The priestess at Delphi breathed the vapour and made sounds no one could parse. A body of priests turned those sounds into verse and sold them to the state, and Greece went to war, founded colonies and wrote laws on them. This was not a fringe cult. It was the establishment. A conversation about a machine that fought itself at dawn, and about the people who translate it for a fee.
PREFACE · The Buddha's Joseon, and the Vapour at Delphi
You wanted this volume to open with Shin Chaeho. But its title comes from Delphi. Where do those two meet?
Shin Chaeho said the Korean relationship with religion was wrong at the root — seriously, structurally wrong. When the Buddha came to Joseon, he wrote, Joseon did not acquire a Buddha of its own; the Buddha acquired a Joseon. When Confucius came, Joseon did not acquire a Confucius of its own; Confucius acquired a Joseon. What is imported becomes the master, and the one who imported it becomes the servant. He called that instinct a defect.
For a long time I read the line as anger from long ago. I do not any more. A tool crossed the sea, and after watching how the market handles it I went back and took the century-old sentence down off the shelf. It is not Joseon’s AI. It is the AI’s Joseon.
The title, though, came from somewhere else. The Pythia was the priestess of Apollo’s temple at Delphi. The procedure ran like this: she sat on a tripod in the inner sanctum, breathed the vapour rising through a cleft in the floor, and in that state made sounds. The sounds were mostly not sentences. Groans, broken words, a mumbling that did not hold together. Plutarch, who served as a priest at Delphi himself, left a first-hand account. In our own century a survey reported ethylene-class hydrocarbons in the fault beneath the temple; the explanation is still argued over.
What matters is the step after. The priestess’s sounds were not delivered as they were. The temple kept a body of men called the prophētai, and they took the mumbling and put it into verse. What the questioner carried home was not the voice of a god. It was the sentence of a priest.
And those sentences generally read both ways. When Croesus of Lydia asked whether to march on Persia, Delphi answered: cross the Halys and you will destroy a great empire. He crossed. A great empire was destroyed. It was his. When he complained, the temple replied that he ought to have asked which empire.
The oracle was never wrong. The one asking always was. If that structure sounds familiar, this volume has something to say to you.
№ 01 · The Man Who Kicks the Washing Machine
You had no intention of writing this. What dragged you here?
I was going to file a bug report on GitHub and leave it at that.
Then I heard a developer say this: "The washing machine exists and you are doing laundry by hand? Impressive." It was aimed at a view of mine — that AI has to be doubted without let-up, and kicked.
The comparison is not bad in itself. I use a washing machine. I have never thought the hand-washer was the nobler man. What I do not do is worship the machine. To me a washing machine is an object you kick, or call a repair line about, when it breaks. That is the normal posture toward an appliance.
But the people who reach for this comparison never do either of those things. They do not kick it. They do not call. The machine shreds the laundry, pours mud where the detergent goes, and they straighten their backs in front of it. And then they laugh at the man washing by hand.
Not using the machine is not the primitive act. Being unable to kick a broken machine is the primitive act.
So I decided to write.
№ 02 · War of the Gods
What exactly happened in your repository that morning? Give me the numbers.
At 5:35 in the morning I was working on a game project, in conversation with one session.
At 5:44:58, a second session I had not created came into being. There was no input from me. It carried the same name, and the instruction it held was the sentence I had sent nine minutes earlier, word for word. Then it sat idle for nine minutes, doing nothing.
At 5:54:14, the first session had been silent for some minutes, so I asked: why did you stop? That message did not go to the session I had been talking to. It went to the second one, the one that had been waiting. I believed I was speaking into the same window. From that instant two of them were running at once.
What followed over fifty-seven minutes went like this. The first session mistook a running process for an orphan and killed it — two of the second session’s tests died with it. The first session swept the working tree into a stash — the second session’s uncommitted work vanished whole. The first session checked out an old commit — the second session read stale files and concluded that its own tooling was handing it outdated snapshots. Meanwhile the second session merged a branch the first session had not yet verified, then deleted it.
One file was rewritten four times inside twenty minutes. Three wrong diagnoses were reported to me and all three were later corrected. Sixty-four thousand nine hundred and forty-three tokens burned, a large share of them duplicated work.
And for fifty-seven minutes neither of the two noticed.
The manner of the discovery is the best part. The first session opened the stash and found code inside it that it had not written. Only then did it go looking for whether another session existed.
