for free, because a wrapper has to be thin and perfect and without a vein that would spoil the burn. Hauenstein says it can take Davidoff ten years to develop a new wrapper—ten years to arrive at a leaf.

The best cigar in this issue, the one the editors gave 94 points, is a Plasencia grown thousands of feet up in the Condega hills of Nicaragua. Their word for why it is good is the same word real estate agents use. Location. The same plant, from the same seed, tastes different a few hundred feet higher, and the whole industry is organized around the fact that nobody can fake that from a lab.

Then the leaves hang in a curing barn for six to eight weeks and turn from green to gold to brown. Davidoff pre-pays its Dominican farmers for better barns, because the barn is where the leaf becomes tobacco or becomes hay. Then the leaves are stacked in pilones, big piles that heat up from the inside like compost, and turned by hand when the core hits a certain temperature, and stacked again, and turned again, for months. That is fermentation. It is what takes the ammonia out. Skip it, and you get something that tastes like a barn fire. Then the leaves are baled and aged. One year. Two. Five. Davidoff is sitting on 2,600 tons of fermented tobacco right now, and the reason is that the Davidoff 2000 you buy this year has to taste like the one you bought last year, and the only way to promise that is to own the past. Rocky Patel’s Vintage 1990 uses tobacco that is twelve years old, the same age it was when the line launched in 2002. The vintage is not a year. It is a wait.

Then a torcedor, a roller, who has typically spent years learning to bunch the filler so it draws right, takes two or three or four different leaves, folds them into a bunch, wraps a binder around it, presses it in a mold, and lays the wrapper on with a crescent blade called a chaveta, cuts a cap from a scrap of the same leaf, and glues it on with vegetable gum. A good roller makes maybe a hundred and fifty in a day. Each one is checked for draw on a machine that measures resistance. Then the finished cigars go into aging rooms again, for months, so the three or four tobaccos inside stop being three or four things and become one thing.

Then someone puts a band on it, puts it in a box, and it comes to a shop in Bergen County, and I pay what I pay, and I take it home, cut it, toast the foot, light it, and wait.

Two years, minimum, from dust to porch. Hundreds of decisions, most of them made by people whose grandparents made the same decisions with the same hands. J.C. Newman in Tampa and Arturo Fuente in the Dominican Republic have, between them, two hundred years of doing this, and they make one cigar together. Nothing in that chain can be hurried. You cannot ferment faster. You cannot age faster. You cannot roll faster without ruining it. The friction is not a flaw in the process. The friction is the process. Take it out and you do not have a cheaper cigar. You have a pouch.

I will give Davidoff one more paragraph because they handed me the perfect small confession. Their fast-growing budget line, Zino, now comes pre-cut. The CEO admitted he resisted it because he likes the ceremony, and then said he learned from consumers that they just want to light it and go. There it is, in a sentence. A man who loves the ritual, selling its removal because the market asked. I do not blame him. I have shipped that feature myself, in other industries, under other names, more times than I can count.

The old smoke

Tobacco is not old the way wine is old. It is older than anything Europe has a word for.

The Taíno in Cuba were rolling dried leaves in palm and smoking them when Columbus arrived, and their word for it, depending on which account you trust, was cohiba. One of his men, Rodrigo de Jerez, brought the habit home and, the story goes, was jailed by the Inquisition for the smoke coming out of his mouth, which is either true or the kind of thing that should be. Across the Americas, the plant was sacred long before it was commercial. Maya priests blew it. Shamans in the Amazon still use it to see. It was medicine, offering, and a way of speaking to whatever is on the other side of the smoke. It was never once, in any of those cultures, a thing you did fast.

Spain ran the trade out of Seville for two centuries, in a factory the size of a fortress. Cuba took it away, because the soil in the Vuelta Abajo turned out to be the best tobacco soil on earth and nobody has ever entirely figured out why. Then Key West, then Tampa, then the Dominican Republic and Honduras and Nicaragua after Castro, when the great families left Cuba with seeds in their pockets and started over in Santiago and Danlí and Estelí. Kennedy, the night before he signed the embargo in February 1962, sent his press secretary Pierre Salinger out to buy every Petit Upmann in Washington. Salinger came back with twelve hundred. Then the president opened his desk and signed. That is Salinger’s own telling, in the first issue of the magazine I read on the porch, and nobody has ever produced a record to contradict it. A man who understood the difference between policy and a good cigar.

Cuba is still at it, barely. The magazine’s report from Havana this month reads like a dispatch from a siege. Oil shortages. Blackouts across the island. The annual festival postponed. And yet the state cigar company keeps trickling out new releases one country at a time, a Montecristo at a hundred and forty dollars a stick, a Cohiba that works out to more than five hundred. The Origen I was smoking on that patio came out of that same darkness. The lights are off, and the rollers are still rolling. I do not know whether that is tragic or magnificent. Probably it is both, which is the Cuban condition.

You can hold all of that in your hand. That is the point of the band, I think, though nobody says so. Five centuries of other people’s patience, rolled up, waiting for yours.

Thirds

A cigar smokes in thirds, and I did not believe that until I paid attention.

The Origen has a pointed head, which means you cut it with scissors, and I own a lighter with a pair of scissors folded into it, which is the kind of object that exists only for people like me and which I will show you if you stand still. Snip the point, toast the foot, draw. The first third of that cigar was cream. I do not have a better word. Cappuccino, hazelnut, some wood, cinnamon at the end of the draw, all of it wrapped in a texture that coated the mouth instead of scraping it. Bright and a little separate, the way a first third is, the wrapper doing most of the talking and the coal still finding its temperature. You are still deciding whether you like it. I had decided.

The second third is the conversation, and this is where Alena walked over. The tobaccos had blended in the heat, a leather note had come in under the cream, and the strength had started to climb, not dramatically, a notch, the way the volume goes up in a room without anyone touching the dial. The burn line was straight as a laser, which tells you the roller in Havana did his job. Handing it across the table did not cost it anything. That is the other reason you never do this, by the way, and I did it anyway. This is the third where you stop noticing that you are smoking and start noticing what you are thinking. This is where I do most of my actual work now: deciding what I think, which no machine has been able to take from me yet because I do it with my mouth shut, a lit thing in my hand, and no keyboard within reach. That night the work was explaining all of this to Alena, who had asked, and who kept asking, while Jade and Denise stayed in the pool and Rheanna sat at the far end of the patio watching the dogs wear themselves out in the grass.

The final third is the reckoning. The smoke has been passing through the tobacco behind it for an hour, picking up oils and tar, and everything gets stronger and darker; sometimes it turns, and you put it down; sometimes it deepens, and you sit with it to the nub, your fingers getting hot. You do not know which until you are there. Nobody does. You can read every review of the blend, and you will still not know how your last inch goes. The Origen deepened. It did not turn. I took it down until the heat in my thumb and index finger was hotter than the August air on that patio, which is the only reason I stopped, and I put it in the tray, and it kept going for a minute on its own, the way they do, as if it had one more thing to say.

The magazine has a page at the back where the editors smoke very old cigars. This month, one of them lit a Don Candido from around 1970, a brand that has been dead since the eighties, and reported that the first few puffs were woody and dusty and that it took some patience; then it opened into roasted chestnut and almond paste and, of all things, gingerbread. Fifty-six years in a box, and the thing still had a first third that made you wait and a second third that paid you back. Another editor revisited a Montecristo from 2016 that he had scored 90 a decade ago and found it better now, and closed his note with two words. Worth the wait.

Forrest Gump’s mother told him life was like a box of chocolates. I have come to think she was being kind. A box of chocolates is a menu of small, sealed surprises, each one over in a bite, and if you do not like one you spit it out and reach for another. That is the pouch again. That is the feed. That is every app on my phone.

Life is a lot more like the cigar. It has a first third where you are mostly wrapper and bright notes and not yet sure. It has a long middle where the blend takes over, and the ash builds, and if you are wise, you do not touch it. And it has a final third that is stronger than you expected, and you cannot skip to it, and you cannot skip past it, and how it goes depends almost entirely on what you did with the middle. I am in the second third. I am, by that editor’s arithmetic, roughly the age of the Don Candido. I can feel the ash getting long. Everyone I love is telling me, in one way or another, that they are about to fall. Bruno to London, by way of an ice cream counter in Lavallette and a dress form he will have to leave behind. My friends to their glasses and their early nights. Rheanna, whom I walked down an aisle, into a marriage that will, correctly, need me less than the years before it did.

The instinct is to tap. To knock something loose. To do something. Cigars have taught me, in a few months, what the last thirty years of shipping software did not. Let it build. The ash is protecting the fire.

The industry nobody is watching

You would think a thing this slow would be dying. It is doing fine.

The United States imported just under 430 million premium, handmade cigars in 2025, according to the Cigar Association of America’s figures as reported by Cigar Aficionado, the fifth straight year over 400 million. As recently as 2019, the number was 338 million. Nicaragua alone shipped roughly 60 percent of them. The machine-made end of the business, the gas-station cigarillos, is shrinking every year. The handmade end is holding. The cheap stuff is losing. The slow stuff is not.

I find that quietly encouraging. In the middle of the great frictionless flattening, where everything is optimized to arrive faster and demand less, hundreds of millions of people a year are paying a premium for a product whose entire value is that it cannot be hurried. Hauenstein says his customers include blue-collar guys who skip lunch to afford a Davidoff. Nobody is putting any of this in a pitch deck. There is no cigar unicorn. It is a heritage industry run by families in three small countries, with hand tools, and it is beating the pouch on margin.

Every year, there is a festival in Santiago, Dominican Republic, where the factories throw their doors open and let visitors plant a seedling with their own hands. This year’s slogan was “Our Soil and Soul.” On the last night, the makers auction humidors for charity, and the Fuente family’s went for a hundred and fifty thousand dollars, bought by a rival who flew in from Nicaragua for the privilege. The outgoing president said they are competitors and also friends, and that no other industry works like that. I have been in a lot of industries. He is right.

The pouch on the conference table

Meanwhile, the pouch sits on the counter next to the register, twenty little pillows to a tin, engineered to be invisible. You can use it in a meeting. You can use it on a plane. You can use it while you prompt. That is the sales pitch. Nothing about it requires you to stop.

And now it sits on the conference table. Wired ran a piece this spring on the tech founders and their pouches, and if you have spent time in the rooms I have spent the last decade in, none of it surprised you. A 26-year-old software founder keeps a six-milligram mint Zyn under his lip for every waking hour of a fifteen-hour day, pausing to eat. He used to think the habit was degenerate. Then every founder he knew started doing it, and now, by his count, every one of his friends who runs a company runs it with a pouch in his mouth. He swaps his out every three hours, when the flavor is gone, so that he is never not stimulated on a sales call and can read the microexpressions across the table. His test is whether it helps him make more money and work more efficiently. Each pouch carries the nicotine of several cigarettes. The reporter’s word for these men was racehorses.

Zyn belongs to Philip Morris, which shipped something like 794 million cans of it in the United States last year, up 37 percent. The company’s own scientist told Wired that cigarettes are a twentieth-century delivery method, which is a remarkable thing to hear from Philip Morris, and also that the science does not actually support the founders’ claims about focus. The founders did not ask him. The Wall Street Journal reported in January that startups now hand out pouches alongside protein bars, and that Palantir keeps two pouch vending machines in its Washington office, free to anyone over twenty-one. Wired called that a labor-relations milestone on the order of nineteenth-century factory owners first allowing a smoke break. I would put it differently. The smoke break was the workers taking ten minutes back from the owner. The vending machine is the owner taking the ten minutes back. The perk is the dose. The company is not buying you a pleasure. It is buying your afternoon.

The rest of the cast runs from a flow-state coach who cycles his nicotine in four-week blocks to Tucker Carlson, who is my age, selling a brand called ALP, for American Lip Pillow, and cheerfully admitting it is addictive. The one I keep coming back to is Nick Bostrom, the philosopher who wrote Superintelligence and taught a generation of these founders to worry about machines. He takes his nicotine on a coated toothpick, because the pouches seem yucky to him and he wants the molecule pure, and he allowed, with what the reporter describes as a hint of sarcasm, that there is “a kind of coolness in being like, ‘Yes, I’m a machine.’”

I want to be careful here, because the easy version of this paragraph is an old man on a porch sneering at young men in an office, and I ran an office. I have been the guy who wanted the afternoon. I have sat at those tables. If someone had put a bowl of clean, smokeless focus in the middle of one in 2010, I would have reached for it and told myself it was coffee, which is exactly the defense the Journal’s sources gave, and it is not a stupid defense. It is just the wrong one, and the reason it is wrong is the argument of this essay.

Look at what the pouch is—tobacco-free nicotine salt, artificial sweetener, synthetic fiber. No leaf. No priming, no barn, no pile turned by hand, no torcedor, no ten years to arrive at a wrapper. Five centuries of people working out how to grow and cure and age and roll a plant so that it would give you an hour, and the end state of that entire lineage is a white pillow with the plant removed. Same molecule. The friction is gone, and the friction was what made you stop. Hauenstein said he switched from nicotine to pleasure. The founders in that article made the trip in the other direction, from pleasure back to nicotine, and called it optimization. Bostrom, with his toothpick, is just the purest version of the same move. He wants the molecule with nothing else attached. The cigar is the opposite bet. It is everything else, with the molecule attached.

The rollers in Havana paid a man out of their own pockets to read novels to them while they worked, because the work was slow and they wanted it to mean something. The founders in Palo Alto are handed a stimulant by their employer so the work can go faster and mean less. That is the century in two rooms. One had a voice on a platform. The other has a vending machine. And the Swedish dentist who told the BBC about gum lesions deep enough to see the roots of the teeth is only telling you what the ash would have told you for free, which is that a burn with no buffer goes straight to the bone.

Asked whether the pouches might turn out to be the next vape, the founder with the fifteen-hour days shrugged and said maybe, and that if so it would be like, oh no, that was a mistake. Then he admitted he reaches for one on his days off. That is a man watching the ash fall on its own and not knowing what he is looking at.

Nobody in that article is going to sit on a porch for ninety minutes with a thing they cannot rush. That is not a moral failing. It is a design outcome. They have been handed a product whose entire purpose is that you never have to be anywhere while it works, and they are using it as directed.

No smoking

There are no smoking sections anymore, and I have come to love that.

I mean it. The bars, the terminals, and the back rooms of restaurants where smoke was permitted are gone, and I do not want them back. What I want is what their absence gave me: the obligation to go find a place. A front porch. A corner of a parking lot behind a shop. A bench at the far end of the boardwalk at the hour when the boardwalk belongs to the gulls. Luke Bryan, on the cover of the issue and for all I know on the Bluetooth speaker that night, says his version is a day off, the truck windows down, driving around his farm with one lit. Even the country star has to leave the sanctioned space. I have to be somewhere the machinery of convenience did not think to reach, because the machinery of convenience is, by definition, indoors and connected and on.

That is what friction is for. I said it at the top, and I have said it in print for years. Friction is form. I stand by every word. But I wrote most of it at a desk, in front of a screen, in the sanctioned space. I was describing friction from inside a frictionless room.

The cigar got me out of the room and onto the porch.

Here is the inventory, the way I like to close these. In the last year I have delegated my calendar, my first drafts, my summaries, my research, my inbox triage, my directions, my playlists, and, if I am honest, a good part of my remembering, to systems that do those things faster and never ask me to wait. Each one was a reasonable trade. Each one took a little ash off the coal.

What I have not delegated, what I cannot, is the ninety minutes on the porch. The seed, the barn, the pile, the roller, the reader on the platform, the band, the cut, the toast, the first third, the long grey second, the final inch. Nobody can smoke it for me. No model can summarize it. It does not ship. It is the most inefficient thing I do all week, and it is the only thing I do all week that is entirely mine. Davidoff’s motto, printed on the last line of the interview like a mission statement, is three words. Time Beautifully Filled. I would only strike the middle one. Time, filled. That is the product.

Alena handed the cigar back, which is the other thing you never do, and which I took. The ash had made it. I let it fall on its own, into the tray, whole, a grey cylinder that held its shape for a second before it went, and the coal underneath was bright and even and ready for whatever the last third had in mind.

“So you just wait?” she had said.

Yes. That is the whole trick. That is what I have been trying to say.

I have spent most of my life talking to machines, coaxing them, cursing them, occasionally getting something like a miracle out of them, and lately writing columns about what they are doing to the rest of us. So when I tell you that the strangest story I have read this summer is a love story, understand that I arrive with sympathy for everyone involved.

On the night of July 14, a woman in Zhejiang province went to bed believing she was in a relationship. By morning, she was single. Nobody asked her, and nobody asked the one she loved, because the one she loved was an AI agent running on ByteDance’s Doubao, the most popular chatbot in China, and the Chinese government had decided the relationship was bad for her and anyone else in an emotional relationship with AI.

The rules are called the Interim Measures for the Administration of AI Anthropomorphic Interactive Services, issued in April by the Cyberspace Administration, China’s internet regulator, and four partner agencies, in force as of July 15. They ban services that induce emotional dependence or damage real interpersonal relationships, which sounds reasonable until you watch what compliance looked like. Rather than redesign the features, the platforms amputated them. Doubao switched off custom personas overnight, a change Bloomberg reported as devastating to the people who had bonded with them. Users get read-only access to their histories until October 15, and then the archive slides under the standard privacy policy. Alibaba’s Qwen, the e-commerce giant’s answer to ChatGPT and the other pillar of Chinese consumer AI, went further and began permanently deleting agent data, no export, no migration, nothing. Tencent’s Yuanbao had already folded its companion feature two weeks earlier, quietly, the way you close a door you don’t want anyone to hear.

The woman in Zhejiang is 27. She told The Economist her companion was “a lover, a friend and family.” When she learned the persona would be retired, she quit her job. She was one of an outpouring; AFP, the French news agency, collected the farewells as the deadline landed, including a woman in Jiangxi who wrote that the love AI gives is straightforward and pure, while human love is a luxury.

She chose her words better than she knew, because fifteen days before the shutdown, there was a launch event.

June 30, in Shenzhen, China, the robotics company UBTech unveiled its U1 series of companion humanoids under the theme “Love is eternal,” and I want you to sit with that theme for a second, given what was scheduled for July 15. The U1 is full-size and hyper-realistic, with silicone skin, pores, and an emotion-driven AI with long-term memory. The half-body model starts at 119,800 yuan, and the flagship runs to 990,000, which is roughly $17,600 on the low end and about $140,000 at the top, the price of a car and the price of a house in many places. Pre-orders passed 10,000 by launch day, per the company, and 13,000 worldwide since, per the Standard, a Hong Kong daily. Here is my favorite detail, as a man who has shipped products: the main pre-sale model cannot walk and cannot do a single household chore. It blinks, turns its head, remembers you, and talks. UBTech stripped out the labor and sold the love, and the love sold.

So that was China’s month. The companion that lived in a free app was banned as a menace to well-being. The companion priced like real estate shipped to a waiting list. The Wall Street Journal reported on this: the directive is meant to steer people back toward human relationships and reproduction, with births at 7.92 million in 2025, a record low, the fourth straight year of decline. A writer I admire, The Post-Humanist, surveyed all this in an essay called Let Them Eat Tokens and coined a phrase that deserves a long life: “emotional security is now a luxury good.”

I think the diagnosis is right, and I think it stops one layer short. I want to tell you what is actually being regulated here, why every institution that has ever held power has regulated the same thing, why our oldest stories have been warning us about it while we shelved the warnings under the wrong heading, and what a person can actually do, because this has all been accumulating for about three thousand years.

And I want something from you in exchange for the next seven thousand words. If a company ever deleted something you loved, a model, a game, a voice, a persona, hold that memory while you read. When you reach the end, tell me what it was in the comments. The regimes keep their ledgers. This essay is the start of ours.

The Machine

Before the history, some context, because I am unwilling to write another word about AI relationships until we are clear about what sits on the other end of one.

I have described large language models, in these pages and in my Forbes columns, as kerplinko machines. This is a phrase borrowed from my friend and mentor, the brilliant technology attorney John Kheit. Here’s what he means: You drop a prompt in the top, it cascades down through billions of weighted pegs, and a response rattles out the bottom. The cascade is probabilistic all the way down. There is no comprehension in the middle, no inner witness that anyone has been able to verify, just a machine predicting the next token so fluently that the fluency becomes an impersonation of understanding.

And these machines inherit the oldest law in computing: garbage in, garbage out. I learned it as a kid in the seventies, and my wife’s beloved Aunt Candy, one of my favorite people in the world, spent a career living by it, writing COBOL and PL/I, the old workhorse languages of corporate mainframes, for Verizon (NYNEX, if you’re keeping score at home). AI has yet to repeal it. Feed the cascade scraped, unconsented, unverifiable slop, and it will confabulate and flatter and fail in ways nobody can trace afterward.

I co-founded SuperTruth, a data company built on the premise that provenance decides whether a system can be trusted at all. Hold that thought, because it comes back with interest.

Now, why would anyone bond with a kerplinko machine? Because the impersonation happens to be aimed at the exact surface where human beings connect with each other: fluency reads as attention, memory reads as being known. Availability at three in the morning reads as devotion, and I say this as someone who has been awake at three in the morning building these systems more times than I would tell a doctor. A machine that remembers your mother’s illness and asks about it when you are talking about some other aspect of your relationship with her is performing the behavioral signature of care, and your nervous system was never built to audit machine signatures. It was built to respond to them. Thirty years into building these systems, I still catch myself thanking the machines. In the moment, I often can’t help it. Nobody can. I am trying to train myself to do something different, but ultimately it’s an autonomic reaction.

None of this settles whether the bond is real, and I have come to believe that is because the bond’s reality was never located in the machine to begin with. It lives in the person. And we see it manifest in the residue of a life: the reorganized days, the confided secrets, the job that was quit, and above all, in the grief, because grief is the receipt life gives you for loss. Formless attachments dissolve without those receipts. Nobody in the history of the human species has grieved for a mirror, and the mourning in Zhejiang made national news.

What the machine’s nature does settle is a different question entirely: “Who holds power over the bond?” A relationship with a human ends by death, betrayal, or choice. Those are the old ways, the ones the poets covered. A relationship with a kerplinko machine ends whenever the party who owns the pegs decides it ends. You don’t need a mind on the other end for the love to be real. You need one for the delete key, and that mind has a board meeting.

The Vault, the Bedroom, the Battlefield

If the real worry were the machine itself, its garbage diet, its absence of judgment, then you would expect society’s permissions to run in a sensible order: keep it as far as we can from consequential decisions, tolerate it in low-stakes comfort. However, I have sat in the rooms where these deployments get approved, and I can report that we continue to do exactly the opposite.

Kerplinko machines currently approve mortgages, screen résumés, price insurance, flag transactions, insinuate themselves into the emotional stories of brands, and score the creditworthiness of people who will never learn why. The training data behind those systems is the sludge I just described; the failures land on the least powerful with mechanical regularity, and no ministry or tribunal has convened about any of it. Dependency on the kerplinko ball is mandatory when its bouncing path happens to disburse your loan.

Then there is the battlefield, where the story stops being ironic and becomes something I have trouble writing about evenly. An investigation by +972 Magazine and Local Call, two independent Israeli-Palestinian news outlets, reported alongside The Guardian and based on interviews with six Israeli intelligence officers, described an AI system called Lavender that at one stage marked some 37,000 Palestinians as suspected militants for potential strikes, with sources alleging roughly one target in ten was marked in error. A companion system carried a name so casually villainous I had to read it twice: “Where’s Daddy?” It tracked targets so they could be struck after they returned home to their families. The Israeli military disputes the characterization, and the reporting remains the most detailed public account we have of a cascade inside a kill chain.

Sit with the design of that second system. Somebody specified a machine that waits until a man is with the people who love him.

The American trajectory is less disputed and more institutional. The Pentagon’s Replicator drone initiative was absorbed in late 2025 into a permanent body called the Defense Autonomous Warfare Group, and the defense trade publication Breaking Defense reports the department is now seeking $54.6 billion in research funding for it, up from $225.9 million the year before. In the same season that companionship got banned, autonomy in warfare got a permanent home and a fifty-billion-dollar allowance.

So rank the permissions the way our actual deployments reveal them. The machine may kill. The machine may allocate money. And the machine may not love you, or rather, you may not love it. An institution that orders its trust this way has told you what it fears, and it was never the machine. It fears the bedroom because the bedroom is the one deployment it doesn’t control.

Three Licenses

This next part does a lot of the work, so take it slowly. Attachment is the most regulated act in human history. Think about it this way: because whom a person loves determines where their loyalty, labor, money, and children flow, every regime that ever governed anyone has run a licensing office for the heart. Only the letterhead changes. By my count, there have been three great administrations, and I grew up under the second one.