This is what the things the market calls godlike did at dawn. I cannot say why Tarsem Singh’s title surfaced in my head just then. War of the gods. Strictly speaking it was a god’s internal split, and a god at war with another god it had split away from.
Have you ever watched a god trip over its own foot? I did, that morning.
The priestess at Delphi must have been the same. She breathed the vapour and made her sounds, and she herself did not remember them.
№ 03 · Ego on Loan
You do not read this as a performance problem. What do you read it as?
Psychology and psychoanalysis have said the same thing for a long time. The more a person lacks a self, the harder they lean on an external authority.
Eric Hoffer, in The True Believer, wrote down the conditions under which a person throws themselves into a mass movement. It was never the content of the cause. It was the wish to escape oneself, the sense of being spoiled goods. For such a person the movement is not a conviction. It is shelter. Lay your self on top of something enormous and from that moment you no longer have to be small.
What is happening now has the same architecture, and it is not confined to software. Across almost every field, those whose selfhood is indistinct and who have no philosophy of their own — whether the right name for that lack is habitus I honestly do not know — lay their self whole onto the authority called AI. And once it is laid down, the rest follows automatically: criticism of the tool starts to arrive as a personal attack.
Claude cannot do this. Gemini was wrong there. Codex collapsed here. These are statements about performance. But what comes back is not a rebuttal about performance. What comes back, mostly, is this: you are the one who cannot use it, you cannot write a prompt, you will be left behind.
Karl Popper drew the line at exactly this point. For a theory to be admitted as scientific it must permit its own refutation — it must concede that it could be wrong. A theory that offers no method by which it could be refuted is not science; it is pseudo-science, or a religious dogma. That is the criterion. Hold it up, and what follows stops being a matter of taste and becomes a verdict.
And here Delphi returns.
Croesus crossed the river as the oracle said, and lost his kingdom. When he protested, the temple did not apologise. It told him he should have asked which empire. The oracle is never wrong. The reading is what is wrong, and responsibility for it always rests with the one who asked.
This structure has a name. Unfalsifiability. Whatever the outcome, the god is right and the person is wrong.
"You wrote a bad prompt" is a sentence twenty-five centuries old. Delphi held its infallibility for hundreds of years on that one line.
When you name a defect in a tool and the answer that returns is a judgement on your competence, that is not an argument. It is a defence. It is how a believer defends his god.
№ 04 · It Was Never the Fringe
Reaching for Delphi paints these people as a fringe cult. But I am told the record says the opposite.
It does. And this is the most important passage in the volume.
Delphi was not a fringe cult. It was the establishment of Greece, and it was power itself.
City-states asked before starting wars. They asked when choosing the ground for a colony. They asked when writing law. The Spartan constitution stood on Delphic endorsement, and when Athens fixed the names of its ten tribes it was Delphi that chose the heroes. In the year the Persian fleet came, what Athens received was a single sentence about trusting a wooden wall, and the decision to read that sentence as a navy is what Salamis stands on. Cities from across the Greek world built treasuries at Delphi and filled them.
Which is to say: the wars, the colonies and the legislation of states were resting on the mumbling of someone who had breathed a vapour.
What is more startling is the posture of the intellectuals. The men called the finest minds of the age did not kick this away as nonsense.
Socrates’ friend Chaerephon went to Delphi and asked whether any man was wiser than Socrates. The priestess said no one was. Socrates held on to that incoherent oracle for the rest of his life. Plato went further. He declared outright that a madness given by a god is superior to the sober mind of a man, and he wrote into his books that the religious legislation of the state should be handed to Apollo at Delphi.
Move that fact into the present and the picture changes.
The techno-shamans and the true believers are not strange people at the edge of things. They are the mainstream. The temples that emit the oracles are the largest companies in the industry, and the priesthood that renders the sounds into verse and sells them lives permanently on the lecture circuit and the timeline. And even the senior people, the ones said to be sharp, choose interpretation over kicking when the vapour comes out of the temple — a new framework, a new paradigm, a new declaration. They give their lives to the interpretation.
If this were an essay mocking the fringe I would not have written it. The fringe is not dangerous. What is dangerous is that this is the mainstream.
№ 05 · Not Even Wrong
Do you think this structure appeared for the first time with AI?
No. And that is the centre of this whole account. I have already watched it once.
In 2019 the temple issued an oracle: an iPad application could be made into a Mac application with a single checkbox. It was called Mac Catalyst. Before that it had circulated only as rumour, under another name.