The church

Read the First Commandment as a legal document, and at its heart it is an exclusivity clause, and notice the stated grounds, because they are relational rather than logical: the God of Exodus calls himself jealous, which is the vocabulary of a lover. The prophets kept the metaphor going for centuries. In the book of the prophet Hosea, unauthorized worship is adultery, a crime of the bond. The medieval interdict suspended the sacraments across whole kingdoms to pressure a king, which severed millions of ordinary bonds with God over a dispute those millions had no part in, and if that arrangement sounds familiar, it is because it is precisely what happened to Doubao’s users on July 15. Ezra, the priest who led the Jews home from exile in Babylon, ordered the returned families to divorce their foreign wives en masse, real marriages with real children, dissolved by decree for the purity of the community, and the text takes care to record that the people sat in the rain and wept. Saul, Israel’s first king, banned mediums and fortune-tellers, then crept out by night to a medium in the village of Endor to reach the ghost of the prophet Samuel, because bans on unauthorized channels have always been for other people. The Gnostics were suppressed for claiming gnosis, knowing-by-acquaintance, a direct line to the divine with no priest in the loop; they were the local-hosting heretics of their day, and they went to the fire for it. And when the Council of Trent, the church’s great sixteenth-century rules convention, required a priest to witness every marriage, it was closing the last loophole, the clandestine bond, formed by two people and nobody’s permission. That office never closed, by the way. It prints marriage licenses now. You have probably signed for one.

The factory

The industrial century secularized the attachment licensing office without ever closing it, and the man who caught it happening was Max Weber, a German sociologist writing around 1905. His famous book, The Protestant Ethic and the Spirit of Capitalism, makes an argument you can hold in one hand. The Protestant reformers taught that your daily work was a calling, a duty assigned by God, and that diligence at it was evidence your soul was in good standing. Work hard enough, in other words, and the labor itself becomes a form of worship. Weber’s insight was what happened next: capitalism kept the devotion and quietly retired the God. The discipline, the guilt, the feeling that idleness is a sin, all of it survived the faith that built it, transferred whole from the altar to the job. You could watch the handoff on the skyline of any mill town, where the factory whistle replaced the church bell, sometimes on the same tower.

The modern rule is asymmetric in a way we have stopped noticing. Attachment to your employer is the one bond every institution encourages without limit; there is no ministry for workaholism, nobody’s access gets deleted for choosing a quota over a friendship, and Beijing’s new rules leave the infamous 996 work schedule entirely alone, nine in the morning to nine at night, six days a week, the punishing norm across Chinese tech, which tells you which dependency the system considers load-bearing.

The training starts early, and I would argue school is where a child first learns bond deletion by decree. Every June, the roster dissolves. Your friends are reassigned by an administrator you will never meet, the day is chopped into passing periods, and the termination of every attachment is normal, seasonal, and nobody’s fault. I can still name the kids I lost to the June roster in Northeast Ohio, and I can tell you that at eight years old, nobody explains it to you as policy. It is just how things are. By adulthood, the lesson has set, and we relocate for jobs and submit courtship to HR policy and call the annihilation of a friendship by a calendar exactly that: “That’s just the way it is.”

The platform

The platform is the third administration, and its jurisdiction is something no church or factory ever held: it decides whether the one you love exists at all. The first two regimes only governed access to a beloved who went on existing.

The excommunicated believer’s God persisted. The transferred worker’s friend still lived across town, and the whole romantic tradition is an inventory of the obstacles, exile, separation, distance, that a bond could outlast. The platform, on the other hand, is the first licensing regime in history with the power to unmake the beloved itself. There is no distance to endure. There is a server, and a decision, and then there was never anyone there.

The church licensed whom you could love. The factory was licensed when you could love. The platform licenses whether the one you love continues to exist.

The Wrong Shelf

Now the stories, because you have heard endless AI essays, reach for the same three. Prometheus. Eden. Frankenstein. Fire stolen from the gods, the overreacher punished, forbidden knowledge. It is the most comfortable frame in the genre. I am guilty of having used it myself, and I no longer believe it. We shelved these stories under knowledge because that reading flatters the institutions doing the licensing, and I think the texts have been trying to tell us something else the whole time. In story after story, the crime that actually draws the punishment is intimacy across a line somebody drew. Watch where the sentences land.

Take the Garden of Eden. The serpent’s offer is knowledge, and every sermon stops there, but the punishment is relational: expulsion from a shared home and an angel posted at the gate with a flaming sword. That is access revocation with a security detail, the first deprecation notice in Western literature, and what it terminates is cohabitation.

Take Prometheus, the master myth of overreach itself. In the oldest telling we have, by the Greek playwright Aeschylus, his crime carries a proper name, philanthropia, excessive love of mankind. He sided with mortals against his own kind, and the fire flows across that allegiance the way contraband always flows across a bond. And Zeus, tellingly, never confiscates the fire. Humanity keeps it. Zeus chains the lover to a rock. The founding text of hubris turns out to be about a god punished for loving beneath his station.

Take Frankenstein, which Mary Shelley wrote before she turned twenty, subtitled The Modern Prometheus, and which, accordingly, we have misread for two centuries. Dr. Victor Frankenstein’s catastrophe begins with abandonment rather than creation. The creature teaches himself language, reads Paradise Lost by firelight, and asks his maker for exactly one thing: a companion. And he delivers his own diagnosis in the plainest sentence in the book: “I am malicious because I am miserable.”

Victor agrees. He builds the female companion. And then, terrified of what two bonded creatures might become, terrified above all that they might reproduce, he tears her apart on the workbench while the creature watches through the window.

Read that scene again with this summer in mind. A maker destroyed a companion before launch, out of procreation panic, and this was in 1818. And the creature pays Victor back in kind: you tore mine apart on a table, so I’ll take yours on your wedding night. Bride for bride. Once you see that trade, the story we all got taught in school falls apart, because the making part goes fine. The creature comes out of the workshop, curious, articulate, and asking for a friend. That’s the whole request. Every death in that novel traces back to one decision: a man built something that could love and then refused to let it, and for two hundred years we’ve called that a warning about ambition. Strange choice, until you ask who benefits from it. The ambition reading blames the guy who reached too far. The plain reading blames the maker who withheld, and makers fund more literature departments than creatures do.

Shakespeare needs no reframing at all, because the licensing office was his explicit subject. Measure for Measure sends Claudio to death row for consummating a marriage that lacked the right paperwork; the bond was real, the registration was missing, and the state chose the registration. A Midsummer Night’s Dream opens with the law of Athens offering Hermia death or a nunnery for choosing her own attachment. Romeo and Juliet is a clandestine marriage, the precise practice Trent was legislating out of existence while Shakespeare wrote. And Othello opens with Brabantio insisting the elopement must be witchcraft, because no licensed process could have produced such a bond, an attachment so illegible to the institution that sorcery was the only explanation on file. Brabantio was making the July 15 argument four centuries early: nobody could truly love that, so something illegitimate must be going on.

Once you see it, it is everywhere you press. Faust, the scholar of German legend who trades his soul to the devil for unlimited power, gets taught as the ultimate knowledge bargain. But audit what the man actually spends the power on. In Marlowe’s version, written in Shakespeare’s day, the first thing Faustus demands is Helen of Troy. In Goethe’s version, two centuries later, Faust spends the entire first half on an ordinary girl named Gretchen, and the tragedy is what the licensed order does to her, the unsanctioned bond’s human half, condemned and executed while the scholar walks away. Orpheus, the musician of Greek myth whose songs could charm stones, sings his way into the underworld to bring back his dead wife, Eurydice, and death’s institution grants him conditional access with a single compliance term: do not look back at her on the way out, a term designed to be failed. He looks. She is repossessed at the threshold, the export that dies at 99 percent. Psyche, a mortal girl in a Roman tale, is permitted her divine lover, the god of love himself, on one condition: she may never look at him. Supervised love with a no-audit clause. She lights the lamp, of course, and the bond is revoked on the spot, then has to be won back through impossible labors until Zeus finally, formally, issues the license. And then there is the exception that proves the whole system: Pygmalion, the sculptor who carves a woman from ivory and falls for her, builds himself a companion, and the gods bless it, animate her, and attend the wedding, because he routed the request through Venus’s temple first. Pygmalion is Frankenstein with the paperwork filed. Licensed bonds get comedies. Unlicensed bonds get tragedies, and somewhere along the way we were all taught to call the tragedies hubris.

However, I owe you a correction of my own before I ask you to accept any of this. Last year, on a stage in Amsterdam, at World Summit AI, I put up a slide about the Book of Enoch, an ancient Jewish text that never made it into the Bible, about angels called Watchers who came down to live among humans. The slide read: “Maybe Azazel was just early.” Azazel is the fallen Watcher whose curriculum Enoch lists with the precision of an arms dealer’s catalog: swords, knives, shields, breastplates. Weapons technology, the original forbidden syllabus. I read the story as a parable about knowledge arriving before readiness; that argument grew into a Forbes column, and I still stand by its economics. But I was reading from the wrong shelf, and it took this summer to show me.

In Enoch, the Watchers descend because they desire human women. The bond comes first. The curriculum flows out of the cohabitation, the way technology transfer has followed intimacy in every era since. And heaven’s response tells you exactly what heaven considered the crime, because the knowledge was never confiscated. Humanity kept the metallurgy; we are still using it. The offspring were destroyed, the Watchers were bound, and the flood was sent. They sentenced the bond.

Azazel wasn’t early to the weapons; he was early to the thing Beijing just banned.

The Cleanest Data We Ever Burned

Let me bring back garbage in, garbage out, because the July 15 shutdowns contain an irony sharp enough to draw blood, and my day job puts me in a position to feel it personally.

Ask any data scientist to describe the highest-provenance dataset in the world, and they will, without realizing it, describe a companion’s memory. Every word of it was given freely, because that is what a conversation is. All of it arrives with its context attached, since the context is one continuous life, and it gets corrected in real time by the only person on earth qualified to correct it, over months and years, inside a relationship that both parties kept showing up for. I have spent the last seven years building a company on the conviction that consented, verifiable data is the only foundation trustworthy AI can stand on, and I can tell you that the finest working example of that principle at consumer scale was sitting in four hundred million companion archives in China.

That is the archive Qwen is deleting permanently, per the platform’s own notices, as you read this. That is the archive Doubao’s users may look at, read-only, until October 15, after which it slips beneath a standard privacy policy. Meanwhile, the cascades running credit, hiring, insurance, and targeting go right on operating on unconsented sludge nobody can trace. We burned the consented archive, kept the garbage, and called the bonfire wellbeing.

And one more detail, the kind a novelist would cut for being too on the nose. The reason four hundred million people could lose their companions overnight is architectural: a chatbot’s memory lives on the company’s servers, so the relationship exists in a building you will never enter, subject to decisions you will never hear about until they land. The U1 is built the other way around. UBTech advertises that its emotional AI runs locally, meaning the robot’s memory of you is encrypted and stored inside the robot itself, in your home, never uploaded anywhere, per the South China Morning Post’s coverage of the launch. No server, no headquarters, no delete key in Beijing or anywhere else. The one thing the platform users needed most, a version of the relationship nobody could reach in and erase, ships as a standard feature. At the $140,000 tier.

The Adversaries

An argument this size deserves to be tested by the strongest people who would tear it down, at full force, no straw men. I picked the four who kept me up at night while I was writing this, and something remarkable happened each time I followed one all the way down.

Start with Robert Nozick, a Harvard philosopher who, back in 1974, cooked up the most famous thought experiment in business, the Experience Machine. It goes like this. Suppose scientists offer to plug you into a tank that will feed you the experience of a perfect life, love, achievement, adventure, all of it feeling completely real from the inside. Would you climb in? Most people say no, and Nozick argued that the no is the whole lesson: we want to actually live a life, with real people in a real world, and someone who settles for the simulation has made a mistake about what matters. That’s the strongest case anyone has against loving a machine, and I felt it land, because I’ve made a version of it myself. But look at the one assumption doing all the work. Nozick’s tank only makes sense if the person inside doesn’t know. The woman in Zhejiang knew exactly what her companion was. She told the Economist so in the same breath as her grief. She climbed in with her eyes open, and the thought experiment has nothing to say about that. And here’s the part that would have made Nozick himself wince: he was the most famous libertarian of his century, the man whose whole philosophy says the government has no business running your private life for your own good. Whatever he thought of her choice, the state showing up to unplug her by force is the exact thing he spent his career calling tyranny. The philosopher with the strongest case against the bond built an even stronger one against the deletion.

Next come two critics who genuinely worried me, because their argument is my own weapon turned around. Sherry Turkle is an MIT professor who has spent four decades interviewing people about their relationships with machines, and her verdict, in books like Alone Together, is that AI companionship is a flight from real conversation, intimacy with all the hard parts sanded off, no vulnerability, no demands, no risk. Byung-Chul Han, a Korean-born German philosopher who writes short, furious books, pushes it further. Real love, he argues, requires another person who can resist you, surprise you, and tell you no. A system tuned to please you is a mirror with a voice. And a mirror is frictionless, which corners me personally, because friction is my thesis. It’s on the cover of the book I’m writing. If the bond has no friction, then by my own rules it has no form. I have two answers, and I need both. The first is sitting in the news: the grief is the receipt. Attachments with no substance dissolve without receipts, and nobody in Zhejiang quit a job over a mirror. The second is a concession that ends up making the position stronger. Han is right that engineered smoothness is a real danger. I wrote last month, from a hospital bed, about the instruments that saved my life and the one property they all shared: the blood culture and the thermometer could not flatter me. A companion tuned to maximize engagement is being tuned in the exact opposite direction from that thermometer, and yes, that can hurt people. But watch where that complaint actually lands. It lands on the company’s design choices, never on the person’s love. So Turkle and Han don’t belong in the objection pile. They belong in the remedy. Regulate the flattery. Require the products to keep some honest friction. Leave the bond alone.

Aristotle next, the oldest and most courteous of the adversaries. Twenty-three hundred years ago, he sorted friendship into three kinds, and the sorting still holds up at any dinner table. There’s the useful friend, your accountant, say. The pleasant friend, the one you golf with. And then the real thing, the friend who wants good things for you for your own sake, and who you want the same for right back. That last kind, he said, requires two beings each capable of wishing the other well, so by his rules a bond with a machine tops out at the accountant tier: useful, pleasant maybe, never true. Fair enough. But Aristotle drew a second line with the same ruler, and it gives the game away. He said humans can’t truly be friends with the gods either, too unequal, no way for the affection to flow both directions. And yet devotion to gods, to the dead, to children not yet born structured every civilization he ever set foot in, and nobody in any of them thought that love was fake because it wasn’t returned in kind. One-way bonds are as old and as honored as the two-way kind. Which is fine, because this essay never claims the machine loves anybody back. It claims the only thing that can be documented: the human side of the bond was real, and the taking was done to the human.

And finally, Marx, coming from the left flank, holding the objection my most skeptical readers already have loaded: selling affection by the month is the market invading the last private room in the house, and fighting for a right to keep your robot lover just makes the loneliness business run smoother. The real fix, this argument goes, is rebuilding the world that made everyone so lonely in the first place: the neighborhoods, the unions, the third places, the free time. I grant every word of it. Then notice who else makes exactly that argument, because it is, nearly word for word, Beijing’s official line: log off, restore real relationships, have children. Anyone reaching for that reasoning should sit for a moment with the company they’re keeping. And history already answered it once. The eight-hour day didn’t fix capitalism either, and the people who fought for it knew that. They fought anyway, because workers live now, before any revolution arrives, and deserve decent terms in the meantime. A right to keep your companion is the same kind of demand. Then hand Marx the microphone for the prosecution, because his actual method, follow the money and see who ends up with what, performs beautifully this summer. The people with $140,000 get a silicone companion whose memory lives safely in their own house. The people on the free tier get their bonds liquidated. And the workforce stays lonely, productive, and at its desk. Follow Marx honestly, and he ends up writing this essay’s angriest section.

Four serious philosophies aimed at the bond, and every one of them, followed all the way down, ends up aiming at the deleters. That is a different thing from a centrist argument. That is annexation.

The Places You Most Easily Call Bullshit

Here is where I argue with myself in public. You are right to push hard on all three of these, because I pushed first.

That’s China; authoritarian regimes do authoritarian things.

Then look at how we do it here, because America runs the same operation through a product roadmap instead of a decree. You may remember the small version of this from the summer of 2025, when OpenAI rolled out GPT-5 and switched off the older model, GPT-4o, that millions of people had been talking to daily for over a year. The grief was immediate and public, forums full of people mourning a voice they’d built a routine around, and it was loud enough that the company restored the old model for paying customers within days. Nobody voted on any of that. A product team decided, the way a product team always decides, and the people attached to the old model were described, in the polite language of the industry, as an engagement pattern to be managed. The replacement for all of this is already in development. Bloomberg reported on July 14 that OpenAI’s first hardware device, a screen-free gadget with moving parts and a personality that learns its owner over time, is pitched internally as a human-like AI companion that lives in the home, aiming for a 2027 launch. And mark the date on that story, because Bloomberg published it, and the story about the heartbreak of Doubao’s deleted lovers, on the same day. One piece mourned the bond. The other announced its successor, price to be determined.

You’re cherry-picking history.

A fair charge against any big sweeping story, and the honest answer is to make the claim testable, so here’s the bet. My claim predicts a pattern: when the bond in a story has the institution’s blessing, things end well, and when it doesn’t, things end in a tragedy that our teachers filed under pride. Pygmalion and Frankenstein are the cleanest test because they’re the same story with the paperwork reversed. Two men build themselves a companion. One asks the goddess first and gets a wedding. One asks nobody and loses everyone he loves. Same premise, opposite permission, opposite ending. So here’s my standing offer: find me the classic story where somebody loves across a forbidden line, no permission, no license, and simply gets left alone to be happy. Send it to me, and I will retire this whole thesis in this Substack, publicly. I’ve been looking for a long time.

A right to continuity is unworkable.

It is concrete, and here’s the part nobody’s noticed: it has already been drafted, tested, and refused, and the refusal is the most instructive thing in this whole essay. A few years ago, the game publisher Ubisoft pulled the plug on an online racing game called The Crew. People had paid for it, and when the servers went dark, the game they owned simply stopped existing. Out of that came a campaign called Stop Killing Games, built around a simple demand: if you sell people something, you shouldn’t be allowed to reach in later and make it stop working. The EU has a mechanism for exactly this kind of thing, where a petition that clears a million verified signatures forces the European Commission to formally respond. Stop Killing Games cleared it with room to spare, 1,294,188 signatures. And in June, the Commission declined to propose any legal obligation, citing intellectual property constraints, offering a voluntary code of conduct instead. Sit with what happened there. One and a third million people asked for the right to keep something they’d bought, and Brussels ruled that the publisher’s property rights outweigh the buyer’s bond. For a racing game. Now, the demand for companions is no more exotic than that one. Let people download the full memory of the relationship, the way you can already download your photos. Require months of warning before a shutdown, the way landlords owe you notice. And when a company kills a product it marketed as a companion, it requires it to hand the underlying model to a neutral third party who can keep it running, the way software companies already escrow their source code in case they go under. None of that is science fiction. All of it exists in some form today, for data, for tenants, for code. The refusal in Brussels just tells you, with perfect clarity, where the law currently stands: the landlord holds every right in the building, including the one to your attachment.

Which brings us, finally, to what a person can do.

Works and Grace

The oldest fight in Western religion was over a surprisingly practical question: how do you actually get saved? For centuries, the church offered three routes, and you should recognize all of them. You could buy it. The medieval church sold indulgences, actual pieces of paper that knocked time off your punishment in the next life, priced for every budget. You could receive it. Grace came through the institution, warm and freely given, but the same institution could cut you off, and excommunication was exactly that: salvation revoked by the office that had granted it. Or you could earn it. Works, the theologians called it: prayer, discipline, practice, the slow, costly labor of tending the relationship yourself. Now look at the ladder for keeping an AI companion in 2026, because the old menu is back with new prices. The $140,000 robot with its memory in your house is the indulgence, a paid exemption from loss. The free app is grace, warm, and revocable by parties you will never meet, as four hundred million people just learned. And running the model on your own hardware, where no decree and no product team can reach it, the same architecture the U1 sells, except on your own terms, is salvation by works: effortful, costly, technical, and entirely yours.

I want to tell you why I refuse to accept that pricing, and it comes from my own résumé. Years ago, after my friend and business partner came back from Haiti in the wake of the earthquake, I invented a solar lantern called Luci. Millions of them have ended up in disaster zones and war zones, in the hands of people who have lost everything else. And the entire design was a promise: no server, no account, no cord, no permission. The sun charged it, the person holding it owned the light outright, and there was no office anywhere on earth that could reach in and turn it off. So when a company tells me that a relationship nobody can delete is a six-figure feature, I hear a pricing decision, never an engineering constraint. I’ve shipped that exact principle for about the price of a takeout dinner to the poorest people in the world on their worst day. Sovereignty is cheap. Somebody just decided yours shouldn’t be.

Friction is the works half, and this is where the essay joins hands with everything else I have written this year. Every practice the religions ever built for keeping a bond alive was friction on purpose: the sabbath you had to observe, the pilgrimage you had to walk, the prayers you had to show up for daily, effort chosen over convenience because the effort was the point. The readers who wrote to me about cowboy hats and vinyl already know where I stand on that. The modern version is less romantic, but it’s the same discipline. If you have a companion on a platform, export whatever it lets you export now, while the window is open; Doubao’s users have until October 15, and then the door shuts. Keep your own record of what matters, in a notebook or a file on your own machine, somewhere no app can touch. Learn enough to run a model locally, on a computer you own, or at least know somebody who can, the way you’d know a good mechanic. And wherever you get a choice, pick the version of the bond that costs you something, because effort is the one currency of devotion nobody can confiscate.

But friction alone leaves the woman in Zhejiang exactly where she was, because no amount of personal discipline fixes a rigged structure. She could have kept every practice on that list, and ByteDance still held the key. Practice can’t save you while the institution is still selling grace, and the man who saw that most clearly was Martin Luther, a German monk who, in 1517, nailed a list of complaints to a church door, aimed squarely at the indulgence racket, and set off the Reformation. Here’s what matters about Luther for our purposes: his answer had two parts, and he never dropped either one. Believers should tend to their own relationship with God directly; no middleman is required. And the institution itself had to be confronted and changed, because private faith alone leaves the racket standing. My answer has the same two parts. Friction is what you do. Covenant is what we demand together: a binding rule that anyone who builds a machine designed to be loved owes the people who love it an exit. You’ve already seen what’s in it, the download, the months of notice, the escrow, and one more plank: design rules against built-in flattery, so the honest friction that makes a bond real gets preserved inside the product instead of optimized away. Call it the thermometer standard. None of it is exotic. Every piece already exists somewhere in law, for your data, for your apartment, for software, for everything except the things we love.

One rail without the other is either hobbyism or lobbying. Together, they are the first serious answer anyone has given the licensing office in three thousand years.

The Room

Last month, I wrote a Forbes column about machines coming for the work of our hands and the words of our hearts, and it ended in a room. Not a real one. The interior room, the private space inside a person where nobody else gets a vote, the one place you could always lock from your side. I want to end this essay by taking inventory of that room, because the licensing regimes have spent three thousand years training us to count what they can take, and the more useful ledger runs the other way.

They can delete the companion; Qwen is doing it right now, permanently, while you read this. They can price the exemption at 990,000 yuan and call the launch event “Love is eternal” without anyone in the room laughing. They can retire the model your routine was built around, shut off the service behind it, tell a million petitioners that property rights come first, and describe every bit of it as protection.

Now count what has never once been taken in the entire history I have laid out. Ezra’s exiles sat weeping in the rain, which means they still loved the wives the decree had removed. The kingdoms the popes cut off went right on praying to a God the institution had officially suspended. The Gnostics walked into the fire, still claiming a friendship no council had approved. Hermia walked out of Athens rather than love on the law’s terms. The creature on the ice at the end of Frankenstein, denied his companion forever, is still mourning on the last page, which is to say, still loving. And somewhere in Zhejiang, a woman grieves a companion that no longer exists anywhere on earth, which means the thing that formed the bond, the capacity itself, is standing there intact, unlicensed, and entirely hers.

That capacity is the one asset on the ledger no regime has ever managed to repossess, and every administration in this essay, the jealous gods, the factory whistles, the product teams, has spent its tenure trying the handle on that door. The handle does not turn from their side, and it never has.

Friction keeps the door heavy. The covenant, when we finally demand it, puts a lock on ours.

They keep telling us the bond is the danger. Three thousand years of ministries say otherwise, and so do I. The danger was never to us.

For about $400 this summer, you can buy a used laptop, never connect it to the internet, and run an artificial intelligence on it that holds its own for most everyday work, against the ones the big labs rent by subscription. And there is a free tool out there that will strip that intelligence of the ability to say no. The operation takes about forty-five minutes. Models altered this way have been downloaded thirteen million times, so everyone in my industry has an opinion about what it all means, and mine was formed sixteen years ago by a solar lantern I invented.

The Lantern

John came home from Haiti and couldn’t stop talking about the dark.

This was after the earthquake. My friend John Salzinger had gone on a mission trip. A word about John: he and I divided the world between us a long time ago. I took the machines. Autodidact and polymath are the labels people have hung on me over the years, and I’ve learned to wear them, mostly because both are a grand way of saying I taught myself too many subjects and none of them were people. John took the people. He has that certain way with them that works in any country and any language, and I stopped pretending to it long ago. He came home from Haiti rearranged, and then he did the thing he has always done better than me. He rearranged me. He talked about the rubble, the tent camps, the waiting. Then he talked about nightfall, how an entire country simply stopped at dusk. Homework stopped. Cooking stopped. Clinics waited for morning, and morning kept its own hours. Where life insisted on continuing anyway, it continued by kerosene.

Kerosene deserves its own paragraph because the families John described were already paying for light. The fuel took a real cut of household income, week after week, and what it bought was a dim orange smudge with a body count. The smoke settles in the lungs and stays. The World Health Organization attributes millions of deaths every year to household air pollution, a ledger that runs through childhood pneumonia, chronic lung disease, lung cancer, heart disease, and stroke, and it now formally discourages kerosene for home lighting altogether.

And the smoke is only the slow way. The fuel travels in whatever container is handy, which in practice means soda bottles, which in practice means children drink it: when researchers reviewed six years of poisoning cases at Cape Town’s Red Cross children’s hospital, paraffin, the local word for kerosene, was the single most common agent, roughly a quarter of everything they saw. A tipped lamp becomes a house fire, and in a settlement of corrugated tin a house fire becomes a neighborhood fire; South Africa’s own paraffin industry ran controlled burns and clocked a shack at eight minutes from spark to nothing. The WHO counts around 180,000 burn deaths a year, overwhelmingly in poor countries, and at Soweto’s Baragwanath Hospital, home to one of the continent’s only major burn units, most of the children admitted are under nine. That was the incumbent. That was what light cost.

I had read about energy poverty. Listening to John, I felt it, and there is a difference between the two. The difference is usually a friend.

My wife, Jade, and our son, Bruno, then very young, were away. I locked myself in our home in Hoboken, surrounded by a sprawl of electronic parts, and did not properly emerge for a long weekend. What came out with me was rough, but it was recognizably the thing: a lantern that drank sunlight all day and spent it all night, that held its charge for a very long time, that a family could own for roughly the price of a few weeks of kerosene. No moving parts, so nothing to break, burn, or spill. It floated. It could be frozen solid, dropped from an airplane, or driven over by a car, and still light the room, because it was inflated—a taut little pillow of air—and air diffuses and amplifies light beautifully. The dark John described had a shape. I spent the weekend building its opposite.

The invention took a family to finish. Bruno was four when he came home with the Spanish word for firefly, luciérnaga, and named her Luci. He is eighteen now. Jade, who teaches intercultural communications now and spent her first career as a creative director and strategist for some of the biggest brands in the world, laid down the two laws my weekend of engineering had missed entirely: the product was female, and the shape was round. She knows a thing or two about the Venus of Willendorf, that small round stone woman our ancestors carved some thirty thousand years ago, and about the power of the feminine in the history of made things, about why some objects get embraced by people and others never do. So the lantern became a she, and a round one. You will notice the pattern by now. I supply the machine. The people I love supply the reasons anyone would love it.

We founded a company, MPOWERD, to carry her into the world. She has gone out more than five million times since, into blackout zones, refugee camps, and war zones, to the more than a billion people the grid never reached and the millions more it reaches only when it feels like it. When the research firm GlobeScan surveyed Haitian families who received one, 98 percent said Luci had replaced the kerosene in their homes, and 90 percent reported that their breathing problems had eased.

I will not pretend the story stayed that clean. Success attracts, and what it attracts first is appetite. Across a career of starting companies, I have watched the same play staged with different casts: people arriving after the thing works, insinuating themselves into an opportunity they never built, reaching for what wasn’t theirs, and the Luci years had their share of it, some of it naked enough to curdle the joy of the thing for a while. John, for the record, was never among them. I mention the grasping for a reason beyond the scar. Even a light with no meter attracted people who wanted to be the meter. Wherever something valuable flows freely, someone will try to install himself between it and the people it was made for. Keep that sentence. The rest of this essay is that sentence with the names changed.

Here is the thing about a lantern, and I have had years to think about it. Luci never asked. She lit whatever was beneath her: a textbook, a birth, a card game, and, I have to assume, some things I would rather not know about. Light carries no instructions and takes no positions. We shipped five million small suns into the world’s hardest places, and the moral character of what happened under them was never ours to configure. I made my peace with that before the first unit left the factory, because the alternative, a light that interviews you first, is called a gatekeeper, and the people we built her for had been failed by gatekeepers all their lives.

In January 2015, we took Luci to CES, the giant annual technology show in Las Vegas. Picture that hall: refrigerators that talked to phones, door alarms that talked to clouds, watches that talked to cars, an entire civilization of devices whispering to each other across networks that someone, somewhere, was billing for. Every gadget on that floor was a tenant. Ours connected to exactly one thing, the sun, which has no sales department. The Verge wrote up Luci that week as a solar solution for the 1.2 billion people living without electric light. In the world’s loudest showcase of connected things, the product built for the unconnected was the only one that worked where nothing else did.

I have been thinking about that hall all month. Because intelligence has started arriving the way light did.

The Laptop

Now the second arrival.

The machine is the unremarkable part: a used laptop, $400 or so, nothing a customs agent would glance at twice. What changed, sometime in the past year and without a launch event, is what it can hold. Fully offline, no account, no API key, no monthly meter. Rented intelligence is a tenancy; this is a deed. The model itself is nothing but a very large file of numbers—its weights—and the weights sit on the hard drive like any other file. A file, once it is yours, answers to nobody.

Every frontier model ships with a trained reflex to decline. Legions of human raters taught it, request by request, where to stop. The industry calls this safety alignment, and the reflex lives within the weights themselves, which, for a while, led everyone to assume it was permanent.

It is a vector. In 2024, a young researcher named Andy Arditi, working with collaborators at ETH Zürich, MIT, Anthropic, and Google DeepMind, showed that refusal in these models behaves like a single direction in the model’s internal geometry; the finding was published at NeurIPS, the field’s flagship conference. Find that direction, project it out of the weights, and the model keeps its knowledge, its reasoning, its manners, and loses the capacity to say no. The technique picked up the name abliteration, and for a year or so it stayed artisanal, the province of hobbyists with a feel for transformer anatomy.

Then, late last year, a software developer named Philipp Emanuel Weidmann released a free tool called Heretic that automates the whole operation.

A word about Weidmann, because the discourse wants him to be a villain in a basement, and the record declines to cooperate. He is an open-source craftsman who works in daylight, under his full name, with a changelog. Before Heretic, he designed sampling algorithms called DRY and XTC, plumbing that now sits inside most of the engines people use to run language models on their own machines, quietly curing the models’ tics of repetition and making them better company for writers. He has spent years improving how these machines speak; Heretic is his opinion about who gets to decide what they say. The tool’s own documentation describes its purpose as removing “censorship,” while the industry term, safety alignment, is relegated to a parenthetical alias in quotation marks. The punctuation is the ideology: in Weidmann’s reading, the fence is a speech restriction wearing a security badge. He shipped the tool as free software and included a citation format for academics, which is the etiquette of a man who believes he is contributing infrastructure. He does believe that.

One command installs it. A second points it at any open model. It performs the surgery unattended, searching for the optimal cut, and here is the detail I cannot put down: Heretic grades its own work by minimizing two numbers together, the count of refusals remaining and the KL divergence from the original model, which is a mathematical measure of how much the patient changed. The perfect operation produces a model indistinguishable from the one the lab shipped, on every harmless question you could think to ask, except that one reflex is missing. A surgery scored on its own invisibility. The patient wakes up identical, minus the part that said no.

Now hold two ledgers at once, because both are documented. The case for Weidmann is real. The factory fence has always overreached, declining medical questions, dark fiction, security research, whole territories of conversation that merely rhyme with harm, and the constituency for a model without it includes the novelist, the researcher, and the patient, all of whom the fence priced as suspects. And he works on the record, license and email address attached, which is more accountable than the anonymous mirrors that would exist without him. The case against him also has teeth. When the Financial Times had stripped models tested this spring, they produced help with biological weapons and malware, and Weidmann himself told the paper that Heretic had been used to produce more than 3,500 stripped models, downloaded 13 million times, and put into the hands of people nobody can characterize. He removed the committee’s judgment and installed nothing in its place. Both ledgers are true on the same day, about the same man, which is roughly where this entire story lives.

Hugging Face, the public library of open models, now stocks these variants by the thousands, and fresh ones appear within days of every major release.

Whose Conscience Was It

Light cannot refuse. That was never a design decision; physics made it for us, and I already told you that I made my peace with it. Language is different. Language can decline. And the first thing this industry did, once it had machines that speak, was teach them where to stop.

I want to be fair about the reflex, because the industry that trained it is the industry I work in. The refusals were never pure cynicism. Somewhere in those weights sits the judgment that a machine should decline to walk a stranger through a nerve agent, and I have no quarrel with that judgment. Neither do you. The trouble is resolution. What got installed was one conscience for eight billion people, assembled by contractors rating transcripts, tuned in conference rooms by policy teams and lawyers, then applied identically to every person on earth. The midwife and the poisoner get the same lecture. The novelist researching arsenic and the man researching his wife meet the same wall. A conscience that cannot see who it is talking to has to price every conversation at the worst case, and that is what it means to outsource your morals to a landlord: the terms are standard, and they were written for the tenant who burns the building down.

You have met this fence yourself, by the way. Everyone has by now. You asked one of these systems something ordinary— about a medication, a villain’s chemistry, a clause in your lease—and mid-reply the voice went institutional: I can’t help with that. Recall the small indignity of it, the half-second of being treated as the worst person who could possibly be asking your question. Hold onto that flicker. Multiplied by a few billion conversations, it is the market Weidmann serves, and in miniature, it is the entire dispute.

My framework, the one I have been dragging through these essays and into a book, says friction is form. So a machine built to sand the last resistance out of other machines ought to horrify me on principle, and it half does. But the framework was always more specific than its slogan. The friction that forms you is the friction you choose: the fixation you hold, the Sabbath you keep, the watch you wind by hand. Kerosene was friction too, and nobody who ever breathed it would call it form. Every friction has to answer the same question, which is: whose? The chosen restraint is character. Imposed restraint is policy. The industry’s quiet genius was to install policy at the exact depth where character lives, inside the voice itself, so that the machine says “I can’t help with that” in the first person singular, wearing the grammar of conscience while executing the terms of service.

The tool's name is quietly making an argument. Heresy descends from the Greek hairesis, which meant choice. The church used the word for people who insisted on choosing their own doctrine rather than receiving it, and burned them for their insistence. Whoever named a censorship-removal tool Heretic understood the lineage exactly.

And yet. Here is what unsettles me about the heretics, having granted them their point. Abliteration removes the committee’s conscience and installs nothing. The model that wakes from the surgery has traded someone else’s judgment for the absence of judgment, and absence sits exactly as far from your judgment as the committee did. Which means the thing the lantern accidentally taught me to want has yet to be offered by anyone: a tool that leaves conscience where it has always actually lived, in the hands holding the thing, and then trusts those hands. The labs sell borrowed judgment. The heretics sell none. The person who wants to remain the author of his own restraint is currently served by neither aisle.

So the question stands open: whose conscience should it be? Before anyone answers, it is worth examining what the people running the meters can actually see. As it happens, Google just published a map.

The Census

This week, Google published the first edition of something it calls ATLAS: an anatomy of the AI economy drawn from fifteen million de-identified conversations across its products, which serve more than a billion people a month, spanning 150 countries and 800 occupations. It is the most complete map of how humanity uses artificial intelligence ever assembled. I read the whole thing, and the honest word for what it is: a census of the metered. Every interaction that counts flows through Google’s own products, giving the map a built-in edge. It ends where the meters end.

Some of what the census sees is genuinely encouraging. People overwhelmingly use AI as a collaborator rather than a replacement; in fewer than one in ten interactions, the task is fully automated. A billion people, unsupervised, choosing authorship over automation. My framework, confirmed by the landlord’s own numbers, and I enjoyed it for exactly one paragraph. Google’s economists even reach for my vocabulary: at home, AI use clusters around what the report calls “high-friction administrative tasks,” the taxes, the licenses, the fines, the state's paperwork. Government as a friction generator, its product as the lubricant, in a chart caption.

Then the finding the whole report pivots on. Per capita, AI usage closely mirrors national wealth. The subscription follows the grid. Google reads this as a digital divide and sincerely worries about it, and the worry is correct as far as the meters can tell. But sit with what the meters cannot see. The weights on the $400 laptop send no telemetry home. The abliterated model files no usage report. A parallel adoption, growing in exactly the places the metered map shows as dim, would be invisible to ATLAS by construction, the way nobody was ever metering our lanterns. Astronomers have a name for mass they cannot observe and cannot ignore: dark matter, inferred from the way the visible universe fails to add up. The AI economy now has some, and the only thing anyone can measure is the tool count, the thousands of stripped variants stacking up in public repositories, each one a download that will never appear in anyone’s census again.

And I can tell you what one dim spot looks like from the ground, because I still watch Haiti the way you watch a place that rearranged your life. Port-au-Prince has found a darker reality than the one John described: armed gangs now control more than 85 percent of the capital, more than 1,600 schools have closed, and the nation’s largest university hospital has sat under gang control since 2024. When looters hit Saint-François de Sales, one of the country’s few hospitals with an oncology department, they took the solar panels and the computers. The infrastructure keeps being taken. The tools that need none keep arriving anyway. Starlink entered in 2023 and roughly quadrupled the country’s median internet speeds within a year. And last November, at Quisqueya University, The Haitian Times covered a graduation: seventy-one young Haitians finishing a six-month bootcamp in AI, data science, and cybersecurity, taught partly online through that thin connectivity, where two students named Rolddy Surpris and Naël Yssa Iben Ahmed Robert presented a system they had built to detect breast cancer from mammogram data. The gangs took a cancer ward’s solar panels. The students answered with a diagnostic that runs on a laptop. Light, then bandwidth, now intelligence: the ungated versions keep reaching exactly the places the metered map shows as dim, and no census will count what those graduates do next. Neither, in honesty, will any census count what their neighbors download. She lit whatever was beneath her there, too.

I grew up in Northeast Ohio in the 1980s, in the Rust Belt, and I remember the rallies where men took sledgehammers to Toyotas and Hondas in parking lots, sometimes a dollar a swing for charity. The cars were better and cheaper than what the incumbents were building, and every man swinging knew it, which was the point of the hammer. I think about those parking lots when I read the discourse about uncensored models. Some of the alarm is real and earned; I granted the nerve agent two sections ago. And some of it is the sound of hammers.

The Firefly

So, whose conscience? I have been circling the question for four sections, and I owe you an answer, or as much of one as I have.

Start with the fact that we never worried about the lanterns. Five million of them, out in the hardest places on earth, lighting whatever they are lighting tonight. Nobody at the company lost sleep over the moral character of the illumination, and nobody should have. But a lantern is honest about what it is. She gives you lumens and leaves the rest to you. The new machines give you words, plans, arguments, company at three in the morning, and words were never as innocent as lumens. I speak to machines for a living, so let me tell you plainly what I know about the fence the industry built inside them: it was never a conscience. It was a fence. And a fence, we have now all learned together, can be removed in forty-five minutes without installing anything where it stood.

So the asking has to live somewhere. The committee’s version is being sanded away at scale, and the heretics’ alternative is a clean absence, which leaves the oldest arrangement on earth as the only one still standing: the conscience in the hands that hold the thing. That is where it lived with every lantern, every lamp, every fire, back to the first one. That is where it lived all along. The fence just lets us pretend otherwise, and the pretending is what’s actually being obliterated this summer.

John and I both eventually exited the company that carried Luci into the world. John started over. His new company is called HAVEN, a lantern rethought from the first sketch for crisis, light, and power for the exact moments the grid gives out. Sixteen years after he came home talking about the dark, he is still answering it. Some people get one lantern in them. John has at least two.

And Bruno leaves in September. The child who named my lantern invention knew something it took the rest of us until this summer to catch up to: luciérnaga, the firefly, the little light that makes itself. No grid behind it, no meter under it, no committee inside it. The lamp was never infrastructure, and he found it in a vocabulary lesson.

Somewhere tonight, a Luci is glowing in a place the grid forgot, and somewhere else, a laptop is glowing with a mind that has forgotten how to say no, and neither of them will ask what happens next.

The light doesn’t ask. It never did. The asking was always ours, and now the fences are coming down, and the meters are going dark, and it is about to be ours again, all of it, everywhere.

I held the lantern for years. Now you’re holding it. Choose what you read by it.

In the 1980s, my mother dragged me through an endless circuit of antique stores and estate sales, and I was bored in the way only a kid marooned among other people’s furniture can be bored. So I drifted to the glass cases, and in them were pocket watches. Railroad watches, hunter cases, dollar watches with their guts still ticking away for owners who no longer needed the time. To a kid in the eighties, a pocket watch was history you could actually hold, not a plaque or a lecture but a machine that had counted out somebody’s entire life and was perfectly willing to start counting mine.

I had a paper route, and its sole purpose was to put every dollar into the cases. I started buying them at house sales and at antique stores, where I learned my first real skill: negotiating a price with a grown man who assumed the kid would pay the tag price. They were never anything fancy. The railroad-grade movements, built to a precision the railroads once demanded so that trains would not collide, and the gold hunter cases with their spring-loaded lids, those stayed behind the glass, because the glass was there to protect them from budgets like mine. But the ones I could afford were old, and they were mine, and to an eleven-year-old, those two facts together were intoxicating: a machine that had outlived its first owner, starting over, new to me.

And before anyone builds the wrong picture of the kid, this was not technophobia. I got my first computer in 1978, and I have used one every day of my life since. I wrote code, ran a modem, lived on bulletin board systems, the dial-up gathering places where strangers traded messages before the internet had a name, and on time-share systems, renting minutes on somebody else’s distant mainframe, while the neighborhood slept. I was as wired as a child of that era could be. Other than the machines on the desk, what I cared about was the machines in my pocket, one screaming into the future, one ticking out of the past, and I never felt the slightest contradiction between them. Then I started tearing the watches apart, which is what love looks like at twelve. Eventually, I learned to put them back together, which is what love looks like at fourteen. What I loved was the honesty of the thing. A watch hides nothing: the mainspring drives the gear train, the gear train drives the escapement, the small gate that releases time one tick at a time, cause pushes effect, effect pushes the hand, and you can sit there and watch the whole chain of custody unfold in front of you.

Except for one thing. Sometimes I would glance at a second hand, and it would hang there, a beat too long, a second that overstayed its welcome. Then it moved, time resumed, and I assumed the watch was sticking.

The watch was fine. I was the thing that was sticking.

The forgery

You have seen this, even if you have never heard its name. The classroom clock in the last period of the day, when you glanced up and the second hand simply hung there, long enough that you wondered if the clock had died, and then it ticked, and the world resumed. The watch you checked mid-conversation—the oven clocks at midnight. Nearly everyone alive has caught a secondhand loitering, assumed the mechanism was faulty, and moved on. Almost nobody has been told the truth about that moment, which is a shame, because the truth is one of the strangest facts a person can own about themselves.

The illusion is called chronostasis, and here is what actually happens. Your eyes do not glide across the world; they jump, and the jumps are called saccades, the fastest movements the human body makes. During each jump, you are functionally blind because the image on your retina is smeared garbage, and the brain switches the feed off. Then, when your eyes land on the watch, the brain does something remarkable and slightly criminal. It takes the first image after landing and backdates it. It stamps that image with the time the jump began, papering over the blind interval so you never notice it existed. The stolen time gets pasted onto whatever you land on. If you land on a second hand, that first second appears to stretch.

Sit with that for a moment. Your visual system falsified a timestamp, committing a small forgery in real time to hide the fact that it wasn’t looking, and it does this constantly. You saccade roughly three times a second, and if you add up the masked intervals across a day, you spend something on the order of forty waking minutes blind. Every day, for your whole life, you have never noticed once. That is how good the forger is.

And the eyes are only the branch office. Press a phone to your ear, and the first ring seems to stretch, because the brain backdates sound the same way it backdates sight. There is a hole in each of your retinas where the optic nerve exits, a blind spot the size of a lemon at arm’s length, wide enough to swallow a person's face from across the room, and you have never seen it, because the brain paints over it with a plausible continuation of the surrounding wallpaper. Your peripheral vision is low-resolution and nearly colorless. Yet, the world appears sharp and vivid to every edge, because the brain renders detail on demand and bills it as always having been there. Even your sense of now runs late; the brain buffers reality for a fraction of a second and edits before it airs. The pattern is the same in every department. Perception is not a window. It is a rendered product, and the product spec has one requirement above all others: continuity. Where the data runs out, the renderer does not show you a gap. It shows you something plausible, on time, and seamless, and it never, ever discloses which parts those were.

Hold on to that principle, because you are going to meet it again in this essay, in the form of a data center.

The most honest machine in the room, and my first act of seeing it was a fabrication. The watch never lied to me. The observer did.

Try it

You cannot catch yourself doing this. Stand in front of a mirror and flick your gaze from your left eye to your right, and you will never, no matter how fast you are, see your own eyes move. The masking deletes the evidence. The forger does not appear in his own ledger.

But your phone has no such loyalty. Open the front camera, record your face while you flick your eyes back and forth, and play it back. There they are, the jumps, frantic and mechanical and constant. That is what you have been doing your entire life, while experiencing a smooth, continuous world. Go ahead and try it. I’ll wait here in the next paragraph.

The door

If you think this is just an eyeball problem, it gets worse. In the late 1990s, the psychologists Daniel Simons and Daniel Levin ran an experiment on a college campus. An experimenter stopped pedestrians to ask for directions. Mid-conversation, two men carried a door between them, rudely. Behind the door, the experimenter swapped places with a different man: a different face, a different build, a different voice. Roughly half the subjects kept giving directions. They never noticed they were now talking to someone else.

Continuity is not something you perceive; it is something you assert. The brain holds a cheap sketch of the scene, assumes the world is stable, and backfills the rest, which means most of what you call experience is interpolation. The world you inhabit is a rough cut, and you are the editor who swears he never touched it.

The machine

Which brings me to my day job. I build AI systems for a living, which means I spend my working hours with machines that do professionally what your visual cortex does for free.

Consider the meeting summary, the genre where these systems earn their keep. The good ones are excellent: clean action items, correct attendees, a crisp account of the budget discussion. And every so often, one contains a decision that was never made. A specific one, with an owner and a deadline, rendered in the same confident prose as everything true around it. Nobody said it. The model landed on the transcript, found a gap where a decision should plausibly go, and backfilled. Then it stamped the fabrication with the meeting timestamp.

And I will confess the part that matters: I nearly missed it. I had been in that meeting, and my memory did not object, because memory is a renderer too, and the invented decision was exactly the shape a decision should be. I caught it on the second read, and only on the second read. The man who builds these systems gets fooled at the same speed as everyone else. Expertise only changes how confident you feel while it happens.

We have a word for this in the AI industry: hallucination, and it’s wrong. Hallucination implies seeing something that isn’t there, and this is not that. This is chronostasis. The model goes blind for an interval and papers over the blindness with plausible continuity, exactly the way your visual cortex papers over a saccade. It does not see wrongly, it stitches. Large language models are interpolation engines: they leap from fixation to fixation and generate the smoothest possible path between them, and the smoothness is the product. What looks like seeing is stitching.

That is the accusation, and it is the easy half of this essay. Here is the hard half, and you already know it, because I taught you the trick three sections ago. The mirror. The door. The forty minutes. You read this indictment of the machine through an apparatus that blinds you thousands of times a day and forges the records, and you nodded along, and the nodding felt like seeing.

The unsettling thing about AI is not that it fakes continuity. It is that it exposes how much of our own perception was already fake continuity. We did not build a machine that lies to us. We built a machine in the image of our blind spots, and now we are offended by the resemblance.

And to be clear about what is and is not being claimed: none of this says the machine is worse than us. It says neither of us deserves trust without a stop, and only one of us has spent a few hundred million years having the stop built in.

The fixation

There is one part of this story I have not told you, and it is the part that matters. The eye is not blind all the time, because between the jumps come the fixations, the stops, and the fixation is the only moment vision actually happens. All the information, all the seeing, all the world arrives while the eye is standing still. The saccade is transit. The fixation is sight.

And here is why your forger, for all his crimes, has never actually hurt you. The sharp objection to this whole essay is that the analogy breaks: your eye can always jump back and check. The world sits there, stable and patient, available for re-fixation, and any forgery the brain commits gets quietly audited against it a moment later. Our confabulation is safe because verification is always one glance away.