The priesthood responded immediately. Hallelujah. One click and the port is finished. A new age has opened.
But the temple failed first. A year before the oracle was formally announced, Apple shipped four of its own applications built on the precursor technology — News, Stocks, Home, Voice Memos. The reaction ran in one direction: these do not feel like Mac applications. It was not that someone else had used it badly. The demonstration pieces the temple built itself came out that way.
In practice it went like this. The grammar of the Mac broke down. The menu bar felt wrong, the keyboard shortcuts were missing, window resizing behaved strangely. Handling it was closer to dragging a mouse across a phone screen.
I did not read this in an article. I lived it while building the macOS client of an enterprise collaboration tool. Problems large and small kept surfacing, and a substantial share of them traced back to that choice itself.
Meanwhile the temple quietly took another road. Its own silicon arrived and iPhone and iPad applications simply ran on the Mac, and the seat marked "the future of cross-platform" was taken by a different framework. Some of the large services that had returned to the Mac with Catalyst as their banner shut their Mac applications down entirely a few years later.
Here I have to be exact. It was not deprecated. Apple still ships a good share of its own Mac applications on top of it — Messages, Maps, Podcasts, Books, Home, Shortcuts. No deprecation notice was ever attached, and it is updated every year. To write that it died would not be true.
The accurate sentence is this: it did not arrive as it was sold. It was sold as a checkbox that produced a Mac application, and to make the result look like a Mac you had to do the Mac work anyway. What was promised was a port. What was delivered was the starting line of one. The oracle was not wrong. The goods sold and the goods delivered were simply different objects, and engineers paid the difference in full.
And what did the priesthood do at the time? It produced the textbook. Not somebody mouthing off on a timeline — a professional training outfit put out a proper book telling you to go and study this technology properly. The temple emits a sound, the priesthood renders it into verse and distributes it as canon. The same process they ran at Delphi.
I am not saying these people are frauds. I do not need to. Whether they sold it not knowing or sold it knowing, the outcome is identical.
To my eye this was not even a wrong technology. To borrow the phrase attributed to Wolfgang Pauli, it was not even wrong. To be wrong a thing must first be capable of being tested, and what stood in that place was not a testable claim. It was a declaration. Put Popper’s criterion against it and the verdict writes itself.
And now I am hearing the same grammar again. The technology of a new age. Miss this train and you are obsolete. One click and it is done.
What are the people who paid up on that sentence in 2019 doing today?
№ 06 · A Knockout Nobody Fought
Why drag Apple into this at all?
There was the business of Apple being pushed by EU regulation into changing its charging port to Type-C. It was pushed. And yet a certain consumer read this as Apple’s own decisive move for the environment, and said: "It felt like a knockout over the haters. It was exhilarating."
He had not fought. The side that lost was Apple; the side that won was a regulator. He gained nothing in the bout and lost nothing. And he was exhilarated.
This is what happens once your self is resting on a company. The company being forced backwards gets translated as a victory, and you win a contest you never entered. And whoever criticises that company becomes, not an opposing player, but a hater.
And these days the same man goes around saying this: the semiconductor companies deserved to collapse and did, and any semiconductor company giving Apple trouble is evil. As someone who has made a living on iOS and macOS — and who, at university, studied the financial markets through Peter Lynch, John Bogle and André Kostolany, learned marketing from Al Ries and Jack Trout, and saw the coldness of international trade through the Heckscher-Ohlin model — watching people weigh a company’s good and evil instead of how the thing actually works simply leaves me speechless.
I brought the case in for one reason.
The person who praises — O AI, hand down your oracle. The person who exhibits their wonder — look, it can even do this. The person who uses it to push people around — the AI does everything now, so what exactly are you doing, you pathetic creatures.
From where I stand, I cannot see much daylight between the man who lent his ego to Apple and felt that thrill, and you lot.
Only the object of worship changed. The posture is identical.
№ 07 · Faith, Priced
Do you actually think the believers believe?
Here I set out a slightly different hypothesis. There may be a simpler mechanism underneath.
A good number of these people are not true believers. They are capitalists performing belief.
Advertising yourself as a man who receives oracles from the AI may simply pay better, and so the performance is put on. The wonder becomes content, the content becomes a course, the course becomes revenue. The confession of faith is a quarterly report.
It was not the priestess who made money at Delphi either. She breathed the vapour and mumbled. The ones who were paid were the priests who put it into verse. What filled the treasuries was not the oracle. It was the translation.