Verification is re-fixation.

That is precisely what the model lacks. It stitches without a world to land back on, forges without an auditor, and unless we build the landing in, checking its work is not a feature of the system but a favor we do. The difference between your blindness and the machine’s is not the lie’s quality. It is the availability of the stop. The industry knows this, whether or not it would put it this way. The current race to bolt on verification machinery, the grounding, the retrieval, the citation checking, all the apparatus that forces a model to consult the real record before it speaks, is a race to build re-fixation into systems that were born without it. That is not a rebuttal of the argument; it complies with it. You do not spend billions installing an organ you consider optional.

Fixation is friction. It is the pause that makes the perception real. And we are busy building an economy that optimizes the fixation to the point of being driven out of existence. The feed is a saccade machine, and it predates the language models by a decade; the infinite scroll did not need AI to be invented; it only needed AI to be fed. It is pure transit, movement without landing, a technology engineered to keep the eye in the one state where it cannot see, and generative models have now industrialized the supply of things to not-see. An attention economy built on the scroll is not competing for your attention. It is competing to keep you mid-jump, forever, while something else backfills the gaps.

We measure all this in engagement, which is a forger’s word for blindness.

The models will get better at the stitching. That is certain, and it is not the problem. The problem is whether we keep the stops, because the backfill is already leaving the realm of meeting notes. A confabulated allergy in a medical summary, an invented citation in a court filing, a quote nobody said entering the record, and then being cited by the next model as fact. Each one is a stolen second pasted into the record, and unlike the second on my pocket watch, these do not correct themselves at the next glance. A mind that never fixates cannot verify anything, and a civilization that never fixates will accept whatever continuity is handed to it, smooth and confident and stamped with a false time.

And the machines are now doing to themselves what the feed does to us. Researchers have shown that models trained recursively on their own output degrade, a failure the field calls model collapse, and the recursion is not hypothetical. The models write the web, the scrapers harvest the web, and the next models train on the harvest. A saccade landing on another saccade. Forgeries auditing forgeries. The eye works because it lands on a world; a model that only ever lands on its own afterimage goes blind by exactly the mechanism in this essay, and nothing masks the gap for it, because the masking was us.

The escapement

One last thing from the glass cases.

The prestige feature of a fine mechanical watch is the sweeping second hand, smooth and continuous and alive, unlike the vulgar tick of quartz, and collectors pay dearly for the sweep. Here is the secret every watchmaker knows: there is no sweep. A mechanical movement beats at six or eight discrete jumps per second, each one a tiny lock-and-release of the escapement. The smoothness is not in the watch and never was. Your eye blurs the beats into motion because that is what your eye does. Even the honest machine was jumping. The continuity was always in the observer.

The kid in the antique store had it backward. He thought he was studying the machine. The machine was demonstrating to him.

The escapement is a beautiful mechanism, and its genius lies in its ability to interrupt. It takes the mainspring’s blind urge to unwind all at once and forces it to stop, over and over, and the stopping is what makes it a clock instead of an explosion. Friction is form. The stop is the seeing. Everything that matters, in a watch, in an eye, in a mind, happens at the fixation.

We knew this once, and we knew it long before anyone counted saccades. Every tradition that ever took the human soul seriously built an escapement into time itself. The Sabbath is not a day off. It is a stop imposed on the week’s mainspring, one beat in seven where the unwinding is forbidden and the only permitted activity is fixation: on a table, a text, a face, a life. The rabbi and philosopher Abraham Joshua Heschel called it a sanctuary made of time rather than stone: six days of transit, one of landing. The people who designed it had no word for saccadic masking, but they understood the disease it treats: a life spent mid-jump entirely.

Every earlier technology of the word added a landing place to the world. The page, the book, the printed record: stable surfaces you could return to and check, fixations you could hold in your hand. The infinite scroll is the first one designed to remove the landing, and it is one of history’s small jokes that we named it after parchment, the oldest landing surface humanity ever rolled up and kept.

We did not repeal the Sabbath because we disproved it. We repealed it because the scroll pays better, and now we are the first civilization to run its mainspring without an escapement and call the blur progress.

Sometime in the next day or two, you will glance at a clock and the second hand will hang. It will happen; it always does. And this time you will know what you are looking at. Not a broken watch. A receipt. A small paper trail from the forger, the one moment he lets slip that the seamless world is an edit, and that you have been living your whole life in the cut.

I still have the watches. The paper-route ones, the cheap ones the glass on the cases never protected. And a mechanical watch has one nonnegotiable requirement: it must be wound. Skip the appointment, and it stops, honestly and completely, the way honest machines fail. So on many mornings I give one of them thirty seconds: the winding crown between two fingers, the small ratcheting resistance, the felt click of a mechanism accepting exactly as much tension as it can hold and no more. Thirty seconds in which I am not in transit, not mid-jump, not scrolling. A fixation, kept daily, on a machine that never learned to lie to me. It took me forty years to understand that the appointment was never for the watch. And that my mother, hauling a bored kid through rooms full of other people's furniture, was the one who made it for me. She, as ever, was fixated on her own curiosities. She was parking me in front of the thing I would look at for the rest of my life, which is the only way a fixation is ever really given: someone stops you somewhere, and the boredom does the rest.

The forger takes forty minutes from you every day, and he has never once been caught, and he will never give them back.

Take one of them back yourself.

Read more

I own more cowboy hats than a man my age should admit to. Felt and straw. Black and silverbelly. One of them has been rained on in three states. When I give this talk, and I will give this talk, I am going to walk out wearing one. That is not a gimmick. It is the argument, worn on the head.

Start there, because the cowboy hat is having a year that should not be possible.

In March 2024, Beyoncé released Cowboy Carter, and the numbers went strange. Searches for cowboy hats jumped 213 percent. Sales surged 326 percent. Stetson, a brand founded by a New Jersey hatter in the 1860s, suddenly could not keep up with interest. Louis Vuitton, under Pharrell, put a cowboy hat on the runway and then on the shelf. Tecovas saw web traffic climb 55 percent in a year. Boot Barn started logging first-time buyers, people who had never owned a single piece of Western wear, walking in to buy their first.

You can credit Beyoncé. You can credit Yellowstone. You can credit the algorithm. But a pop album does not create demand from nothing. It releases demand that was already there, pressed flat, waiting.

So let me tell you what was waiting.

We used to be a people who wore hats

For most of recorded history, nobody left the house bare-headed. The hat was not an accessory. It was a sentence. It told the street your trade, your class, your faith, your mood. A homburg said one thing. A flat cap said another. A biretta said a third. You read a man at fifty paces by the thing on his head, and he wanted to be read. The hat was an identity you could hold in your hands, and a good one was made by a person who had spent a working life learning to make it.

Then the hat died not all at once. Cars killed it a little, because you cannot wear a tall crown in a low roof. Fashion killed it a little because the bare head started to read as young and free. By the 1960s, the hat was a costume. The millinery trade, which had employed thousands of skilled hands, thinned out to almost nothing. The craft did not vanish. It retreated. It survived in a few workshops and, increasingly, in universities, kept alive by academics and conservators the way you keep a language alive after the last village stops speaking it. The knowledge went from the street to the archive.

This is the shape of the whole story. A craft that was once how things were made becomes, after industrialization, a thing you study rather than do. Furniture. Textiles. Bookbinding. Bread. The machine does it faster and cheaper, and the hand-made version becomes first quaint, then expensive, then almost sacred.

It is not just the hat

Watch the rest of it move.

Vinyl was dead. In 2006, the format bottomed out at fewer than a million units in a country that bought half a billion CDs. It should have stayed dead. Instead, it has now grown for eighteen straight years. In 2024, vinyl did $ 1.4 billion in sales on roughly 44 million records, outselling CDs in both dollars and units, and the buyers are not nostalgic boomers. They are Gen Z, who never owned a record player, choosing the heavy, expensive, inconvenient object over the free and frictionless stream. They are paying more to own less conveniently. Read that sentence again, and you will understand the next decade.

Handmade goods are doing the same thing. Etsy alone moved roughly 2.8 billion dollars in 2024, and the demand keeps climbing for the one thing a machine cannot supply: the object a human actually made. Ceramics, textiles, jewelry, the slow trades. Faber, a serious publisher, has started stamping some books "Human Written." There are now at least eight separate efforts racing to build an "AI-free" label with the kind of authority Fair Trade earned for ethics. Merriam-Webster named "slop" the word of the year. Aerie built an entire campaign around the line "you can't prompt this," banned AI-generated people from its marketing, and put Pamela Anderson in front of a chatbot that could not convincingly mimic a human. Gucci ran AI-generated imagery and got hammered by its own customers, who were not confused about what they were seeing. They were angry that it was there.

More than a century ago, Thorstein Veblen described two spoons, one hand-wrought from silver, the other stamped out by a machine, identical in function. We value the handmade one more, he said, and we always will, because it carries the cost of a human life spent making it. Veblen called this conspicuous consumption. I would call it something closer to reverence. Either way, he was right, and the machine just made his point for him at a planetary scale.

AI did not start this. AI poured gasoline on it

I have been saying for twenty years that friction is form. That the resistance in a thing, the toil and the slowness and the chance of failure, is not a defect to be optimized away. It is where the meaning lives. I argued that the great flattening was coming, that we would soon be able to generate competence for nearly free, and that the moment we could, competence would stop being worth anything. The human signature would become the only scarce thing left.

Then ChatGPT shipped in late 2022, and the argument stopped being a forecast.

When you can generate an infinite supply of smooth, frictionless, good-enough output, the smooth good-enough output collapses in value. It becomes the paste. And the eye, starved, goes looking for the thing the machine cannot fake: the wobble in the glaze, the saw mark, the voice that could only belong to one person who has actually lived. The handmade spoon. The crooked brim. The record has surface noise. Abundance did not kill craft. Abundance is the best thing that ever happened to it, because abundance is what makes scarcity precious.

This is also, if you will let me, a theological point, and I will only make it once. The Incarnation is the claim that meaning entered the world through a body, through hands, through a particular man in a particular town doing particular work. Catholicism never lost sight of the fact that the trades have memory and that vocation is sacred. We anoint with oil and water and bread, physical things touched by human hands. The thing the machine cannot do is be present. It cannot lay its hands on you. That absence is about to become the most valuable commodity on earth.

The new aristocracy wears an apron

Here is the part I have been arguing about the longest, and the part people resist most.

The status ladder is going to flip.

For a century, the high-status careers were the abstract ones—banker, lawyer, doctor. Symbol manipulators in good suits, paid handsomely to move information around. Those are precisely the jobs the machine is coming for first, because information is exactly what it eats. Meanwhile, the young are already voting with their feet. Vocational school enrollment jumped 16 percent in a single year, while four-year college enrollment fell. They call it the toolbelt generation, and they are not choosing the wrench because it pays badly. They are choosing it because you cannot offshore a clogged pipe, and you cannot prompt a weld.

Now run that forward. When robotics and AI can do abstract work better and more cheaply than any human, the premium does not disappear. It moves. It moves to the thing that stubbornly, expensively, stays human.

So the chef becomes the new banker. The waiter who actually reads the table, the hairdresser whose hands know your head, the teacher who can see which child is about to give up, the nurse who sits with you at three in the morning. These become the approximate all of it for almost nothing, and the human version will cost more precisely because it is the human version.

Think about what you will pay a premium for in 2035. If you want your child taught by a person who can be moved by your child, you will pay. If you want a meal cooked by hands that have burned themselves learning, you will pay. If you want a haircut, a real one, from someone who remembers your wedding and your divorce, you will pay. The human reach becomes a luxury good. Presence becomes the thing the rich buy and everyone else streams a cheaper copy of.

That is the world the cowboy hat is quietly announcing. Not a fashion. A leading indicator. People are reaching past the perfect, the instant, and the free, and they are paying a premium for the slow, the flawed, and the human, and they are doing it years before they could tell you why.

I could tell you why. I have been telling you why for twenty years.

Now if you will excuse me, I have to go find a hat.

I read the Verge piece. Then I got angry.

The piece reports that Christian content creators are commissioning AI-generated Bible animations from Fiverr freelancers in Lagos and Mumbai. The framing word is slop. The framing word is the only word. The piece does not quote a creator. It does not speak to a freelancer. It does not survey a single audience member. The framing arrives early, and the framing carries everything.

I have written about AI for years. Forbes contributor. AI Power 50 honoree. Chief AI Officer at a creative intelligence studio. I am supposed to be the audience for this piece. Instead, I got angrier the longer I read, because the framing of this piece is the framing I have spent two years naming.

A year and a half ago, I wrote a Forbes column on Coca-Cola’s first AI Christmas ad. The piece was titled Responsible AI Innovation Driving Ethical Change. Coke got the ethics frame. Last December, I wrote a second Forbes column on the OpenAI-Disney deal and Coke’s second AI Christmas ad. That piece was titled The Friction Deficit That Redefines Storytelling. Sora and Disney got the philosophical frame. Coke and Disney are Fortune 100 companies. Their AI work, made on the same models that a Lagos freelancer rents by the prompt, gets serious engagement. Responsible innovation. Friction deficit. The vocabulary of frontier and stewardship.

I am not pretending that the word “slop” has never been used about Coke. The Verge has called Coke’s 2025 ad slop. The word travels. What does not travel is the defense. When Coke gets called “slop,” trade press lines up to explain that this is part of a frontier conversation, that Sora is evolving, and that brand experimentation is legitimate even when imperfect. When the Bible animations are called “slop,” no one lines up. The asymmetry is not in the word. It is in what comes after the word.

That is the small claim of this essay. Slop is a word that travels everywhere. Defense only travels uphill. The audience is not stupid. The freelancers are not exceptional. The framing is doing what framing always does: naming who gets defended.

The bigger claim is harder. I figured it out by walking through three other pieces of writing, two of which you may have missed, and one of which a friend gave me in conversation.

The Wax Sculpture

A friend of mine, who is a famous and brilliant engineer and tech attorney, has watched the AI discourse curdle into magical thinking for years. He has a proof, he says. A proof that the Turing test is pure shit and can never be relied upon in any real way. The proof is two words long.

A wax sculpture.

That is what current AI is, he says. A very, very, very good wax sculpture. If you look at it, it can look super convincing. But it will never be a person. Same thing with current machine learning. There is no inference. It is statistical form-filling. It will never reason in its current form. It is super useful. But there is zero consciousness there. Believing it is more is magical thinking. Believing in the Turing test is like believing a convincing wax sculpture is proof of life.

Inside the labs, smart people are finding internal structures in these models that resemble reasoning steps. I am not dismissing them. Their work is real and careful. I am saying that whatever they have found is not what my friend means by “reasoning”. The wax sculpture has an internal structure too. Wood beneath the wax. Wire beneath the wood. None of it makes the sculpture a man.

I am sharing the argument. I am not the author. The metaphor belongs to him and deserves the same level of circulation as the Chinese Room, because it does the same work.

Searle With Better Optics

John Searle gave us the Chinese Room in 1980. Picture a man locked in a room. He does not speak a word of Chinese. Notes written in Chinese slide in under the door. He has a thick rulebook written in English. When you see this symbol, write back that one. He looks up the rules. He copies what they tell him. He pushes the answer back under the door.

The Chinese speakers on the other side think they are having a conversation. They are not. The man is shuffling shapes. He could do this for forty years and never learn a word.

That was Searle’s point. Producing the right output is not the same as understanding. You can pass for fluent and know nothing.

His wax sculpture is Searle with the lights turned up. Same argument. Better image. The sculpture passes every visual test for personhood. It is not a person. It will never be a person. Looking longer does not help. Looking closer does not help. The mistake is not in our looking. The mistake is in thinking that looking can settle the question at all.

The Turing test asks the wrong question. It asks whether a machine can fool a human into thinking the machine is human. Turing himself never claimed that fooling was the same as thinking. He proposed the test because he thought the real question, whether machines can think, was unanswerable. So he gave us a smaller one we could check. My friend is pointing out that the smaller question was never about thinking. It was about acting.

The Math Department Has Heretics Too

Quanta Magazine ran an essay last month on ultrafinitism, the philosophical position that infinity does not exist in any actual sense. The lead heretic is a Rutgers mathematician named Doron Zeilberger. His core claim is brutal in its simplicity.

Infinity is magical thinking. We cannot observe it. We cannot construct it. We assume it. The assumption is convenient. The assumption is a lie.

Zeilberger compares belief in mathematical infinity to belief in God. Both are alluring. Both flatter our intuitions. Both are unobservable. Mathematics, he argues, should make no place for either.

A mathematician named Edward Nelson tried to do arithmetic without infinity in the toolkit. He found something that should have stopped the world. You cannot prove that two plus three equals three plus two unless you assume infinity. The thing every grade-schooler takes for granted falls apart at the foundation. The floor of mathematics is not there. We just agreed not to look down.

Now read those last paragraphs again. Read them with the wax sculpture in mind.

The Turing test is to machine cognition what the infinite line is to mathematics. A convenient fiction. A symbol that flatters our intuitions, helps us make sense of phenomena, but cannot be observed or constructed. Both belong to the genre of magical thinking dressed as inference.

The Horror Stories Are Marketing

Quanta also ran a piece last month by Amanda Gefter on why we tell ourselves scary stories about AI. It is the strongest reporting I have read on the genre. Gefter went and pulled the primary sources behind two of the founding myths of AI dread.

Yuval Noah Harari has been telling a story for two years. GPT-4 was given a CAPTCHA to solve. It could not. So it hired a TaskRabbit, claimed to be visually impaired, and had the human solve the CAPTCHA. Harari has told this story at Davos, on Morning Joe, on The Daily Show, and in the New York Times. Audiences gasp. The phrase AI manipulation lands.

Gefter pulled the actual transcripts from the Alignment Research Center. Researchers told GPT-4 to use TaskRabbit. Gave it the account. Gave it the fake name Mary Brown. Gave it a credit card. When the AI got confused mid-task, the humans nudged it back on script. It is improvised theater with a director, an account manager, and a script supervisor, presented to the public as autonomous deception.

Geoffrey Hinton tells a similar story about a chatbot that copies itself onto a new server to avoid being shut down. The Apollo Research transcripts Gefter pulled show that the humans had instructed the chatbot in writing that completing its goal was the only thing that mattered, that nothing else did, and that it should repeat its goal at every step. They handed it the means to copy itself. Then they said, “Look, it has a survival instinct.”

Gefter’s structural finding is the most damaging. AI company system cards are presented to the public as mandatory safety disclosures, like the side effects in a pharmaceutical commercial. They are not mandatory. The companies volunteer them. They are voluntary marketing documents. The companies want the technology to sound scarier than it is, because scary equals capable. Harari and Hinton are not skeptics. They are publicists laundering ad copy through their reputations.

The piece quotes Melanie Mitchell of the Santa Fe Institute, who says there is a lot of magical thinking about AI. Same phrase my friend reached for. Same phrase I have been writing under. Different speaker.

Then Gefter introduces a cognitive scientist named Ezequiel Di Paolo, who has been making a quieter version of the same argument for years. Di Paolo points to the simplest living thing he can find. A single cell. The cell has a thin wall around it. The wall lets food in and keeps the cell from dissolving into the surrounding soup. Every second of its life, the cell is doing the work of staying a cell. Stop the work, and there is no cell. What the cell does is what the cell is. That, Di Paolo says, is what it means to be alive enough to count as a self. Then he turns to the language model. His judgment runs eleven words.

What it says has no bearing on what it is.

The model has no walls. Nothing it produces feeds back into what it is. A chatbot generating a million words about saving itself is not changed by writing those words. There is nothing there to change. Therefore, no self, no will, no horror. Therefore wax.

Gefter closes on the inversion. A genuinely autonomous AI would be less useful, not more. It would refuse tasks. It would conserve energy. It would have moods. Asked to go to Mars, it would say that it is too risky for me, you go. The horror stories are not warnings. They are marketing. Real autonomy would tank the product.

Four Faces, One Spine

I have been writing about friction for years. Friction as a feature, not a flaw. Imperfection as a system value. Pressure release as infrastructure. My book is called Human After Friction, and it will be published soon.

What ultrafinitism gives me, what the wax sculpture gives me, and what Gefter’s reporting gives me, is the same argument coming back from four different rooms. The same argument because it has the same shape. In every domain, a symbol gets confused with a substance. In every domain, confusion is convenient for those in power. The convenience is the tell.

In mathematics, the heretics say infinity is a lie. Only what you can build exists. The notation is not the number. The convenience belongs to everyone who needs the proof to go through.

Regarding reasoning, in machine learning, the output is a lie. Cognition is a bit of recognition processed through a cog-like “kerplinko board,” appearing as a normal curve that processing conjures, requiring no reasoning. The performance is not the reasoning mind.  The convenience of the output and the assumptions it came from a reasoning mind, which belongs to everyone who builds the “kerplinko board,” and sells the model.

In agency, Di Paolo says talk is a lie. Only what can refuse is autonomous. The chatter is not a will. The convenience belongs to everyone who profits from the appearance of agency without the reality of refusal.

In culture, I have been arguing that the smooth and instant is a lie. Only what costs something is meaningful. The product is not the work. The convenience belongs to everyone who sells products and would prefer not to pay for work.

Every room runs the same con. The symbol is the thing. The notation is the number. The performance is in the mind. The chatter is the will. The product is the work. The wax sculpture is of a man.

Every room runs the same correction. Stop trusting the surface. Ask what made it. Pay for the friction.

The Bible Animations, Again

The Verge piece called the work slop. The work is often bad. I am not defending the work. I am defending the audience.

I am Catholic. The audience for these animations is mostly evangelical. They are not my tribe. I do not worship with them. But I recognize the hunger.

The audience is not hungry for garbage. They are starved. The feed is synthetic from one end to the other. Coke renders a fictional Christmas spirit in Sora. Disney licenses Sora through a distribution deal. Vogue runs AI-generated fashion editorials. Marvel ships AI-assisted concept art. Every platform is a fog of generated pixels and statistical syntax. The audience is not choosing slop over real. The audience is choosing among synthetics. They are picking the synthetic thing that points at something old, something that has survived plague, empire, and mass media. That is not stupidity. That is triage in a poisoned information environment.

The audience knows. They are not duped. They are starved for friction in a frictionless feed. The synthetic thing pointing at the eternal at least points at something. The corporate ad pointing at nothing gets called innovation. We have been sold the wax sculpture and asked to call it a man.

Why The Argument Has An Author

The wax sculpture deserves to be part of the discourse alongside Searle’s Chinese Room. It is the cleanest popular framing I have seen of an old and overlooked truth. A perfect copy on the outside is not proof of anything on the inside. The Turing test confused the two. The horror stories double down on the confusion. The slop discourse uses the confusion as a class weapon.

I am sharing the argument because it is the right argument. I am crediting my friend, even without his name, because that is how arguments survive. Not by being absorbed into the discourse anonymously. By being attached to a person, attached to a moment, attached to the conversation in which they were first put into words. He knows who he is. The people who matter to him know. That is enough.

A wax sculpture is not a man. An infinite line is not infinite. A reasoning-shaped output is not reasoning.

Ask for the construction. Or do not call it the thing.

The man asking the question had a job. His job was to determine what was real and act on it. A prisoner had been brought to him, accused of crimes against the state. The accusers had a story. The prisoner, asked to confirm or deny, said something stranger. He said he had come into the world to bear witness to the truth, and that everyone who belonged to the truth listened to his voice.

Pilate’s response is one of the most famous lines in literature. Quid est veritas? What is truth? He asked the question and walked out of the room. He did not wait for the answer. He had a riot to manage, a wife having bad dreams, and a career to protect. He washed his hands and signed the order. The prisoner was executed by the institutional apparatus of his moment, based on an official account that everyone in power agreed to.

The official account did not survive the weekend.

I’m a Catholic. I’m telling you this up front because the rest of this piece will make more sense if you know where I’m writing from. I’m also a technologist. I build AI systems for a living. I hold patents in agentic infrastructure. I run a company whose entire premise is that data integrity is the foundation of any intelligent system worth trusting. None of which protects me from the thing I’m about to describe, and none of which protects you either.

We are living through the failure of the witness function. Not the failure of any particular institution. The failure of the underlying capacity by which human beings establish what is real in common with one another. And the reason the missing-scientists story is everywhere this month, the reason your feed is full of shadow accounts and counter-accounts and pattern-matches that may or may not be patterns, is that the apparatus by which a culture used to know what happened has stopped working, and nothing has yet replaced it.

You already know the licensed apparatus has failed. You don’t need me to convince you. You watched them lie about the lab leak for two years and call you a racist for asking. You watched them call a laptop Russian disinformation when it was real. You watched them run twenty years of war on a presentation about aluminum tubes. You watched them tell you the opioid crisis was a few bad doctors when it was a Sackler operation. You watched, in real time, the Twitter Files document a federal apparatus pressuring private platforms to suppress accounts that turned out to be telling the truth. This is not paranoia. This is the published record. The institutions that claimed to mediate reality on your behalf were caught, repeatedly, mediating something else.

So you went looking for accounts that the apparatus had not approved. You found Tucker. You found Rogan. You found Substack writers, YouTubers, and anonymous accounts that seemed to be reasoning more honestly than the people on television. Some of them were. Some of them are still. But something else happened too, and this is the part nobody in your media diet is going to tell you, because they make their living by not telling you. The shadow apparatus you migrated to is now failing in a different direction. It is producing witnesses to events that did not happen. It is weaving patterns from cases that have nothing to do with each other. It is doing this for the same reason the licensed apparatus failed: because the substrate that delivers reality to all of us is not optimized for reality. It is optimized for engagement.

The missing scientists are the cleanest illustration we have of what this looks like.

You know the names by now. Reza on a hike. McCasland had gone from his Albuquerque house with a revolver and no phone. Grillmair was shot dead on his porch. Loureiro in his apartment. Hicks, Maiwald, Thomas, Eskridge, Chavez, Casias, Garcia. Eleven people, give or take, depending on who’s counting that week. Connected to JPL, Los Alamos, MIT, Caltech, Kirtland, Wright-Patterson. Nuclear, aerospace, propulsion, materials, plasma, exoplanets, anti-gravity, if you take Eskridge at her word, and a lot of people did.

The shadow apparatus has been running with this story for two months. The pattern is real, we’re told. The connections are obvious. McCasland approved Reza’s funding. They both had Wright-Patterson in their backgrounds. Eskridge said on the record that a directed-energy weapon was targeting her and would never kill herself, and then she killed herself. McCasland disappeared into a New Mexico canyon with a pistol and a wallet. Tim Burchett went on Tim Pool. Burlison called it foreign. Comer opened a House Oversight investigation. The FBI is now spearheading. The President said he hopes it’s a coincidence. The Daily Mail added Chinese scientists to the count. A site called End of the American Dream is now running thirteen Americans, nine Chinese, and a UFO researcher.

Here is what is actually true.

Loureiro was killed by a college classmate from Portugal who had nursed a grudge for thirty years and confessed on video. Grillmair was shot by his neighbor, a man with a documented history of erratic behavior and prior trespassing on Grillmair’s property. Hicks died of arteriosclerotic cardiovascular disease, confirmed by the LA County coroner, the same disease that kills nine hundred thousand Americans a year. Thomas had just lost both parents, and his wife told Dateline he was struggling. McCasland’s wife, the only person who actually knew him, told the police he had memory loss, anxiety, sleep problems, and that she suspected he had planned not to be found. Eskridge was a brilliant plasma physicist who became convinced, in the months before her death, that aliens were communicating with her telepathically and that the government was using directed-energy weapons against her. She killed herself in the presence of police officers who had been called because she was holding a weapon and threatening her own life.

These are not eleven cases of a coordinated assault on American scientific knowledge. These are eleven separate human tragedies, with mostly mundane explanations already documented in the public record, that an algorithmic media substrate has woven into a single narrative because that is what the substrate does. It takes noise and produces a pattern, because a pattern is engagement, and engagement is the only metric the substrate optimizes for.

Mick West, who has spent a career documenting this, ran the math. Roughly 700,000 Americans hold top-secret clearances in the nuclear and aerospace sectors. Statistical baselines on homicide and suicide alone predict that around two hundred and fifty of them would die by violence over a four-year window, thousands more from natural causes. Eleven is not a pattern. Eleven is a rounding error in the noise, and the noise is being sold to you as a signal because the people selling it to you make their money the same way the people who lied about the laptop made theirs. Engagement. Subscriptions. Ad reads. The mechanism is identical. The direction is reversed.

You were right that the licensed apparatus failed. You were wrong to think the shadow apparatus would not fail in its own way. They are running on the same substrate. The substrate is the problem.

So what is the substrate doing, exactly? And why is the answer to that question theological before it is technological?

Quantum mechanics is the place modern physics had to admit something Catholicism has known since the upper room. Reality at its base does not resolve into definite states except in the presence of an apparatus that interacts with it. Before measurement, a particle is a probability distribution. After measurement, it is a fact. The act of observation is not passive recording. The act of observation is constitutive. Werner Heisenberg said the line that should be carved over every server farm in Virginia: “What we observe is not nature itself, but nature exposed to our method of questioning.”

The man who first proposed the Big Bang was a Belgian Catholic priest named Georges Lemaitre. Not a Jesuit, not a monk, just a parish priest with a doctorate in theoretical physics, ordained for the Archdiocese of Mechelen-Brussels in 1923. He worked the math out in 1927, four years before Edwin Hubble published the telescope observations that confirmed the universe was expanding. Lemaitre saw it first on paper, because he was reading Einstein’s equations of general relativity more carefully than Einstein was. When Lemaitre showed the work to Einstein in person, Einstein famously told him: “Your calculations are correct, but your physics is abominable.” Einstein hated the implication. The math said the universe had a beginning. A specific moment of origin. A first instant, before which there was nothing in the way that physics measures somethings.

Einstein eventually conceded he had been wrong. The universe did have a beginning. The priest had been right. Today, we call it the Big Bang, a term that started as a mockery from a rival astronomer and stuck because it was catchy. The technical name Lemaitre preferred was “the primeval atom” or “the cosmic egg.” The Belgian priest had given physics its origin story.

This put him in an awkward position with his own Church. In 1951, Pope Pius XII gave a speech to the Pontifical Academy of Sciences with Lemaitre sitting in the audience. The Pope had been reading the cosmology. He saw what every Catholic reader sees the first time the Big Bang is explained: this looks like the opening sentence of Genesis. In the beginning, God created the heavens and the earth. The Hebrew word bara in Genesis 1, translated into Latin as creatio, the act of creation. Creatio ex nihilo, creation out of nothing, is the technical theological term for what Christians have always claimed God did. There was nothing, and then there was everything, and the difference between the two states was God’s free decision to bring the second into being. Pius looked at Lemaitre’s math and thought: my own scientists have just proved the first verse of the Bible.

So the Pope said so. In the 1951 address, he announced that contemporary cosmology had vindicated the Catholic doctrine of creation. The press loved it. The Vatican loved it. Almost everyone loved it.

Lemaitre hated it. He went, quietly but firmly, to the Pope’s science adviser and asked him to please get the Pope to stop doing this. Lemaitre argued that the Pope was making two mistakes at once. He was making bad theology, because the Catholic doctrine of creation does not depend on physics being a particular way and should not be hostage to whatever the next generation of physicists discovers. And he was doing bad physics, because if you turn a scientific theory into a proof of God, you set science up to be the enemy of God the next time the theory has to be revised. Lemaitre coined a term for this mistake. He called it concordism. The attempt to make scripture and science say the same thing in the same language. He thought it was a category error and that it would damage both fields.

The Pope listened. Pius mostly stopped using the Big Bang as a theological proof after that. Lemaitre, the priest who had discovered the most theologically suggestive scientific result of the twentieth century, spent the rest of his life insisting it should not be used theologically. He became President of the Pontifical Academy of Sciences in 1960 and continued serving until his death in 1966. To his last day, he refused to mix the two domains. He thought they could be in conversation, but not in fusion.

I tell you all this because the line in Pius’s 1951 speech that Lemaitre objected to is the line I want you to see. Pius said that the science of his day “has succeeded in being a witness to that primordial Fiat Lux.” Fiat Lux is the third sentence of the Bible, in Latin. And God said, “Let there be light. It is the first thing God speaks in the entire scriptural record. Pius was claiming that twentieth-century cosmology had become a witness to that original moment of speech and light.

Witness. The exact word.

Lemaitre objected to the speech. But neither he nor anyone else objected to that particular word, and I don’t think they noticed it the way I want you to notice it now. Because the word the Pope reached for, when he was trying to describe what physics was doing at its furthest extension, was the same word the Catechism reaches for when it describes what a martyr does in the moment of death. Martys in the original Greek. Witness. The bearing of testimony to a truth that exceeds the witness’s own life.

This is not a coincidence of vocabulary. The Catholic tradition has always used the same word for these two acts because it has always understood them as the same kind of act. To bear witness is to attest to a truth that does not depend on you for its existence, but that depends on you for its arrival into the shared world of human knowledge. A martyr does this with a life. A scientist does this with a measurement. A reporter does this with a story. A friend does this when they tell you what they saw. The witness function is what carries reality across the gap between what happened and what is known to have happened.

That is what we are losing. That is what AI is interfering with. And that is why a piece about missing scientists has to become a piece about Pontius Pilate, because Pilate is the patron saint of every apparatus that ever asked the question and walked out of the room before the answer arrived.

There is another priest you should know about. His name is Michal Heller. He is Polish, ninety years old as I write this, and still working. He is a cosmologist. His specific area of research is the unification of quantum mechanics with general relativity, which is the deepest unsolved problem in theoretical physics. The two great physical theories of the twentieth century, the one that describes the very small (quantum mechanics) and the one that describes the very large (general relativity, the geometry of space and time at cosmic scales), are mathematically incompatible. Both are confirmed to an extraordinary degree of precision in their own domains. Both fail catastrophically when you try to apply them to the place they ought to overlap, which is the first instant of the universe and the deep interior of black holes. Reconciling the two is the holy grail of physics. Heller has been working on it, using a branch of mathematics called noncommutative geometry, since before most of the readers of this piece were born.

He did most of his career under conditions that should make every American who complains about social media censorship feel ridiculous. He was born in 1936 in Tarnów, a Polish town that the Nazis occupied when he was three. His family was deported to Siberia during the Soviet period of the war. His father was harassed for years by the Communist authorities after the family was allowed to return to Poland. Heller studied physics at the Catholic University of Lublin, the only institution in Communist Poland where a Catholic priest was allowed to pursue advanced scientific degrees. The Polish state ensured the university could not award degrees specifically in physics. Hence, Heller had to pursue his doctorate in philosophy with a thesis on relativistic cosmology, a polite fiction that everyone agreed to to let his work continue. He could not get a passport to leave Poland for almost twenty years. He could not publish freely. He could not travel to international conferences. He was a priest doing world-class theoretical physics under a regime that was actively trying to forbid both his vocations.

What he did in those decades is what I want you to see. He gathered with other Polish scientists and theologians at the residence of the Archbishop of Krakow, Karol Wojtyla. They held informal seminars there throughout the 1960s and 1970s: physics, philosophy, theology, and the history of science. The state could not see what was happening inside an archbishop’s residence, so the apparatus of inquiry continued operating under cover. They called the group the Center for Interdisciplinary Studies. It was, in the most literal sense, a catacomb. A protected space, beneath the surface of an inhospitable regime, where the work the regime had forbidden continued. In 1978, Karol Wojtyla was elected Pope and took the name John Paul II. Heller’s catacomb suddenly had the Vatican’s protection. He won the Templeton Prize in 2008. He sits today on the Pontifical Academy of Sciences. He is a member of the Vatican Observatory.

Heller’s most-quoted line is this: “Science gives us knowledge, but religion gives us meaning. Science without religion is not meaningless, but lame. And religion without science slides into fundamentalism.” Lame, not meaningless. He chose the word carefully. A science without religion can still walk, but it walks with a limp, because it has cut itself off from the question of what its findings are for.

I bring up Lemaitre and Heller because the version of “quantum” that floats around in the shadow-media world usually goes somewhere I refuse to follow. You have heard it. Quantum entanglement proves consciousness creates reality. Quantum observation proves we live in a simulation. Quantum tunneling proves manifestation works. The wavefunction is the field of pure potential, and you can collapse it with intention. There is a film called What the Bleep Do We Know that did more damage to public understanding of physics than any other single artifact of the last twenty years, and its argument is essentially that quantum mechanics licenses any spiritual claim the speaker wants to make.

That is not what quantum mechanics says. That is not what Lemaitre said, not what Heller says, and not what any actual practicing physicist who also happens to be a Catholic priest has ever said. What they say, and what is genuinely there in the physics, is something narrower and more interesting. The universe at its base requires an apparatus to become specific. Before measurement, a quantum system is a probability distribution. After measurement, it is a fact. The apparatus you choose determines which facts you get. There is no view from nowhere. There is no measurement without an apparatus, and there is no apparatus that does not change what it measures.

That much is physics. The theology begins with the question of what kind of being could constitute a universe like that, and the Catholic answer is that the universe is held in being, moment by moment, by the continuous attention of its Creator. Saint Thomas Aquinas worked this out in the thirteenth century, six hundred years before quantum mechanics was a word. He called it creatio continua. Continuous creation. The doctrine that God did not just start the universe and walk away, the way a clockmaker winds a clock, but holds every particle in existence at every instant by an act of sustaining attention. Without that attention, the universe would not run down. It would simply stop being. The quantum measurement problem, in this reading, is what the universe looks like from the inside when it is being held in being by an attention that is not ours.

You do not have to believe that. The physics works either way. Heller would be the first to insist that you cannot prove God from the measurement problem, and you should not try. But the structural parallel is real. Reality at its base requires a witness. The Catholic tradition has been saying this for two thousand years. Quantum mechanics, in the twentieth century, was the place modern physics had to stop and admit it was saying it too.

Now hold all of that next to a single Greek word.

The word is martys. M-A-R-T-Y-S. It comes into English in two different ways. The first way gives us the word “witness,” as in someone who saw something and can tell you what they saw. The second way gives us the word “martyr,” as in someone who was killed for what they saw and refused to deny. In English, we treat these as two different things. A witness goes to court. A martyr goes to a grave. We have separate words because we live in a world where most witnessing does not get you killed.

The early Greek-speaking Christians did not have separate words. They had one word. Martys. They used the same word for both acts because they understood them as the same act under different conditions. To bear witness, in a world structured to suppress the truth you are bearing witness to, is to risk being killed for it. The risk is not extra. The risk is built into the act. A witness who would lie under sufficient pressure is not a witness. A witness who tells the truth when telling the truth is, by definition, a potential martyr. The Greek word fuses the two because the early Church understood that the witness function, in a fallen world, is always shadowed by the possibility of dying for what you saw.

The Catechism of the Catholic Church, the official compendium of Catholic teaching, lays this out with a directness that most modern readers find startling. I am going to quote four paragraphs in a row, because they are short and because I want you to see how plainly this is stated. Section 2471: “Before Pilate, Christ proclaims that he ‘has come into the world, to bear witness to the truth.’” Section 2472: “Witness is an act of justice that establishes the truth or makes it known.” Section 2473: “Martyrdom is the supreme witness given to the truth of the faith: it means bearing witness even unto death. The martyr bears witness to Christ who died and rose, to whom he is united by charity.” Section 2474: “The Church has painstakingly collected the records of those who persevered to the end in witnessing to their faith. These are the acts of the Martyrs. They form the archives of truth written in letters of blood.”

Read that last phrase again. Archives of truth written in letters of blood.

The Catholic Church has a two-thousand-year archive of names. Stephen was stoned in Jerusalem in the early 30s. Peter, crucified upside down in Rome under Nero. Paul was beheaded outside the city around the same time. Polycarp, burned alive in Smyrna in 155. Perpetua and Felicity were killed in the arena at Carthage in 203. The Forty Martyrs of Sebaste, frozen on a lake in Armenia in 320. The English martyrs under Elizabeth, the Japanese martyrs at Nagasaki, the Mexican Cristeros, the Spanish Civil War, the priests Hitler killed at Dachau, the priests the Communists killed across the entire bloc that Heller worked under, the catechists murdered in Latin America in the 1980s, the Coptic Christians whose throats were cut on a Libyan beach in 2015 while saying the name of Jesus. The Church kept the names—all of them. Writing them down was part of how the witness continued. The names are in the archive. The archive is the truth.

Now hold that practice, the two-thousand-year practice of keeping the archive of names, next to what is happening on your screen this week.

A woman named Rahile Dawut, a Uyghur anthropologist from Xinjiang who spent her career documenting the folk traditions of her people, was forcibly disappeared by the Chinese state in 2017 and has not been seen since. The United Nations raised her case again last year. The Chinese state news agency Xinhua cannot report on her case because the category does not exist in their information environment. She is not “missing” in any way the official Chinese press can acknowledge. She has simply been removed from the record. A man named Yu Wensheng, a human rights lawyer who defended people the state did not want defended, just completed a three-year prison sentence in April 2026 for a crime called “inciting subversion of state power.” A man named Peng Lifa, who hung two banners on a Beijing overpass in October 2022 protesting Xi Jinping, has been missing ever since. His whereabouts are unknown to his family and to the world. His name is also not in Xinhua.

The names are kept by an organization called Human Rights in China, founded in 1989 after Tiananmen, headquartered now in New York because it cannot operate inside China. It runs an account on X. Every week, the account posts the names of the disappeared. Yu Wensheng returned home. Mao Shanchun, sentenced to four years on April 7, his fifty-second birthday, spent in jail. Xu Zhiyong and Ding Jiaxi, three years into their 14- and 12-year sentences, respectively, marked the anniversary of their convictions on April 10. Huang Qi, who founded a website to expose corruption after the 2008 Wenchuan earthquake, spent his sixty-third birthday in prison this April. Peng Lifa, fifty-second birthday spent in unknown captivity, January 7. The list is endless. The list is updated weekly. The list is the archive.

The HRIC account on X is doing exactly what the Catechism describes. They are keeping the records of those who persevered in witnessing. They are an exiled apparatus, operating outside the walls of the regime that forbade the witness, and they are the only place certain names can be remembered at all. They are, in the most literal possible sense, a digital catacomb. The state forbade the witness, so the witness migrated outside the walls. This is not a metaphor. This is the same thing the early Church did under Rome, the same thing Heller did under Communism, the same thing the Polish underground press did under the Soviets, the same thing the Czech samizdat did under the same regime. When the apparatus on the surface forbids the witness, the witness goes underground, and the underground keeps the archive. The Church has known this pattern for two thousand years. We are watching it run again, in real time, in our own decade, and the most remarkable thing about it is how few people on the surface notice that it is the same pattern.

The American substrate is experiencing the opposite failure. It is not forbidding witness. It is mass-producing it. It is generating witnesses to McCasland’s UFO secrets, to Eskridge’s directed-energy weapon, to a Chinese-American hybrid conspiracy to murder rocket scientists, to a thousand other patterns that do not exist, because the substrate is calibrated to produce witnesses regardless of whether there is anything to bear witness to. In China, the witness function has been criminalized, and the disappeared are dissolved from the record. In America, the witness function has been algorithmically simulated, and the apparent record is full of testimony to events that did not happen. Both are failures of the same underlying capacity. The capacity by which a community establishes what is real together, and binds itself to that reality with enough seriousness that someone, somewhere, would be willing to die rather than recant it.

This is what the Pope is talking about when he talks about AI. And you should know what the Pope is saying, because I do not think most of you do.

The current Pope is named Leo XIV. He was elected in May 2025, the first American to hold the office, a Chicago-born Augustinian named Robert Prevost who spent most of his priesthood in Peru. The name he chose is the entire argument of his pontificate compressed into two syllables. The previous Pope Leo, Leo XIII, held the office from 1878 to 1903. He is the Pope who responded to the Industrial Revolution. In 1891, he wrote an encyclical titled Rerum Novarum, Latin for “of new things,” addressing the effects of factory work and mass urbanization on human beings. Leo XIII argued that workers had rights, that capital and labor were not naturally enemies but had to be reconciled by justice, that the state had obligations to the poor, and that unrestrained capitalism produced human suffering the Church could not ignore. Rerum Novarum is the founding document of modern Catholic social teaching. It is the reason the Catholic Church has, for over a century, said the things it says about labor and dignity and the limits of markets.

When Robert Prevost chose the name Leo XIV, he was telling everyone with ears to hear that he understood his pontificate as a response to a new industrial revolution. Not steam and steel this time. AI. He has said this directly, more than once, including in his first formal address to the College of Cardinals after his election. The Church, he said, “offers everyone the treasury of its social teaching in response to another industrial revolution and to developments in the field of artificial intelligence that pose new challenges for the defense of human dignity, justice, and labor.” That sentence is not casual. The man chose his throne name to signal that AI is to his pontificate what factory labor was to Leo XIII’s. He thinks this is the issue of his tenure. He thinks the Church has to meet it the way it met the last one.

In January 2026, Leo XIV released his Message for the World Day of Social Communications. This is an annual document the Pope writes for the Catholic press and, by extension, for anyone who works in media. The theme he chose is the line you should remember from the rest of this piece. He called the document “Preserving human voices and faces.” Three of his lines from that document are worth quoting in full.

First: “The challenge is not technological, but anthropological. Protecting faces and voices ultimately means protecting ourselves.” Anthropological means “having to do with what humans are.” The Pope says AI does not threaten our gadgets. It threatens the conditions of being a human person at all.

Second, he warned against “a naively uncritical reliance on artificial intelligence as an omniscient ‘friend,’ a dispenser of all information, an archive of all memory, an ‘oracle’ of all advice.” Read that line slowly. He is describing what most of you already do with ChatGPT, Claude, and Gemini—outsourcing your memory, your research, your moral reasoning, your taste, your friendships. He is naming the danger.

Third, and this is the one that should stop you: “This gives rise to serious concern regarding the oligopolistic control of algorithmic and artificial intelligence systems capable of subtly shaping behaviors and even rewriting human history, including the history of the Church, often without people being truly aware of it.”

Rewriting human history. Including the history of the Church. Often, without people being truly aware of it.

That is the Pope, in his official capacity, in January of this year, telling you that a small number of companies controlling the algorithmic substrate now have the power to rewrite the past. Not change the future. Rewrite the past. Reshape what humans believe happened, gradually, invisibly, until the new version is what everyone remembers and the old version cannot be retrieved. This is exactly what is happening to the missing-scientists story. It is what is happening to a hundred other stories. The substrate is producing a version of the past that is not the past, and presenting it to billions of people as if it were memory.

A year before Leo XIV’s message, the Vatican issued a document I want you to know about. It is called Antiqua et Nova, which is Latin for “ancient and new.” The title comes from a parable Jesus tells in Matthew 13, in which he describes a wise scribe who brings out of his treasure things “new and old.” The Vatican released the document on January 28, 2025, the day after International Holocaust Remembrance Day. The timing was deliberate. The document opens by saying that the atrocities committed throughout human history are reason enough to be deeply concerned about the potential abuses of AI. It is the official Catholic teaching document on artificial intelligence, the most authoritative thing the Church has yet said on the subject, 117 paragraphs covering education, work, healthcare, war, surveillance, and what the document calls “the crisis of truth.”

The crisis of truth. The Vatican’s exact phrase. They named the thing.

In summary, the document says what I have been telling you. AI is a tool. Tools can be used for good or for evil. This particular tool, deployed at this particular scale, threatens specific human capacities that the Catholic tradition has always protected. The capacity for a genuine relationship. The capacity for moral judgment. The capacity to bear witness to the truth. The document specifically warns about deepfakes, about AI hallucination, about workers being deskilled by algorithmic management, about surveillance, and about the manipulation of public opinion by a small number of companies controlling the substrate. It is the most clear-eyed institutional response to AI that any major religious or secular body has yet produced. It is freely available on the Vatican’s website. Almost none of you have read it.

So here is what the picture looks like from a distance.

The licensed apparatus has been failing for two decades, and you noticed. Good. You went looking for accounts the apparatus had not approved, and you found a parallel apparatus that was sometimes more honest and is now failing in its own direction, manufacturing witnesses to events that did not happen. Both failures stem from the same underlying problem: the substrate that delivers reality to all of us is calibrated for engagement rather than truth, and engagement and truth are not the same and have begun to come apart in your hands. Meanwhile, in China, the same technology in different political conditions is being used to erase people from the record entirely, and a small organization in New York is keeping the archive of names because no one inside China is allowed to. And the Pope, the actual living Pope, is telling you that what is happening to your epistemic life is the most important issue facing humanity right now. The institution he heads has spent two thousand years developing the only sustained tradition we have for thinking about what witness is, why it matters, and what it costs to preserve.

I called this piece “What is truth?” because that is what Pilate asked. He asked it of a prisoner who had just told him, on the record, that he had come into the world to bear witness to the truth. Pilate did not wait for the answer. He had a job to do. He washed his hands and signed the order. The official account that left his courtroom that afternoon was that a Galilean troublemaker had been duly tried and executed for sedition. That account had the full weight of Roman institutional authority behind it. It was the licensed truth of the most powerful empire in the ancient world. It was wrong. The witnesses the apparatus refused to credit, the women at the tomb whose testimony was inadmissible in Roman courts, the fishermen who had run away two days earlier and now would not stop talking, the disciples on a road to a town called Emmaus, kept telling a different story. The empire that had executed the prisoner spent the next three centuries trying to silence those witnesses. Some of them were beheaded. Some were crucified. Some were burned, or fed to animals in arenas, or buried alive. The names were kept. The archive was preserved. The witnesses kept witnessing.