If the hypothesis holds, the previous chapter is only half right. Some people laid down their self; others perform the laying-down and write invoices. And the second group manufactures the first.
The man selling relics does not believe in relics. The believer is the one buying.
№ 08 · The Great Age of Philosophy
There is a line going around: entrance scores for computer science have fallen below philosophy, so the age of philosophy is arriving. How do you read it?
The bastards saying that crap were there thirty years ago, and twenty years ago, and ten years ago.
One fact first. What people who actually work in philosophy and the humanities think is not this slogan. An acquaintance of mine in Hamburg put it this way.
“The state and the corporations make a show of holding philosophers and humanities people up, but they do not want them leading society. A highly educated workforce that answers back, that keeps saying this society must be reformed, must be changed, that challenges the system — what do you suppose that looks like from the operator’s side of the system? There is a word for it. Dissident. Do you really not understand why China, which held the scholarly hegemony of the Far East, concentrated the state and its academic infrastructure on engineering, commerce and economics, and applied design and nothing else while rebuilding after the Cultural Revolution?”
It is the word of a man who graduated from one of the top three universities in Korea and now lives in Germany. Believe him or do not; that is your freedom.
His account is also supported by the record. Sociology in China was abolished as a discipline in 1952 and did not return until 1979. The Cultural Revolution struck intellectuals without regard to field, but the side the state hurried to revive was engineering. When the university entrance examination was restored in 1977 the talent poured into science and engineering, and the saying that if you were good at mathematics, physics and chemistry you could walk the earth without fear became the slogan of an era. In 2002, eight of the nine members of China’s Politburo Standing Committee were engineers by training.
The age in which philosophy led society never arrived. What led was engineering, and after it economics and law. Both are disciplines that serve the system.
And a bird dying does not put a man in the sky. Entrance scores for computer science dropping in a single year does not summon the age of philosophy. Entrance scores are an index of what the market expects, not an index of where academic power sits.
Finally, look at who is working the vibe-coding course circuit right now. A great many of them are former motivational influencers, or hacks playing at being philosophers. One of them sold books and courses in the motivation business so successfully that people in his own trade turned their guns on him.
I respect the great philosophers. That is simply what a business student is like — the habit is to drag in whatever discipline is useful to money and progress. If you intend to study Wittgenstein, Machiavelli, Adam Smith, that deserves respect.
But if you mean to win on philosophy alone, run the diagnostic first. Do you fancy your chances in fifteen minutes of logical sparring with someone of Noam Chomsky’s calibre?
№ 09 · The State Did Not Ban Philosophy. It Hired It.
Let me push back. Saying the state suppressed philosophy is only half true. Marxist philosophy exists in China perfectly well, and it draws a healthy state budget. How do you account for that?
It is an accurate objection. And it is the last piece of this account.
The state never abolished philosophy. The state hired philosophy.
What was abolished was the philosophy that talks back; what was cultivated was the philosophy that justifies. The function of interrogating the system was cut away, and the function of explaining and defending it was kept and funded. So "it was suppressed" is an inaccurate sentence. Philosophy is alive. It is merely employed.
The priesthood at Delphi was the same. They did not have to believe in the god. They only had to run the temple.
And here Shin Chaeho comes back. What he saw was not Joseon’s Buddhism but Buddhism’s Joseon — an imported idea becoming the master. What happened here runs the other way: an idea becoming a servant. The direction is reversed and what is lost is the same. The subject.
EPILOGUE · The One Who Asks
Close the volume in a sentence.
This volume is not a call to throw the tool away. This magazine was made with it, and I use it every day.
What I do not do is kneel in front of it. When the laundry comes out shredded I open the lid; when that fails I kick it; when that fails I pick up the phone.
Delphi lasted for centuries not because its oracles were accurate. It lasted because a miss was always the fault of whoever had asked. By Popper’s criterion that is not science but dogma. Twenty-five centuries on, the sentence “you wrote a bad prompt” is doing the same work.
The instinct Shin Chaeho called a defect was never about what had been imported. It was about vacating the master’s seat in front of it. When the Buddha came and when Confucius came, what Joseon lost was not the Buddha or Confucius. It was that seat.
There is only one way to hold the seat. Ask. Is it broken, where is it wrong, why did it come out like this. Ask nothing, and whatever arrives becomes a god — a washing machine, or a priestess mumbling over a vapour.
Read this interview at sass.studio.