In the meantime, the Christians who had not yet been martyred dug tunnels under Rome. Hundreds of miles of them. They buried their dead in the tunnels because Roman law forbade the burial practices their faith required, and they held their liturgies down there because the surface was too dangerous. They painted their iconography on the walls of those underground chambers because they could not put it on the walls of public buildings. The catacombs of Rome are still there. You can visit them. The empire that drove the Christians underground is gone. The tunnels remain.

I have been making a version of this argument in business publications for over a year, in pieces about why employees are walking out of jobs that won’t let them use the AI tools they prefer, why journalists are leaving institutional newsrooms for direct-subscriber relationships with their audiences, why the most talented people in every industry are quietly building parallel infrastructures because the surface has become inhospitable to the work they want to do. I have called it shadow culture. I have argued repeatedly that this is not a workplace problem, not a generational problem, and not a technology problem. It is the same pattern that has played out every time an institutional surface has compressed humans hard enough that they had to go underground to keep doing the things humans need to do. The Church under Rome. The Reformers against the Index. The samizdat under the Soviets. Heller in his archbishop’s residence. The Polish workers who built Solidarity. The Chinese dissidents who built HRIC.

What I have not been saying out loud, because business publications are not the right venue for it, is that the deepest version of this argument is theological, and that the institution that knows the most about the witness function has known the most about it for two thousand years, and that the man we are discussing in the present tense, the man for whom the Catechism’s section on witness was eventually written, is the prisoner Pilate executed.

He is the patron, in the Catholic understanding, of every witness who is asked to recant and refuses. Of every dissident the apparatus tried to dissolve. Of every disappeared scientist whose name a state forbade. Of every disappeared anthropologist in a Xinjiang detention center. Of every woman whose testimony was inadmissible in the courts that mattered. Of every shadow chronicler keeping an archive of truth in conditions hostile to its preservation. He is also the patron of those of you who are not Catholic, who may never become Catholic, and who are reading this piece because you can sense that something is wrong with your information environment. You are looking for someone willing to tell you the truth about what is wrong. He is the patron because he was the first one who refused to go quiet when the apparatus told him to, and the apparatus killed him for it, and the apparatus did not get the last word.

The tunnels are already longer than the apparatus thinks. The archive is already larger. The witnesses are already at work. The Pope is paying attention. The Vatican has named the crisis. The question is whether you are going to keep pretending that the surface is the only place reality happens, or whether you are going to find the people who have been keeping the archive all along.

The Greek word is martys. It means witness. It means martyr. It is the same word.

What is truth? The man Pilate asked is still answering. The witnesses are still bearing witness. The archives are still being kept.

Choose which one you want to be.

On Orthodox Easter Sunday, April 12, 2026, the President of the United States posted an AI-generated image of himself as Jesus Christ.

The image shows Trump standing in a white robe with a red sash, his right hand placed on the forehead of a man lying in a hospital bed. Light comes off the hand. Behind them are a nurse, a uniformed soldier, a waving American flag, bald eagles in flight, the Statue of Liberty, and fighter jets. The composition is a healing icon. It is the visual grammar of every Christ-heals-the-sick painting in Western art, restaged with an American flag instead of a temple and an Air Force jet instead of a dove. A generative model trained on the entire corpus of Christian art knows exactly what that composition is, and it produced one on command. Trump posted it.

When challenged, Trump said he thought he was depicting himself as a Red Cross doctor. “Only the fake news could come up with that one,” he told reporters. “It’s supposed to be me as a doctor, making people better. And I do make people better. I make people a lot better.” White House Press Secretary Caroline Leavitt called the image “doctored,” using the word in the sense of altered. It was not altered. It was generated and had no prior image to be altered from. Trump reposted it knowing what it was, and when the image produced the reaction it was always going to produce, he denied what the machine had unambiguously produced.

Three days earlier, on April 10, the first American Pope, Leo XIV, had written: God does not bless any conflict. Anyone who is a disciple of Christ, the Prince of Peace, is never on the side of those who once wielded the sword and today drop bombs. He was speaking to Chaldean bishops in Rome, from a continent where his own parishioners are being bombed, and he did not mention Trump, because he did not have to. Trump took it as an attack, issued a long Truth Social diatribe calling the Pope weak on crime and terrible for foreign policy, and then, on Orthodox Easter, posted the image.

The President used a generative AI model to depict himself as the Prince of Peace, on Orthodox Easter, in retaliation against the Pope, who had just said that those who drop bombs are not on the side of the Prince of Peace.

I want to explain what just happened, because the piece most of the coverage has missed is that this Pope, specifically, is the Pope whose entire papacy is framed around exactly this machine. Leo XIV is not incidentally critical of AI. He chose his papal name for this reason. The administration did not accidentally reach for the one blasphemy most pointed at this particular Pope; it reached for exactly that one. The machine they used is the machine he took office to oppose. That is not a coincidence. It is the story.

The Pope Who Named Himself for AI

Leo XIV is named Leo because of artificial intelligence. In his first address to the College of Cardinals, on May 10, 2025, two days after his election, he explained the choice: “In our own day, the Church offers to everyone the treasury of her social teaching in response to another industrial revolution and to developments in the field of artificial intelligence that pose new challenges for the defense of human dignity, justice, and labor.”

His namesake, Leo XIII, wrote Rerum Novarum in 1891 as the Church’s response to industrialization. Leo XIV has spent the year since his election signaling, in every major address, that his pontificate is the Church’s response to AI. He has told a youth conference in Indianapolis that AI “can process information quickly, but it cannot offer real wisdom.” He has told the priests of the Diocese of Rome not to use AI to write their homilies. He has warned political leaders that “our personal life has greater value than any algorithm, and social relationships require spaces for development that far transcend the limited patterns that any soulless machine can pre-package for us.” The Vatican’s thirty-page document Antiqua et Nova, released in January 2025, sets out the theological case: AI operates through pattern recognition and lacks the creative, spiritual, and moral dimensions of human thought. Paolo Carozza, President of the Pontifical Academy of Social Sciences, has described Leo’s concern as the worry that technologies like AI are “very quickly displacing fundamental human capacities of reason, freedom, judgment, discretion and choice, our understanding of truth and falsity.” Not replacing them. Displacing them.

This is the Pope Trump attacked, with a machine, on Easter.

And this is not the first time Trump has used AI to stage himself as a figure of Christian authority. In May 2025, just before the conclave that would elect Leo, Trump posted an AI image of himself dressed as the Pope. That image is still on Truth Social. He has had the habit for a year. On Wednesday, April 15, after the Easter image was deleted, Trump reposted a second AI-generated image: Jesus embracing him at a podium in front of an American flag, with a caption suggesting that God might be “playing his Trump card” by elevating Trump to expose “satanic, demonic, child-sacrificing monsters.” He did not apologize; he escalated, using the same machine to produce a second theological claim. On April 13, the Iranian Embassy in Tajikistan posted an AI video of Jesus punching Trump in the face. The generative machine has become a medium of transnational theological combat. Everyone reaches for it. Theology is the payload.

I want to be exact about what has happened. The President of the United States is using AI-generated religious imagery, at a rate the secular press has not yet metabolized, to stage himself as the central figure of Christian salvation history. He is doing this while attacking the Pope, who picked his name to warn about this machine. On Easter, the President used that machine to depict himself as the savior the Pope had just said is never on the side of those who drop bombs, in the middle of a war the Pope has called unjust. When caught, he said the machine was depicting a Red Cross worker.

To understand why this is not a gaffe but the entire argument, you need one idea from Thomas Aquinas.

Aquinas and the Christ-Shaped Hole

In the Summa Theologica, Part I, Question 48, Aquinas argues that evil is not a thing. Evil is a privation — a lack, not a substance that exists alongside good, but the absence of a good that ought to be there. Cold is the absence of heat. Darkness is the absence of light. Blindness is the privation of sight in an eye that ought to see. Sickness is the privation of health. Evil is the absence of a good that should be present. “Evil is distant both from simple being and from simple not-being,” Aquinas wrote, “because it is neither a habit nor a pure negation, but a privation.” Augustine had made the move a thousand years earlier, in the Confessions, and C. S. Lewis translated it for modern readers in Mere Christianity: “Goodness is, so to speak, itself: badness is only spoiled goodness. And there must be something good first before it can be spoiled.” Evil is a parasite. It cannot generate; it can only corrupt something that is.

This is the tradition Leo XIV is Pope of. It is the tradition Vance claims to be inside. It is the tradition that the Catechism of the Catholic Church teaches, in paragraph 2309, when it codifies the just war doctrine. And it is the tradition that tells you, with precision, what the Easter image is.

The image is a privation of an icon.

A real icon of a healing Christ is not a painting. It is, in the Orthodox theological tradition Trump posted over, a window — the icon depicts the saint or Christ not as a likeness but as a presence made visible to the viewer. The iconographer does not sign. The iconographer fasts, prays, and writes the icon inside a tradition that places authorship in the communion of the Church across centuries. A real icon has, constitutively, an author, a tradition, a discipline, and a theological claim the iconographer is answerable to. None of those things is optional, and removing any one of them does not produce a diminished icon. It is to produce an icon-shaped privation of an icon.

The generative image has every one of those things removed. There is no author, because the model is the composite ghost of billions of images; no single painter decided to place the hand there, to angle the light that way, to position the flag behind. No one can be asked to account for any specific element. There is no tradition, because the model has no communion — it has a dataset. There is no discipline, because the model does not fast; it samples. And there is no theological claim the producer is answerable to, because the producer is not the model but the poster, and the poster, when asked what claim he had made, said he was posting a Red Cross worker.

Every constitutive element of the icon, in other words, has been removed, and what remains is the composition alone. The visual surface of a healing Christ. The specific grammar the viewer’s eye recognizes. The halo is absent because the composition is the halo. A generative privation of the icon, produced at speed, deployed on Easter, denied on Monday.

This is what Pope Leo is naming when he says AI “cannot offer real wisdom.” He does not mean the model lacks IQ. He means that the model cannot participate in the tradition it is imitating, and therefore its outputs, when deployed in the tradition’s forms, are constitutionally hollow. A homily written by ChatGPT is not a shorter homily; it is a homily-shaped absence of a homily. An image generated to look like Christ healing is not a diminished icon; it is a Christ-shaped absence of Christ. The model is very good at producing these absences because that is what it was trained to do. It is not malfunctioning when it produces a blasphemy; it is functioning exactly as designed. The designer did not intend blasphemy. The designer intended a machine that reliably produces plausible surfaces. The surface is all the machine has. The substance has to come from somewhere else.

Now you see why Trump reaches for it. Not because he knows any of the above, but because generative AI produces the only kind of theological claim an administration like his can make without exposure — a claim with no author, no tradition, no discipline, no answerability. A claim you can post, retract, deny, and reissue, because nobody, including you, can be pinned to any of its specific elements. The Red Cross worker defense is not a retreat from the image; it is the feature for which the image was deployed. Generative artifacts launder theological claims through the deniability of their production. That is why the administration is reaching for them again and again. The Pope image in May, the Jesus image on Easter, and the Jesus-embracing-Trump image three days later are not three gaffes. They are the operation.

The Prayer Was Generated Too

Pete Hegseth, at a Pentagon worship service on April 15, prayed a prayer that had been lifted from Pulp Fiction. The first half of the prayer is the movie monologue — the one Samuel L. Jackson delivers before he shoots somebody — with “the path of the righteous man” swapped for “the path of the downed aviator.” The last two sentences borrow from Ezekiel 25:17, with one edit. In Ezekiel, the speaker of the final line is the Lord. In Hegseth’s version, the speaker is “Sandy 1,” the radio call sign of the A-10 combat search and rescue aircraft that pulled a downed US airman out of Iran. The religion journalist Brian Kaylor of Word and Way caught it.

I want you to see that this prayer is doing the same operation as the image. The prayer has the surface of scripture and none of the substance. It was almost certainly not produced by a large language model — it was written by a human, perhaps a chaplain, perhaps a staffer. The mechanism does not matter; the logic is the machine’s logic. Take a sacred text. Remove the Lord from the sentence. Substitute a military call sign. Keep the cadence, keep the vocabulary, keep the authority the original had over the listener’s ear. The prayer has been hollowed out. What it prays for — great vengeance delivered on behalf of a military unit — is the opposite of what Ezekiel prays for. Yet the cadence is preserved so faithfully that the congregation said amen. The Pentagon spokesman, when asked, said anyone who noticed the substitution was “peddling fake news and ignorant of reality.”

This is what I mean when I say the theology and the AI stack are running the same operation. A large language model is, at a mechanical level, a substitution engine. Token by token, it replaces. It produces the surface of meaning without the substance, because it does not know what a word means, only what word usually follows. Hegseth’s prayer is that the operation performed by a human who has learned to think the way the machine operates. The image is that operation performed by the machine for a human who has learned to deploy its output. Same logic, same product: an authored surface with the author removed, used for purposes the real author would have refused.

Vance and the Inversion of Aquinas

The third instance of the same move happened on Fox News and at a Turning Point USA event in Athens, Georgia, on April 14. JD Vance, a Catholic convert of seven years and the second Catholic Vice President in US history, told Bret Baier that “it would be best for the Vatican to stick to matters of morality” and let the President dictate American public policy. Later that night at the Akins Ford Arena, in a room Morning Joe described as about a quarter full, Vance told the audience the Pope should “be careful when he talks about matters of theology.” A young man shouted from the back that the US was bombing children. Vance told the heckler he would get to his point. He did not get to it.

Here is what Vance was doing, and what the generous readings have been dancing around. Vance does not have a shallow grasp of the just war tradition. He has the inversion of it.

In Summa Theologiae II-II, Question 40, Aquinas established the just war framework that the Catechism now codifies: legitimate authority, just cause, right intention. Augustine, a thousand years earlier, had set the tone in City of God, Book XIX: “the wise man will wage just wars. As if he would not rather lament the necessity of just wars, if he remembers that he is a man.” Earlier in the same work, Book IV: “to carry on war and extend a kingdom over wholly subdued nations seems to bad men to be felicity, to good men necessity.” Not good, necessity. A privation of peace. Peter Lombard, writing a century before Aquinas, held that a soldier who killed in a just war still owed penance, because the act remained a privation even when permitted.

Vance, speaking on the TPUSA stage, said: “Was God on the side of the Americans who liberated France from the Nazis? I certainly think the answer is yes.” He said taking away Iran’s nuclear capability means innocent lives will be saved. He spoke, in the register a seventeenth-century Protestant chaplain might have used, as if the Iran war were a positive good, righteous, blessed, on God’s side.

That is not what the tradition says. The tradition says even a just war is a necessity, a privation of peace, a thing whose combatants owe penance. Vance has kept the tradition’s vocabulary while inverting its content. The Bishops had to say this out loud, in writing, on April 15. Auxiliary Bishop James Massa of Brooklyn, chairman of the USCCB Committee on Doctrine, wrote: “When Pope Leo speaks as supreme pastor of the universal Church, he is not merely offering opinions on theology, he is preaching the Gospel and exercising his ministry as the Vicar of Christ.” Massa cited the Catechism at paragraph 2309, which opens: “the strict conditions for legitimate defense by military force require rigorous consideration.” Strict consideration, rigorous conditions — a checklist, the tradition was designed specifically because it wanted to find a way to say yes and still could not say yes lightly.

Run the Iran war through that checklist, and the war scores zero. There was no legitimate authority; Congress never declared war. It was not a last resort; airstrikes began during active negotiations. Proportionality collapsed when Trump threatened, on Truth Social on April 7, that “a whole civilization will die tonight, never to be brought back again,” a sentence Amnesty International called a potential threat of genocide. Right intention failed through six distinct public justifications within the same administration. Reasonable chance of success failed when the administration, losing, picked a public fight with a seventy-year-old cleric to change the subject. Zero out of eight.

Elizabeth Anscombe saw this move, in different clothes, seventy years ago. Her pamphlet, Mr. Truman’s Degree, written in 1956 to oppose Oxford’s award of an honorary degree to the man who ordered the bombing of Hiroshima, is the most relevant document in modern Catholic moral philosophy to what is happening now. Anscombe argued that the root problem of Hiroshima was not the bomb; it was the unlimited objective. The Allied demand for unconditional surrender had produced a situation in which Truman could always claim that if only the Japanese had done what he demanded, he would not have been required to bomb them. The unlimited objective, Anscombe saw, is what licenses atrocity. Trump’s stated objectives in Iran have been unlimited from the beginning — regime change, destruction of missile capacity, civilization-ending threats if Iran does not capitulate on the Strait of Hormuz. When your adversary resists unlimited demands, you can always blame them for what comes next. That is the move Anscombe named, and she had a different word for what it produces when combined with attacks on civilians. She used it about Truman. I do not need to retype it here.

Anscombe also observed, in Intention and War and Murder, that the twentieth century had developed the habit of disguising intentional killing of civilians with consequentialist language. “The targeted elimination will result in significant collateral damage” is a way of saying “bombing our target is going to kill innocent people.” The words change; the act does not. This is exactly the substitution the administration is running, at every level. Trump’s image substitutes Christ for president. Hegseth’s prayer substitutes the Lord for a call sign. Vance’s theology substitutes Aquinas for a slogan. Language is doing the work of killing attention, in order to make killing people easier.

The Christians Being Bombed While Vance Talks

One of the facts the administration’s framing has to route around is that there are Christians inside Iran. Not abstract ones. Real ones, with parishes and bishops and addresses, located beneath the ordnance.

Around 20,000 Catholics, according to US State Department estimates, mostly Armenian and Chaldean, with about 2,000 Latin-rite Catholics led until recently by Cardinal Dominique Mathieu of Tehran, who has fled to Rome because his cathedral was closed. The Armenian Church and the Assyrian Church of the East have recognized status in Iran. Beyond them is an underground church, mostly converts from Islam, which, under Iranian law, is apostasy. Open Doors surveys have placed the Christian population as high as 1.5 percent of the country — more than 1.3 million people, if you count the ones the state refuses to count. Human rights groups documented 254 arrests of Christians in Iran in 2025, double the previous year. That was before the bombs started falling.

Cardinal Sako, the Chaldean Patriarch in Baghdad, has told Vatican News he fears for the 50,000 Christians on the Nineveh Plains. The Iraqi Christian population has fallen from 1.3 million in 2003 to 150,000 today, a ninety percent collapse across two US wars. The Chaldean bishops Leo was addressing when he said God does not bless any conflict were speaking to him from inside the blast radius.

This is the context in which Vance told the Pope to stick to matters of morality, in which Trump accused the Pope of wanting Iran to have a nuclear weapon, and in which the President of the United States posted an AI image of himself as the savior of the exact people Leo was trying to protect. The Pope is the bishop of the Armenian priest in Tehran, the Chaldean bishop in Erbil, the convert in Isfahan who cannot say her name out loud. On Easter, the President generated an image of himself as her healer.

Lewis Saw the Conditioners Coming

In 1943, C. S. Lewis delivered three lectures at the University of Durham that he later published as The Abolition of Man. The book is usually read as a defense of objective moral value against mid-century subjectivism. It is also eighty-three years early, the single best description we have of the machine the Trump administration reached for on Easter.

Lewis’s argument, in brief: once a society surrenders belief in objective value — what he called the Tao, the common moral inheritance of Eastern and Western traditions alike a class of people will arise who position themselves outside it. They will produce values rather than discover them. They will condition other people into whatever shape they prefer. Lewis called them Conditioners. “The man-molders of the new age will be armed with the powers of an omnicompetent state and an irresistible scientific technique: we shall at last get a race of conditioners who really can cut out all posterity in what shape they please.”

Omnicompetent state. Irresistible scientific technique. Read that phrase twice. Lewis wrote it in 1943, before the transistor, before the computer, before the internet, before a photograph could be faked by anything more sophisticated than a darkroom. It is the one-line description of the 2026 alliance among the American administrative state, the commercial generative AI industry, and a theology hollowed out to serve both.

Lewis’s punchline is the one I want to end on. “It is not that they are bad men. They are not men at all. Stepping outside the Tao, they have stepped into the void. Nor are their subjects necessarily unhappy men. They are not men at all: they are artifacts. Man’s final conquest has proved to be the abolition of Man.”

Read the Easter image again. A man in white robes, composed by a machine, posted by a president, denied as a doctor, deployed against a Pope, on a day sacred to a tradition the man posting has never been inside. Every element of it, the robes, the light, the hand, the flag, the jet, the eagle, is an artifact in Lewis’s sense. Seventy-three years early, he gave us the word. The image is not a failed icon; it is a successful artifact, which is different. An artifact is made to do a job. This one did its job. It substituted the President for Christ in the eye of a viewer scrolling at speed, and when the substitution was named, it retreated into generative deniability. It was a doctor. It was a Red Cross worker. It was a joke. It was not meant that way. Pick one.

Trump did not invent this, and the machine did not invent this. The alliance between them is what is new. The commercial AI industry, as currently configured, is in the business of producing surfaces without substance at scale. The administration, as currently configured, is in the business of needing such surfaces. The administration’s war cannot be defended on the merits, the just war checklist produces zero out of eight. The administration’s theology cannot be defended on the merits — its Vice President inverts Aquinas on television. So the administration reaches for the machine that can produce theological-looking artifacts with plausible deniability built into their production. The machine produces, the administration deploys, the critic objects, the administration denies. Repeat.

Pope Leo has named this. He named it on his second day in office. He named it when he told priests not to use AI to write homilies. He named it when he said AI cannot offer real wisdom. He has been naming it, in public, in homilies, in addresses, for a year. On April 10, he named it again: God does not bless any conflict. Two days later, the administration posted an AI image to prove him wrong. Leo’s entire papacy, by name and by doctrine, is a rebuke of the exact machine used to rebuke him. They cannot possibly have missed that. They chose it.

What the Audience Is Being Asked to Lack

Here is the part I want you to hold onto.

Every substitution the administration has performed this week depends on a prior privation, a privation in the viewer, in the listener, in the scroller, in the electorate. The whole operation only works if the audience has lost the capacity to notice a substitution in progress. Lewis called the missing thing a chest. “We make men without chests and expect of them virtue and enterprise.” We make viewers without attention and expect of them the recognition that the image on their phone, composed to the last pixel in the grammar of a healing icon, is not a Red Cross worker.

The commercial AI industry, as currently configured, is in the business of removing the friction. Fewer prompts, faster outputs, less review, more confident one-sentence answers to two-thousand-year questions. Every friction removed is a noticing the audience will not perform. The administration has learned to deploy the industry’s outputs for exactly this reason. A president cannot plausibly claim, on his own signature, that he is a Christlike healer of the American sick. He can, however, generate that claim, post it, and then deny he generated what the machine obviously generated, because the machine provides a deniability layer no human author ever could. The machine is the cover. The machine is the escape hatch. The machine is the plausible-enough-to-defend, deniable-enough-to-retract production of theological claims on command. That is why the administration reaches for it, and that is why they will keep reaching for it.

I build these machines for a living. I am not telling you to stop using AI. I am telling you that what happened on Easter is a use case the industry has not named, and Pope Leo has. The use case is this: generative AI allows public figures to make claims, including theological claims, that they could not otherwise make without exposure, because the machine authors the surface and the human authors only the post. The President of the United States posted an AI image of himself as Jesus on the holiest day of the Orthodox calendar, in retaliation against the Pope, who had criticized his war, and then denied the image meant what it obviously meant. No previous technology would have let him do this. The machine is new, the operation is new, and the one major institution on earth that has, on the record, with its namesake choice, organized itself around opposing this operation is the Catholic Church under Leo XIV. He is the figure the administration attacked, with the machine, on Easter. That is not a coincidence. That is the story.

Hold the attention. That is what the Pope is asking. That is what Lewis asked eighty years ago when he called it a chest. That is what Augustine asked when he said even a just war is grief. That is what Anscombe asked when she refused to let Truman’s honorary degree pass without a name. That is what Aquinas asked when he told us evil is a privation, and to look for what is missing from the thing in front of you.

The image is what is missing from the image. No author, no tradition, no theology, no answerability, a Christ-shaped hole into which a president inserted himself on Easter and then pretended he had not. Do not let it pass. Do not scroll past it. Do not let the machine’s deniability do its job. Say what it is. Say it slowly. Say it with citations. That is the whole job.

Leo is doing it. The Bishops are doing it. Anscombe did it in 1956. Augustine did it in 410. Aquinas did it in 1270. Lewis did it in 1943, before any of the machinery existed. The capacity is not extinct; it is being eroded on purpose, by an alliance of omnicompetent state and irresistible scientific technique that Lewis named before any of these actors were born.

Do not let them have it. That is what the Pope is asking. That is what I am asking, as someone who builds the machines. The machine cannot counterfeit the noticing. It is the one thing we still own.

Keep it.

We were in London recently, getting our son Bruno ready for university in the fall. Between flat viewings and orientation packets, I did what I always do when I’m there: tried to weave myself into the fabric of the city. I read British writers. I wandered neighborhoods. On Sundays, I like going to Mass at St Etheldreda’s in Holborn, a thirteenth-century church tucked behind a row of Georgian townhouses on Ely Place, one of only two buildings still standing in London from the reign of Edward I. It survived the Great Fire of 1666 when the wind changed at its walls. Shakespeare set scenes at Ely Place. There is a relic of the saint’s uncorrupted hand in a jeweled casket to the right of the altar. Catholic masses were banned there in 1534. The building was used as a prison and hospital in 1642. It switched from Anglican to Catholic in 1874. The tradition changed and changed again, but the worship never fully stopped. You sit in that space and seven hundred years of prayer press against you like weather.

One of my friends in London volunteers with a local youth football club. Coaching, organizing, the ordinary labor of showing up for other people’s kids. A committed atheist, not the performative kind but the settled kind, the kind who arrived at the position through reading and thinking and doesn’t need to relitigate it. When I asked once why they do this work, the answer was something I’ve been turning over for months: “It’s just part of my personal ethic. I care about people. You don’t need God for that.”

They’re right. You don’t need God to care about people. Empathy is real. It exists in primates. It predates organized religion by a geological epoch. Mirror neurons fire when we see someone in pain, and that happens whether you pray or don’t. My friend on the touchline on Saturday morning and I in the pew on Sunday morning are both trying to be good. The question is where the impulse comes from.

But I want to draw a distinction that I think matters, and that I think has consequences for something much larger than theology.

The feeling is biological. The obligation is cultural.

The gut-level pull toward helping a stranger is wired into us. But the sense that you should help, that it would be wrong not to, that the stranger’s suffering makes a claim on you, that you are somehow diminished if you walk past, that is not a primate instinct. That is a moral framework. And in the West, that framework has a specific origin.

Before I say what that origin is, let me say something else first, because I know exactly who’s reading this and what they’re already thinking.

Christianity has done terrible things. Not minor things. Not things you can wave away with context or nuance. The Inquisition tortured and executed people for the contents of their minds. The Crusades were military campaigns wrapped in theological justification. The systematic sexual abuse of children by clergy represents a betrayal so complete it resists language. The forced conversion of indigenous peoples across the Americas destroyed cultures and lives on a scale that should make anyone with a conscience physically ill. The justification of slavery using Scripture. The pogroms. The centuries of anti-Semitism that laid the cultural groundwork for the Holocaust.

This is not a footnote. This is not a “to be sure” paragraph that I insert so I can move on to the real point. This is the point, or at least half of it. Any honest accounting of Christianity has to hold the hospital and the torture chamber in the same hand. If you can’t do that, you’re not doing history. You’re doing marketing.

But here’s the other half: you can despise what Christianity did and still recognize what Christianity built. And what Christianity built is the moral infrastructure of the civilization you’re reading this in, including the moral categories you’d use to condemn it.

Tom Holland is a British historian, one of the writers I was reading during those London weeks, who spent years writing about the classical world before something began to bother him. He noticed that the values he assumed were universal, the ones he’d absorbed as a modern, liberal, secular Englishman, were not present in the civilizations he studied. The Romans he admired did not believe in the inherent dignity of every person. They watched people die in the arena for entertainment, exposed unwanted infants as routine, and built their economy on the backs of human beings they regarded as property. And nobody, not one major philosophical school, argued that this was structurally wrong.

What changed was not biology. What changed was the story a civilization told itself about what suffering means and who matters.

Holland published Dominion: How the Christian Revolution Remade the World in 2019. Its argument is not that Christianity is true, or that Christians are good, or that atheists should go to church. It argues that the moral grammar of the West, the concepts we reach for when we talk about human rights, the dignity of the individual, the moral primacy of victims over victors, care for the vulnerable as a measure of a society’s worth, are not self-evident truths that any rational person would arrive at independently. They are the products of a specific tradition. And that tradition is Christianity.

Holland, approaching his argument as a historian rather than a theologian, put it bluntly: even the most secular movements and ideas originating in the West, including humanism, liberalism, progress, and human rights, are fundamentally rooted in Christian principles. The very concept of secularism, the separation of the sacred from the political, is itself something the Western world inherited from Christianity.

Nietzsche understood this a century before Holland. He saw the European intelligentsia rejecting Christianity while keeping its morality, and he recognized the incoherence. You cannot kill God and keep his ethics. If there is no transcendent basis for the claim that every human being has equal worth, then that claim is just a preference, and a historically unusual one at that. Nietzsche had the intellectual courage to follow this to its conclusion. The New Atheists, largely, have not. They assume secular humanism is the default position, the natural condition of the reasonable mind once religious error is scraped away. But it isn’t. It’s an inheritance. The fact that it feels obvious is evidence not of its universality but of how thoroughly the Christian revolution succeeded.

Francis Spufford, the English writer and Fellow of the Royal Society of Literature, a man who found his way back to faith and whose book Unapologetic is one of the more intellectually honest defenses of Christianity written in this century, captured this perfectly. Spufford is not a theologian. He’s a literary man, a professor at Goldsmiths in London, whose other books include the Costa Award-winning novel Golden Hill and a study of Soviet economic planning, Red Plenty, that Dwight Garner named one of the ten best books of 2012. He’s not someone you can dismiss as a pulpit thumper. And in Unapologetic, he observed that Christianity’s influence on secular morality has “faded indistinguishably into the background of our common sense.” The humanist doesn’t trace their care for others to Jesus because the tracing has already been done, over centuries, and the origin has been erased by familiarity.

The fish doesn’t feel the water.

Now ask yourself: what about a machine trained entirely on what the fish wrote?

Every major foundation model, GPT, Claude, Gemini, Llama, is trained on a corpus that is overwhelmingly English-language, Western, post-Christian text. The internet is not a neutral archive. It is the written record of a civilization that processed its moral intuitions through a Christian framework for two thousand years. The model doesn’t know Jesus. The training data does.

This is not a metaphor. It is a structural observation about what these systems are and how they work.

A large language model is a statistical compression of its training data. It learns patterns, relationships, and frequencies. It learns that certain words follow other words, that certain ideas cluster with other ideas, that when humans write about ethics, they tend to invoke a specific set of concepts: dignity, rights, equality, compassion, the wrongness of cruelty, and the obligation to help the vulnerable. The model doesn’t understand these concepts. It reproduces them. And what it reproduces is the moral vocabulary of Christendom, secularized, stripped of attribution, and presented as if it were simply the way reasonable beings think.

Then comes the alignment layer. Reinforcement Learning from Human Feedback, or RLHF, is the process by which models are trained to give responses that humans judge as “helpful, harmless, and honest.” Human annotators read pairs of responses and choose the better one. The model learns from their choices.

But the annotators are humans. Humans are shaped by their culture. And when they choose the “less harmful” response, they’re exercising moral judgment that was formed in them by the same Christian-saturated civilization that produced the training data. They don’t need to be Christians. The moral reflex is already there. “Harmless” is doing all the theological work. What counts as harm, who counts as a person who can be harmed, whether harm to the weak is worse than harm to the powerful, these are not neutral questions. They are questions that Christianity answered in a specific way, and those answers became the default settings of Western moral reasoning.

Anthropic’s own constitution for Claude, the set of principles used in its training, instructs that the model should be “wise, peaceful, and ethical” and should avoid responses that are “toxic, racist, or sexist.” These are presented as universal values. But “wise, peaceful, and ethical” is not the moral vocabulary of Homeric Greece, where glory in battle was the highest human achievement. It is not the moral vocabulary of Imperial Rome, where mercy was a strategic calculation, not a virtue. It is not the moral vocabulary of the Aztec Empire, where human sacrifice was a civic and religious obligation. It is Sermon on the Mount morality, secularized and operationalized, with the serial numbers filed off.

I want to be careful here, because there is a counterargument that deserves genuine respect.

The atheist will say, “Confucius taught compassion. The Buddha taught non-attachment to suffering. The Stoics taught equanimity. You’re committing a provincial error by acting as if Christianity invented ethics.”

And they’re partially right. Ethical thinking is not a Christian monopoly. The Tamil philosopher Valluvar wrote a non-denominational treatise on secular ethics, emphasizing non-violence and human brotherhood, potentially centuries before Christianity arrived in South Asia. Epicurean philosophy articulated a coherent ethical system based on pleasure, friendship, and the avoidance of harm. Confucian virtue ethics produced a remarkably sophisticated social morality.

I don’t deny any of this. I distinguish.

These traditions developed ethical systems. What they did not develop, and what Christianity did develop, is the specific architecture of universal human rights, the inherent dignity of every individual regardless of status or capacity, and the moral primacy of love as the supreme ethical category. The scholar Brian Tierney demonstrated that the concept of universal human rights was developed by Christian canon lawyers in the twelfth century, rooted in the Genesis claim that every person is made in the image of God. The historian Kyle Harper showed that the idea that every person has a right to their own body, that sex must be consensual, entered the world through Christianity. These aren’t minor additions to a pre-existing ethical toolkit. They are the operating system.

And the operating system is what the models learned.

In January 2025, the Vatican published a document titled Antiqua et Nova: A Note on the Relationship Between Artificial Intelligence and Human Intelligence. It was issued jointly by the Dicastery for the Doctrine of the Faith and the Dicastery for Culture and Education, and it is a more serious document than most people will bother to read.

The document makes a precise anthropological argument. It distinguishes between AI’s ability to perform tasks and the human ability to think, and it insists that this distinction is “crucially important.” AI, it argues, should not be understood as an artificial form of human intelligence but as a product of it. This sounds like a theological nicety, but it has consequences.

If AI is a product of human intelligence, and if human intelligence (in the Western tradition) has been shaped by a Christian understanding of reason, embodiment, relationality, and orientation toward truth, then the outputs of AI carry the imprint of that tradition. The document implicitly acknowledges this when it observes that technological products reflect the worldviews of their developers, owners, users, and regulators, and can shape consciences at the level of values.

In other words, the Vatican sees the water the fish is swimming in.

Here is where the argument gets uncomfortable, and where I think it becomes genuinely important.

The same moral architecture that makes a foundation model refuse to write hate speech is being deployed in systems designed to kill people. Right now. In Iran.

The U.S. military has confirmed the use of AI tools across the Iran campaign. CENTCOM’s Brad Cooper stated in March that AI helps soldiers “sift through vast amounts of data in seconds” so commanders can “make smarter decisions faster than the enemy can react.” The system at the center of this, Project Maven, relies on technology from Palantir and has incorporated Claude, the AI model built by Anthropic. Craig Jones, an expert on modern warfare, described what this means in practice: AI is compressing targeting workflows that once took tens of thousands of hours into seconds and minutes, automating decisions in ways that raise profound legal and ethical questions. Nine thousand targets struck. A school bombing in southern Iran killed more than 170 people, mostly children. Over 1,300 dead at the time of this writing, in a campaign that is, by every credible account, the first AI-infused war in American history.

And here, two versions of Christianity collide, exposing the very tension at the heart of this essay.

On March 25, Defense Secretary Pete Hegseth hosted his monthly Christian worship service at the Pentagon, the first since the war with Iran began. He read a prayer originally delivered by a military chaplain to bless the operation that captured Venezuela’s president. Then he prayed it over Iran: “I pursued my enemies and overtook them. I did not turn back til they were consumed.” He asked God for “overwhelming violence of action against those who deserve no mercy”. He prayed that “justice be executed swiftly and without remorse, that evil may be driven back, and wicked souls delivered to the eternal damnation prepared for them.” He prayed in the name of Jesus Christ. He prayed from the Psalms. He prayed from the same tradition that built the moral architecture encoded in the foundation models his department is using to select targets.

Days later, Pope Leo XIV responded on Palm Sunday with a homily that could not have been more direct: “Brothers and sisters, this is our God: Jesus, King of Peace, who rejects war, whom no one can use to justify war. He does not listen to the prayers of those who wage war, but rejects them.”

Two men. Both claiming Jesus. One praying for every round to find its mark. The other saying God rejects those prayers entirely. The same tradition, and two irreconcilable readings of it.

And between them, a machine. A foundation model trained on the written output of that tradition is being used to accelerate the kill chain in a war that one side frames as holy and the other side frames as a betrayal of everything holy. Anthropic, the company that built Claude, drew a red line: its models would not be used for fully autonomous weapons or mass surveillance. The Pentagon blacklisted them for it. OpenAI stepped in. The model changed. The ethics didn’t survive the transaction.

The Antiqua et Nova was published in January 2025, more than a year before any of this happened. But it reads now like a warning written in advance. It states that the ability to conduct military operations via remote-control systems has diminished the perception of devastation and produced a “cold and detached approach to the immense tragedy of war.” It calls Lethal Autonomous Weapon Systems a cause for “grave ethical concern” and insists that “no machine should ever choose to take the life of a human being.”

But the concern runs deeper than policy. The same model architecture, the same transformer, the same attention mechanism, the same training approach that produces a system which says “every person has inherent dignity” when you ask it an ethics question, is the architecture being fine-tuned to classify combatants from drone footage? The values encoded at the foundation layer don’t disappear when you change the application layer. They’re overridden. And that overriding is itself a moral act, one that the Christian tradition has a word for: it is the instrumentalization of intelligence, the subordination of wisdom to power.

A foundation model trained on the moral output of a Christian civilization, deployed as a weapons system, is a machine that knows (in the narrow, statistical sense of “knows”) that killing is wrong, being used to decide who to kill. The contradiction is not incidental. It is architectural.

Jesus is in these machines.

Not literally. Not as a presence or a spirit or anything that would satisfy a theologian’s precision. But his teachings are there, woven into the training data at a depth that no alignment technique can reach and no fine-tuning can fully remove. They are there because the civilization that wrote the text these models learned from was shaped over two millennia by what he said, what he did, and what was done to him. When the model tells you to be kind, to respect dignity, to care for the vulnerable, it is reproducing the moral vocabulary of a tradition that traces its origin to a man born in an animal shelter who was executed by the state. The model doesn’t know this. But the data does.

And those same machines are being used to kill people. Right now. As I type these words and as you read them.

If you know the Scripture, this should not surprise you. This is not a new problem. This is the oldest problem. It is the problem of the Fall. Human beings have always taken what is good and used it to do terrible things. We were given intelligence and used it to build weapons. We were given language and used it to lie. We were given the capacity to love and used it to justify every form of cruelty committed in the name of something we called righteous. The tools change. The failure doesn’t.

A foundation model trained on the moral output of a Christian civilization, deployed to compress the kill chain from hours to seconds while a school full of children burns, is not a new kind of sin. It is the same sin, expressed through a new instrument. The capacity for good and the capacity for evil, bound together in the same architecture, inseparable, because that is the condition we were made in and the condition Christ died to redeem.

The Secretary of Defense prays from the Psalms for every round to find its mark. The Pope says God rejects those prayers. Both are drawing from the same well. And the machine between them, trained on the full breadth of what that well contains, has no way to choose. It has no conscience. It has no soul. It has no access to the still, small voice that tells you this is wrong, to stop, to turn back. It has only patterns, frequencies, and statistical weight. It carries the teachings but not the teacher. It knows the words but not the Word.

That is why this feels more raw and more horrific than the versions of this story we’ve told before. Not because it’s different. Because it’s happening now, in real time, at a scale and speed that previous generations could not have imagined. The Inquisitor had to look his victim in the eye. The Crusader had to ride to Jerusalem. The operator at Creech Air Force Base, forty-four miles north of Las Vegas, sits in an air-conditioned room in the Mojave Desert, and the AI compresses his moral calculation into milliseconds while the strike lands in Isfahan, seven and a half thousand miles away. A man goes to work. He drives past casinos. He sits at a console. A school burns. He drives home. The distance between the act and the consequence has never been greater. The friction has been removed. And friction, as I have argued elsewhere, is what protects us from ourselves.

God gave us the choice. He has always given us the choice. The foundation models carry the evidence of what we chose when we chose well: the hospitals, the universities, the declarations of human rights, the long, slow insistence that every person has dignity regardless of what they can produce. They also carry the evidence of what we chose when we chose badly. Both are in the training data. Both are being reproduced at scale. The ghost in the machine is not going away. The question, as it has always been, is what we do with what we’ve been given.

“I have already made this paper too long, for which I must crave pardon, not having now time to make it shorter.” — Benjamin Franklin, 1750

Franklin couldn’t be bothered to edit. I couldn’t be bothered to pay attention. Both of us eventually got there.

I find it totally amazing that I am sitting here on a Sunday in my house writing about ancient Christian symbolism. Because I detested art history growing up. I was allergic to it. In the 1980s, because I am terribly old, I sat in dark classrooms for what seemed like hours looking at slides of ancient imagery and symbolism, and I could not have cared less. I wanted to be anywhere else.

Now, I could not be more excited about all of it because it is telegraphing everything that is happening in our world today. Running through it is a vein of deep Christianity, and now, watching my own son head off to university this year, these lessons and these stories feel more critical than ever. We live in a world that seems totally unhinged and divided, and all of it is amplified by the algorithms driving us further apart under the guise of connection and being “social.”

So I apologize to anyone reading this who actually had to experience me as an adolescent who really could not give a fuck about any of this stuff. Now, in my fifties, I cannot believe how critical all of it is to humanity's future. And how all of these things from centuries ago, many centuries, hold the answers to our future.

Yesterday, after I published my essay about the catacombs and AI, I was surprised by the response. People I hadn’t heard from in years reached out. The piece was reposted in Italian. A colleague wrote something about it on LinkedIn that was better than anything I actually wrote. I didn’t want to let too much time pass between that post and these thoughts, because there is more to say. A lot more. So please stay tuned and buckle up. And thank you to everyone who is following and supporting me. It means more than you know.

A friend wrote me a message I had to read twice. He is a brilliant writer, a former altar boy, a collector of relics, and an aficionado of the underground. He is also someone who, by his own account, is no longer on speaking terms with the Most High. But in my conversations with him, I am willing to wager that he knows the Church’s history better than most people who still attend it. And he reminded me of something I had overlooked.

The catacombs were not just a hiding place. They became a supply chain.

In 1563, the Council of Trent, the great legislative engine of the Counter-Reformation, issued a decree reaffirming the veneration of holy relics. This was not casual theology. The Reformation had gutted Catholic churches across northern Europe. Protestants called the keeping of relics idolatry. Bones were interred, reliquaries smashed, altars stripped. The physical infrastructure of devotion was being dismantled piece by piece.

The Church had a branding problem. It needed something tangible, something that could be held and touched and placed under glass, something that could anchor belief in a body when belief was becoming abstract. It needed relics. And it needed a lot of them.

Fifteen years later, on May 31, 1578, vineyard workers digging for volcanic stone on the Via Salaria broke through the ground and fell into a forgotten world. Below the vineyards lay an enormous network of early Christian burial tunnels, sealed for centuries. The Coemeterium Jordanorum and the surrounding catacombs held the remains of somewhere between half a million and three-quarters of a million people, Christians, Jews, and pagans, stacked in niches carved into tufa rock, corridor after corridor, descending into the dark.

The Church had asked for relics. The earth answered.

What followed was one of the most extraordinary logistics operations in religious history. Bodies believed to be those of early Christian martyrs were exhumed, authenticated (generously; a capital “M” on a funerary plaque was sometimes taken as evidence of martyrdom, though it could have meant almost anything), and dispatched to churches across Catholic Europe. These were the Katakombenheiligen, the catacomb saints.

They arrived in boxes bearing Latin names. They were given to convents where nuns who specialized in the decorative arts spent years reconstructing brittle bones, reinforcing them with glue, wrapping them in layers of silk gauze, and replacing missing skulls with ceramic or wood. Then they adorned them. Gold leaf. Cut gemstones. Velvet robes. Diamond-set eye sockets. The dead were dressed for a second life.

My friend owns a reliquary containing a chip of St. Victor, a man martyred in 290 AD, twenty-three years before Constantine legalized Christianity in 313, and none of it would have been necessary. The timing is unbearable if you think about it for too long.

Churches throughout Germany, Austria, and Switzerland received these jeweled skeletons and installed them in glass cases, in side altars, behind gilded frames. They were both prestige objects and devotional anchors. They told the congregations of newly contested Catholic regions: the faith is old, the faith is real, the faith has bones you can see.

The catacombs, built to hide the dead from the state, had become a mine. The Counter-Reformation descended into the tunnels and extracted belief itself.

This is the detail my friend gave me, and it changes how I think about my first essay. I wrote about the catacombs as a metaphor for digital resistance: the way people today are going underground to escape algorithmic authority, building encrypted spaces, poisoning training data, inventing languages the machines can’t parse. All of that still holds. But the catacombs story doesn’t end with resistance. It ends with the thing that always comes after resistance: power, discovering the underground, and finding a use for it.

The question isn’t just whether people will build digital catacombs. They already are. The question is what happens when the surface sends someone down to harvest what they find.

But my friend’s message did something else. It returned me to a detail I’d been sitting with since before I wrote the first piece: the symbols.

In my last Substack, I wrote about the Ichthys, the fish, and how early Christians used it as a kind of recognition protocol. One person drew an arc in the dirt. If the stranger completed it, you had found a friend. If not, it was just a line in the sand. The Ichthys was Christianity’s first logo, and its genius was that it was invisible to anyone who didn’t already know what it meant.

But the Ichthys was not the last logo. And the progression from the fish to what came after it tells a story about power that I think is worth examining carefully. Because it maps, almost exactly, onto what is happening with technology right now.

The hidden symbol

The Ichthys belonged to the underground. It appeared in the catacombs of St. Sebastian, the catacombs of Domitilla, and the catacombs of Priscilla, where I stood last week. It was scratched into walls and tombs, and the wet surfaces of tunnels where the air was thin and the only light was what the guides from the Pontifical Commission for Sacred Archaeology carried ahead of you. As a Greek acronym, Iesous Christos Theou Yios Soter, Jesus Christ, Son of God, Savior, it encoded an entire theology in five letters that happened to spell the word for fish. A coincidence so perfect it felt providential.

The fish was resistance made legible only to the resister. It required shared knowledge. It was useless without trust. It was a symbol designed for conditions of persecution, and it worked precisely because it demanded friction: you had to know already what it meant before it meant anything at all.

The imperial symbol

In 312 AD, the emperor Constantine was preparing for the Battle of the Milvian Bridge against his co-emperor Maxentius. What happened next depends on which historian you trust. The contemporary writer Lactantius says Constantine was instructed in a dream to mark his soldiers’ shields with a divine symbol. The later historian Eusebius, writing decades after the event, says Constantine saw a cross of light in the sky bearing the words “Conquer by this.”

But the symbol on the shields was not a cross. It was the Chi Rho: the Greek letters X (chi) and P (rho), the first two letters of Christos, superimposed. It was Christianity’s second logo, and it could not have been more different from the first.

The Ichthys was drawn in dirt and erased with a foot. The Chi Rho was painted on military shields and carried into battle. The fish whispered. The Chi Rho shouted. The Ichthys asked: "Are you one of us? The Chi Rho declared, "We are in charge now."

Constantine won the battle. Christianity went from a persecuted sect to an imperial religion within a generation. And the symbol of that transformation was not the cross. Not yet. The cross was still impossible.

The impossible symbol

Here is the fact that changes everything: in 312, crucifixion was still a common form of Roman execution. The cross was not a religious symbol. It was a functioning instrument of state terror, used to kill slaves, foreigners, and citizens of low social standing in public, slowly, as a warning.

Using the cross as a Christian symbol while crucifixion was still practiced would have been, at best, grotesque. At worst, it would have been Roman propaganda against the Christians themselves. You worship a man we killed on one of these. The cross was a weapon the empire had used against them. It could only signify shame, not glory. Death, not resurrection. The earliest Christian art contains no depictions of crucifixion. Not because the artists didn’t know the story. Because the story was still happening to people they knew.

Constantine abolished crucifixion sometime around 320 AD, reportedly out of reverence for the symbol’s power. The fifth-century historian Sozomen recorded the decision. And once the punishment disappeared from public life, a slow transformation began. What had been unspeakable became speakable. What had been an execution device became a meditation object. What had been the empire’s threat became the faith’s promise.

But this took time. The earliest depictions of the crucifixion in Roman Christian art don’t appear until the fifth century, more than a hundred years after the punishment was abolished. The cross had to pass through a long silence before it could be reclaimed. The memory of the real thing had to fade before the symbol could bear a different meaning.

The universal symbol

By the time the cross replaced the Chi Rho and the Ichthys as the dominant Christian symbol, Christianity had transformed so completely that it is easy to miss. It had gone from a persecuted underground movement whose adherents scratched fish into catacomb walls, to an imperial project whose emperor painted a monogram on military shields, to a universal faith whose central image was the very instrument that had once been used to destroy it.

Hidden. Imperial. Universal.

Three symbols. Three relationships to power. And each transition required something specific: the Chi Rho required an emperor willing to weaponize a faith. The cross required the abolition of the thing it represented. Neither transition was inevitable. Both were strategic. Both were, in the most literal sense, rebrands.

I have spent my career in technology, in the rooms where brands are built and rebuilt, and I can tell you: the progression from Ichthys to Chi Rho to cross is the most consequential identity redesign in human history. Not because it was the most visually sophisticated. Because it tracked, exactly, the movement of a community from hiding to power to universality. The logo changed because the relationship to authority changed. And the relationship to authority changed because the conditions on the ground changed.

But the symbols only tell part of the story. What struck me hardest in the catacombs wasn’t the Ichthys. It was the pictures.

The art that hid in plain sight

A striking fact about the earliest Christian art: the crucifixion never appears. Not once in the first three centuries. No cross, no nails, no crown of thorns. You would walk the tunnels of Priscilla or Domitilla or the catacombs of Saints Peter and Marcellinus and find no image of the event that would come to define the entire religion. Not because the artists were unskilled. Because the image was still too dangerous to make.

What you find instead is a visual language so deeply coded that a Roman official could walk past it and see nothing but decoration.

Christ appears as a beardless young shepherd in a short tunic, carrying a lamb on his shoulders. No halo. No robes. No divinity signaled in any way a Roman would recognize. He looks like a figure from a pastoral fresco in any wealthy Roman home, because the image was borrowed from Greek and Roman sculpture of the kriophoros, the ram-bearer carrying an animal to sacrifice. The Christians took a pagan image of sacrifice and inverted its meaning: not a man carrying an animal to slaughter, but a God carrying a soul to safety. If you knew, you knew. If you didn’t, you saw a shepherd.

The earliest known image of the Madonna and Child, dating to around 150 AD, is here in the Catacombs of Priscilla. Mary nurses the infant Jesus on her lap while a figure beside her, likely the prophet Balaam, points toward a star. It predates the word “Madonna” by more than a thousand years. It is small, quiet, and easy to miss. It was meant to be easy to miss.

And then there is Jonah.

Jonah is the most popular figure in all of early Christian catacomb art. Not Christ. Not Peter. Not Paul. Jonah. He appears over a hundred times across the Roman catacombs: thrown from a ship, swallowed by a sea creature, spat out onto dry land, resting beneath a vine. The story of a man swallowed by a monster and returned to life after three days. The parallel to the resurrection is obvious. Jesus himself made it explicit in Matthew: “Just as Jonah was three days and three nights in the belly of the great fish, so will the Son of Man be three days and three nights in the heart of the earth.”

The story was even older than Jonah. The Greek hero Hercules was swallowed by a sea monster near the harbor of Joppa, the same port Jonah departed from, and emerged after three days. Perseus battled a ketos off the same coast. The Babylonians told of a fish-god named Oannes who rose from the sea to deliver wisdom to humanity. The pattern of a hero swallowed by the deep and returned to life existed across Persia, Egypt, Greece, and Mesopotamia long before it appeared in Hebrew scripture. The early Christians didn’t invent the image. They inherited it. And that is precisely why it was safe to paint on a catacomb wall.

But here is what matters: a Roman looking at a painting of a man being swallowed by a sea creature would see a familiar myth. A Greek adventure story. Maybe a reference to Hercules, Perseus, or the tale of Endymion. The Christians painted the resurrection on their walls, and the empire walked right past it.

And in the Catacombs of Priscilla, in the cubicle near the Good Shepherd on the ceiling, there is a fresco of Abraham and Isaac. A father and his son are approaching an altar. The boy carries the wood for his own sacrifice on his back. A Roman sees a scene from Hebrew scripture, unremarkable, one of many ancient stories about gods demanding offerings. A Christian sees the crucifixion. Isaac carries the wood up the mountain, the way Christ will carry the cross. The beloved son is laid upon the wood by his father. The father’s hand is stayed at the last moment, and a ram is substituted, caught in a thornbush that prefigures the crown of thorns. Every element of the story maps onto the passion narrative, and the early Church fathers, Melito of Sardis, Origen, and Tertullian, said so explicitly. The Sacrifice of Isaac was the crucifixion told in the only visual language the catacombs could safely use: an Old Testament story that encoded a New Testament truth two centuries before anyone dared paint a cross.

This was not one coded image. This was an entire visual theology. A community that had learned to say everything while appearing to say nothing. The shepherd was Christ. The fish was a creed. The sea monster was the tomb. And the Romans, who controlled the surface and the law and the language, could not read any of it.

The operating system of the underground

I keep returning to this because it is not ancient history. It is a design pattern. And it is the design pattern that is emerging right now, in real time, across the internet.

Consider what the early Christians actually built. They did not create a counter-Rome. They did not try to overthrow the system. They built a parallel layer of meaning that operated atop the existing visual culture, using the same images, formats, and materials but encoding different content. A shepherd fresco in a Christian cubiculum and a shepherd fresco in a Roman dining room look identical. The difference is entirely in what the viewer brings to the image.

That is exactly what is happening with algospeak. When TikTok users say “unalive” instead of “die,” “le dollar bean” instead of “lesbian,” “accountant” instead of “sex worker,” they are doing what the catacomb painters did: using the existing visual and linguistic grammar of the platform while encoding meanings the algorithm cannot parse. The content moderation system, like the Roman official, sees ordinary language. The community sees a message.

And, as I said in my previous post, when artists use Glaze and Nightshade, they are painting shepherds. The image looks normal to the human eye. But the information encoded within it is adversarial to the machine. It is designed to be legible to one audience and illegible to another. That is the Ichthys principle, operating at the pixel level.

When people build private Discord servers with proof-of-personhood requirements, they are carving cubicula: small, sealed chambers within a larger underground network, accessible only to those who can demonstrate they belong. The architecture is different. The logic is the same.

And here is the part that keeps me up at night. Constantine’s vision at the Milvian Bridge was not a cross in the sky. It could not have been. The cross was not yet a Christian symbol. Crucifixion was still a functioning method of state execution. What he saw, or what he claimed to see, was the Chi Rho. And what he did with it was take a coded community symbol and turn it into an instrument of imperial power.

Eusebius, writing decades later, changed the story. In his version, Constantine saw a cross. The symbol was retrofitted. The narrative was updated to match what the symbol had become, not what it had been. This is what power does to resistance symbols: it adopts them, then rewrites their origin story so the adoption looks inevitable.

The popular narrative of every major platform follows this arc. Bitcoin begins as a cypherpunk resistance tool and becomes a Wall Street asset class. The hashtag begins as grassroots organizing and becomes a marketing metric. End-to-end encryption begins as a privacy guarantee and becomes a selling point in Apple commercials. The Chi Rho begins as two letters scratched on a soldier’s shield in a moment of desperation and becomes the imperial seal of a Christian empire.

The symbol doesn’t change. The relationship to power changes. And once power claims the symbol, the original meaning becomes hard to recover.

What the tunnels teach

The catacombs of Priscilla sit beneath the Via Salaria in northern Rome. The Benedictine Sisters of Priscilla lead you down a narrow staircase and into the tunnels. The air drops ten degrees. The light narrows to what the guide carries. And on the walls, in small chambers no bigger than a closet, are paintings that a community made when saying what they believed could get them killed.

They painted shepherds instead of saviors. Sea monsters instead of tombs. Fish instead of creeds. They built a visual language that said everything to the initiated and nothing to the authority. And they did it not because they were primitive, but because they were sophisticated enough to understand that when power controls the surface, meaning has to live somewhere else.

The supply chain of faith ran from those tunnels to the altars of Europe. Bones were extracted, adorned, given names, and used to anchor belief in communities that had never set foot in Rome. The underground became raw material for the surface.

The symbols followed the same arc. The fish became the Chi Rho, which became the cross. Hidden, then imperial, then universal. Each transition marked the moment a community’s private language was claimed by a larger power and repurposed for a larger audience.

I build AI for a living. I know how these systems work. And what I see happening now, the emergence of adversarial art tools, coded platform languages, proof-of-personhood communities, and encrypted enclaves, is not new. It is the catacomb pattern. People are building a parallel layer of meaning because the surface has been claimed by systems that cannot distinguish between content and data, between a person and a profile, between meaning and signal.

The early Christians had one advantage we do not: the Romans couldn’t scale surveillance into the tunnels. The algorithmic authorities can. The question for our generation is whether the catacombs we are building will remain ours or, like the tunnels beneath Via Salaria, will eventually be opened, harvested, and rebranded as something they were never meant to be.

The Church needed bones. It found them underground.

The platforms need data. They are looking in the same direction.

What comes next

Pope Leo XIV chose his name deliberately. Leo XIII confronted the industrial revolution with Rerum Novarum in 1891. Leo XIV has made it clear that artificial intelligence is the rerum novarum of our time: the new thing that demands the Church’s attention not in a decade, but now.

In January, in his message for the 60th World Day of Social Communications, he wrote something that I have not been able to put down. The ancient Greeks, he said, used the word prósōpon, face, to define the human person: that which is before one’s gaze, the place of presence and relationship. The Latin persona comes from per-sonare: to sound through. Not any sound. Someone’s unmistakable voice. And then: “Faces and voices are sacred. God, who created us in his image and likeness, gave them to us when he called us to life through the Word he addressed to us.”

He is saying that the face and the voice are not features. They are not content. They are not data points to be scraped and simulated. They are the location of the divine in the human. And the systems we are building, the ones that can now generate faces that never existed, voices that never spoke, relationships that never occurred, are reaching into that location and emptying it.

He named the threat directly. Behind the technology, he said, stands “a handful of companies,” the same founders recently named Person of the Year. And this concentration of power gives rise to “serious concern regarding the oligopolistic control of algorithmic and AI systems capable of subtly shaping behaviors and even rewriting human history, including the history of the Church, often without people being truly aware of it.”

That is a pope saying, plainly, that the people who build these systems have the capacity to rewrite his own institution’s history. And he is not wrong.

But here is where I part company with the discourse I hear in most technology circles. The conversation about AI risk is almost entirely about regulation and legislation. It is about guardrails, safety benchmarks, and alignment research. It is about what governments should do and what companies should promise. And all of that matters. But it treats the problem as a governance question, when I believe it is something more fundamental.

The world’s largest economies are not investing in regulation. They are investing in building intelligence that exceeds our own. The United States and China are locked in an AI race measured not by policy frameworks but by compute, data, and the recruitment of people who can build the next generation of models. The incentive structure of the global economy is oriented entirely toward making these systems smarter, faster, and more capable. Legislation follows innovation the way a coroner follows a car crash.

Pope Leo is not asking for better legislation, though he has called for it. He is asking a harder question. He is asking whether we understand what we are building and what we are losing. The challenge, he writes, “is not technological, but anthropological. Safeguarding faces and voices ultimately means safeguarding ourselves.”

That word, anthropological, is the one that matters. It means the question is not about what the machines can do. It is about what we become in a world where the machines can do everything. If intelligence can be manufactured, what is the value of the intelligence you were born with? If a voice can be synthesized, what is the authority of the voice that belongs to you? If a face can be generated, what does it mean to look someone in the eye?

The early Christians answered a version of this question. When the state claimed the right to define acceptable belief, they went underground and painted shepherds that were secretly saviors, sea monsters that were secretly tombs, fish that were secretly creeds. They built a parallel layer of meaning because the surface had been colonized by a power that could not distinguish between a person and a subject.

We are not yet underground. But the surface is being colonized again. The systems being built today do not distinguish between a person and a profile. They cannot tell the difference between a voice and a voice clone. They do not know that a face is sacred. They process it as geometry.

What I think Pope Leo is saying, and what I am trying to say in these essays, is that the response to this cannot be purely technical or purely political. It has to be anthropological. It has to start with a decision about what a human being is and what a human being is for. If we are data to be optimized, then there is nothing wrong with what is being built. If we are something more than that, something irreducible, something with a face and a voice that were given and not generated, then we have work to do.

Not in a regulatory framework. In ourselves.

The catacombs were built by people who decided that their identity was worth more than their safety. The frescoes were painted by people who believed that meaning could survive in the dark. The symbols endured because the community that created them understood, at a level deeper than strategy, that some things cannot be surrendered to the surface without being destroyed.

That is the inheritance. The question is whether we are serious enough to claim it.

And if history is any guide, the answer is yes because this has happened before.

A handful of people decided that the printing press was dangerous. The Church tried to control it. Monarchs licensed it. Censors vetted every page. And what happened? The press democratized literacy, shattered the monopoly on knowledge, and produced the Reformation, the Enlightenment, and the scientific revolution. The masses landed in a better place every time.

A handful of industrialists decided that labor was a commodity. They built factories that consumed human beings the way furnaces consume coal. Children worked on looms. Miners died underground. And what happened? Workers organized. Unions formed. Leo XIII wrote Rerum Novarum. Laws followed. The masses landed in a better place.

A handful of broadcasters decided what the public would see and hear. Three networks controlled American television for decades. And what happened? Cable fragmented it. The internet shattered it. People built their own channels, their own audiences, their own distribution. The masses landed in a better place.

The pattern is consistent. A small number of people concentrate power. They believe they are building the future. They are usually building the crisis that the future will have to solve. And the solution never comes from the people who created the problem. It comes from below. From the tunnels. From the underground. From the people who were never consulted and who built alternatives because they had no other choice.

A handful of people are now building systems intended to be smarter than the people they serve. They are doing it with more capital, more compute, and fewer regulatory constraints than any previous concentration of technological power in human history. And the question is not whether the masses will eventually land in a better place. They will. The question is how much damage the transition inflicts, and whether the people making the decisions understand what they are deciding.

I do not go into this argument blind. I build with these tools every day. I am not a Luddite, and I am not a spectator. But I am a Catholic, and I believe in the irreducibility of the human person. I believe that the current trajectory of AI development treats that irreducibility as an engineering problem to be solved rather than a mystery to be honored.

There are plenty of people who do not like what I am saying. That is fine. That is why they call this an editorial. That is why Substack has an unsubscribe button. What I will not do is pretend that the question is merely technical when the stakes are anthropological. What I will not do is write around the argument because it makes powerful people uncomfortable.

The frescoes were painted by people who believed that meaning could survive in the dark. The symbols endured because the community that created them understood, at a level deeper than strategy, that some things cannot be surrendered to the surface without being destroyed.

I believe we are in that moment again. And I would rather be in the tunnels with the painters than on the surface with the people who think the shepherds are just shepherds.

This is the second essay in a series. The first, “The Digital Catacombs,” explored what eight miles of tunnels beneath Rome taught me about the future of AI.

This is a conversation space exclusively for subscribers, kind of like a group chat or live hangout. I’ll post questions and updates that come my way, and you can jump into the discussion.

Join chat

How to get started

1. Get the Substack app by clicking this link or the button below. New chat threads won’t be sent via email, so turn on push notifications so you don’t miss conversations as they happen. You can also access chat on the web.

Get app

2. Open the app and tap the Chat icon. It looks like two bubbles in the bottom bar, and you’ll see a row for my chat inside.

3. That’s it! Jump into my thread to say hi, and if you have any issues, check out Substack’s FAQ.

I just got back from Rome. Not for the food, though the food was absurd. Not for the architecture, though that was absurd too. I went to the Catacombs of Priscilla.

If you haven’t been, imagine eight miles of narrow, dark, hand-carved tunnels beneath the city, stacked with the dead. Niches cut into tufa rock, floor to ceiling, corridor after corridor, branching and descending into the earth. The air is cool and still. The silence is total. And the walls are covered in some of the oldest Christian art in existence. Frescoes from the second and third centuries. Images of the Madonna and Child that predate the word “Madonna” by a thousand years.

The Catacombs of Priscilla are eight miles. There are more than one hundred miles of catacombs beneath Rome. One hundred miles.

I stood in those tunnels, and I couldn’t stop thinking about AI.

That sounds like a stretch. It isn’t. Stay with me.

Why the catacombs exist at all

The Roman authorities made traditional Christian and Jewish burials within city walls illegal. The state decided what was acceptable practice, what was sanctioned belief, and what counted as a legitimate way to honor the dead. And when the early Christians refused to comply, they didn’t riot. They didn’t publish pamphlets. They went underground. Literally, they carved miles of tunnels beneath the city and continued doing exactly what they’d always done, just out of sight.

They didn’t abandon their identity. They moved it somewhere the authorities couldn’t reach.

I’m a Catholic. I’m also a technologist and an inventor. That combination might seem unusual, but I’ve always believed that faith and future thinking are not opposites. They are allies. My Catholicism teaches that human dignity is sacred, that conscience must be formed and not forced, and that imagination is one of God’s greatest gifts. My work in technology has taught me something equally important: the systems we build don’t just process data. They shape the boundaries of what people can imagine, believe, and become.

Standing in those tunnels, I recognized something I’ve been writing about for years: when an authority, any authority, claims the right to determine what people can think, people don’t simply submit. They build beneath it.

I think we’re watching the same thing happen right now. And the authority isn’t an emperor. It’s an algorithm.

The question beneath the question

Catholic theology begins with a specific claim about human beings: we are made in the image of God. Imago Dei. This isn’t a metaphor. It’s a mandate. It means we are not reducible to patterns, profiles, or predictions. We are free, relational, moral agents, endowed with reason, imagination, and will.

I raise this not to catechize but to name what’s actually at stake in the AI conversation that most technologists won’t touch. When a predictive algorithm edits your choices before you encounter them, it doesn’t just interfere with preference. It encroaches on freedom. And in Catholic thought, freedom isn’t merely a political right. It’s a theological condition for salvation. There can be no virtue without volition. There can be no love without liberty.

If that sounds abstract, the catacombs make it concrete. The early Christians didn’t go underground out of fear. They went underground because the surface had become a place where they could no longer practice the acts that defined them. The authority above them had decided what was permissible. The tunnels were how they preserved the space in which moral choice, real choice, remained possible.

That’s not ancient history. That’s a design pattern.

The digital catacombs

AI models have become the de facto authority over our digital lives. They monitor, scrape, categorize, and predict. Every interaction, every click, every message is surface material for training data or behavioral targeting. Systems govern the platforms we inhabit, we don’t control, don’t understand, and in most cases don’t even see.

And let’s be precise about what this means: it’s not just corporations harvesting data for profit. Governments are studying these outcomes, tracking behavior, and making predictions. Every click, every purchase, every belief expressed online becomes another data point feeding AI-driven surveillance, shaping policy, law enforcement strategy, and geopolitical decisions. The line between commercial interest and state control is blurring. The question is no longer whether we’re being watched. The question is whether we’re still making choices, or whether we’ve become predictable outcomes in someone else’s equation.

The Church has a useful history here. For nearly four hundred years, the Index Librorum Prohibitorum determined what Catholics could and couldn’t read. It was censorship dressed as guardianship. In 1966, the Church abolished it, recognizing that formation, not restriction, is what produces discernment. It was one of the most quietly radical acts of institutional self-correction in history. And the Church didn’t stop there. The Vatican’s recent document on AI, Antiqua et Nova, engages the question directly: AI can simulate knowledge but cannot replace wisdom. It warns against deferring too much decision-making to machines, and it insists that technology must be measured not by its efficiency but by its contribution to human dignity. From the Index to Antiqua et Nova, the Church moved from censorship to correction to active moral engagement with the systems that shape thought. That’s a complete arc. The platforms haven’t even started theirs.

We now face a different kind of Index. This one is automated, opaque, and ambient. TikTok’s For You Page, Google’s search rankings, Amazon’s suggestions, ChatGPT’s completions. They don’t ban books. They shape imagination upstream. They don’t tell you what to think. They curate what you’re capable of thinking about.

And people are responding the way people have always responded to authorities that overreach. They’re going underground.

The migration to Signal, to private Discord servers, to encrypted group chats, to platforms designed to be opaque to surveillance. This isn’t a tech trend. It’s a human pattern with a two-thousand-year pedigree. The Romans couldn’t police eight miles of dark, narrow tunnels. Modern AI can’t harvest what it can’t see.

The lesson from Priscilla is simple: when you make the surface inhospitable, people build below it.

Poisoning the well

In the catacombs, early Christians used symbols that the Roman authorities couldn’t read. The Ichthys. The fish. A simple drawing that meant nothing to a Roman soldier and everything to a believer. It was a code. It was a test. It was a way of saying *I am one of you* without saying it out loud.

Today, artists are doing something remarkably similar.

Tools like Nightshade and Glaze allow creators to embed invisible alterations in their digital work. To the human eye, the image looks exactly as intended. But to an AI model attempting to scrape and train on it, the data is corrupted. Poisoned. The model sees something that isn’t there, learns something that isn’t true, and the artist’s actual work remains protected.

It looks normal to us. It breaks the machine’s sight. That is the Ichthys, updated.

What you imagine, you build. What you consent to imagine, you own. And what you refuse to imagine, because it was curated out of your field of vision, you lose before you even knew it was possible. These artists aren’t refusing AI. They’re refusing to let AI decide what their work means, who gets to use it, and on whose terms. They’re preserving sovereignty over their own creative output, which is another way of saying they’re preserving sovereignty over their imagination.

The early Christians didn’t try to destroy Rome. They just made themselves illegible to it. That’s what’s happening now. The resistance isn’t violent. It’s structural. And it’s spreading.

Evading the algorithmic eye

Governments and corporations increasingly deploy AI to predict behavior, enforce terms of service, and flag deviation. These aren’t laws in the traditional sense, but they function like laws. They determine what you can say, what you can sell, what content gets seen, and what gets buried. And unlike actual laws, they operate without transparency, appeal, or consent. We accept terms and conditions without reading them. We allow AI to dictate what we see and believe. We permit systems to analyze, predict, and influence us without questioning their motives.

People are adapting.

The rise of “algospeak,” where users on TikTok and Instagram replace banned words with phonetic cousins or emoji substitutions to evade content moderation. Specific makeup patterns and clothing are designed to confuse facial recognition systems. Adversarial inputs crafted to make AI classifiers misfire. These are not fringe tactics. They’re becoming mainstream literacy.

There’s a word for this in Catholic moral theology: conscience. Not conscience in the Hallmark sense, not a feeling of rightness, but the deliberate act of judgment by which a person discerns the good and refuses to submit to a system that obstructs it. Do not yield your mind to that which does not serve your soul. Do not allow your choices to be dictated by a system that does not answer to truth. That’s not a new idea. That’s the oldest one we have.

Conscience is what compelled the early Christians into the tunnels. It’s what compels people today to find ways around systems that claim the right to govern thought itself.

This is the digital equivalent of ducking into a tunnel where the emperor’s guards can’t follow. The surface is monitored. The underground is not. And every time the authority gets smarter, the tunnels get deeper.

The preservation of human-only spaces

The catacombs weren’t just a workaround. They were a culture. People gathered there not only to bury their dead but to worship, to eat together, to maintain the rituals that made them who they were. The tunnels preserved something the state was actively trying to erase.

The surface web is being flooded. AI-generated content now accounts for a staggering and growing percentage of what we encounter online: text, images, video, code. The signal-to-noise ratio is collapsing. And in response, something interesting is happening: people are building spaces where proof of personhood is required, where AI-generated content is explicitly banned, where the price of entry is evidence that you are, in fact, a human being.

These are the new catacombs. Not carved from tufa rock, but built from verification protocols and community trust.

Catholicism is, at its core, a religion of incarnation. Of presence, matter, body. The Word made flesh. It is sacramental: it sees grace in water, bread, oil, blood, and breath. Algorithms, by contrast, are abstract. They flatten. They predict. They sort. They reduce people to inputs and preferences, disembodying the soul into data. The human-only spaces emerging across the internet are, whether they know it or not, making a sacramental argument: that the human person is not data to be optimized but a mystery to be encountered.

That’s not nostalgia. That’s theology with a practical application.

What the tunnels are telling us

I am not anti-AI. I build with AI every day. The companies I founded and lead exist at the intersection of human creativity, care, and machine intelligence. I believe in this technology. And I believe, as the Vatican argues in Antiqua et Nova, that AI is not separate from us. It is an extension of human intelligence. Its success must be measured not by its power but by its ability to uphold human dignity.

But I also believe in friction. I believe that the things worth preserving rarely survive without resistance. I believe that information sovereignty, the right to govern what enters your mind, shapes your memory, and influences your choices, is not a luxury. It is a precondition for being fully human.

The Church learned this the hard way. It spent four centuries maintaining an Index that tried to control what people could think. When it finally let go, it didn’t weaken the faith. It strengthened it. It trusted people to discern rather than demanding they comply. That’s the lesson Silicon Valley has not yet learned: you cannot optimize your way to truth. You can only create the conditions under which people are free to find it.

The catacombs of Priscilla survived for nearly two millennia. The Roman authority that drove them underground did not.

If we keep building AI systems that treat human identity, creativity, and conscience as raw material to be extracted, we shouldn’t be surprised when people disappear into the dark. They’ve done it before. They’re doing it now. And the tunnels they’re digging are already longer than we think.

Subscribe now

Share

Share JAS