A few years ago, a software company in Stockholm ran an experiment. It gave its developers access to an AI coding assistant and measured what happened. Productivity rose. Code was produced faster. Certain routine tasks, boilerplate functions, unit tests, documentation, nearly disappeared from the developers' workdays. The company declared the experiment a success.
But the developers told a different story when interviewed six months later. Several described a creeping unease. They were producing more, and understanding less. The assistant handled the parts of the work that had once forced them to think carefully, to trace logic, to sit with a problem until it yielded. With those tasks removed, what remained felt like supervision: reviewing outputs, approving suggestions, nudging the system when it drifted. One developer, a woman who had spent a decade building her skills, put it simply: "I am becoming the person who watches."
She did not say the tool was bad. She said something more unsettling: that it was good, and that its goodness was changing what she was.
This is not a story about lost jobs. It is a story about lost becoming. And it points toward a question that neither the accountability frameworks of Kantian ethics nor the metaphysics of suffering in Schopenhauer quite reaches. The question is not who holds the line, or what lies behind it. The question is what kind of human beings we are turning into, now that the machines have arrived to handle the hard parts.
Friedrich Nietzsche, writing through the voice of a fictional prophet in the 1880s, saw this question coming from a direction no one else was looking.
The Prophet and the Marketplace
Also sprach Zarathustra is an odd book. Part philosophical novel, part prose poem, part prophetic sermon, it tells the story of a hermit who descends from his mountain to share his wisdom with humanity and discovers that humanity does not want it. Zarathustra's first speech, delivered to a crowd in a marketplace, introduces two figures that have haunted European thought ever since: the Übermensch and the Last Man.
The Übermensch is not a superhero. Nietzsche's German is precise: über means over, across, beyond. The Übermensch is the human being who overcomes herself, who creates new values rather than inheriting them, who treats life not as a problem to be solved but as a material to be shaped. The concept is less a destination than a direction: the human being who is always becoming, never finished, never comfortable with what she already is.
The Last Man is the opposite. He is the endpoint of a civilisation that has optimised away every difficulty, every risk, every source of discomfort. "We have invented happiness," the Last Man says, and blinks. He wants nothing that is hard to get. He risks nothing. He creates nothing. He is warm, fed, entertained, and safe. He has, by every measurable standard, succeeded. And he is, in Nietzsche's account, the most contemptible figure imaginable: not because he suffers, but because he has arranged his life so that he never has to.
The crowd in the marketplace, when Zarathustra describes the Last Man, does not recoil. It cheers. "Give us this Last Man!" they shout. This is the detail that makes the passage prophetic rather than merely cautionary. The Last Man is not a fate imposed from without. He is a fate chosen from within, eagerly, by people who see comfort as the highest good and cannot imagine why anyone would refuse it.
The Comfortable Apocalypse
Read the passage today and the resonance is difficult to avoid.
The direction of consumer technology for the past two decades has been the systematic removal of friction. One click to purchase. One prompt to generate. One subscription to access everything. The labour of choosing, searching, comparing, struggling with difficulty, has been progressively automated. Each individual removal is a genuine convenience. Taken together, they describe a trajectory: toward a life in which less and less is demanded of the person living it.
Artificial intelligence accelerates this trajectory. A system that writes your emails, summarises your reading, generates your presentations, plans your meals, and manages your schedule is not oppressing you. It is serving you. And with each task it absorbs, it narrows the space in which you are required to exercise judgment, to struggle with expression, to sit with uncertainty until your own thinking clarifies. The developer in Stockholm was not complaining about oppression. She was describing something more subtle: the atrophy of capacities that had once defined her professional identity.
Nietzsche's insight is that this atrophy is not an accidental side effect. It is the logical outcome of a value system that treats comfort, efficiency, and the removal of suffering as the highest goods. If the goal is to minimise friction, then the machine is the perfect instrument, and the Last Man is the perfect user. He does not resist the machine. He welcomes it. He cannot imagine why anyone would choose difficulty when ease is available. He blinks.
The danger Nietzsche identifies is not that machines will take over. It is that we will hand over, voluntarily, the activities through which human beings have historically become more than they were. Learning is difficult. Creating is painful. Exercising judgment under uncertainty is exhausting. These are not flaws in the human condition. They are the mechanism by which the human condition develops. A civilisation that outsources them has not been liberated. It has been diminished, and it will not notice, because noticing requires the very capacities it has allowed to decay.
The Three Metamorphoses
Zarathustra's first discourse after the marketplace describes three transformations of the spirit: the camel, the lion, and the child.
The camel kneels and takes on the heaviest burdens. It asks: what is most difficult? And it bears the weight willingly, because bearing weight builds strength. The camel represents the discipline of apprenticeship, of mastering a tradition, of submitting to standards one did not create. The medical student who memorises anatomy, the musician who practises scales for years, the programmer who traces algorithms by hand before ever touching a keyboard: these are camels. They are loading themselves with difficulty because they understand, at some level, that the difficulty is the point.
The lion destroys. It confronts the great dragon whose scales read "Thou shalt" and roars "I will." The lion cannot create new values, but it can clear the ground by refusing the old ones. The lion is the moment of rebellion: the scientist who rejects the paradigm, the artist who breaks with tradition, the engineer who refuses to build what the market demands simply because the market demands it.
The child creates. Having borne the weight (camel) and refused the old commands (lion), the spirit achieves what Nietzsche calls "a sacred Yes": the ability to begin, to play, to create values out of nothing but its own fullness. The child is not innocent in the sense of ignorant. It is innocent in the sense of unburdened by resentment: it creates not against something but out of something, an overflow of life that needs no justification beyond itself.
The three metamorphoses, read as a developmental sequence, describe how human beings become capable of genuine originality. And each stage requires something that artificial intelligence threatens to remove.
The camel requires difficulty. If the AI assistant handles the tedious parts of learning, the camel never builds the strength that comes from carrying weight. The medical student who uses a diagnostic system from the first day of training may never develop the deep pattern recognition that comes from years of looking at scans with nothing but her own eyes and a textbook. She will be more efficient. She may be less formed.
The lion requires confrontation with inherited values. But if the system has already optimised the workflow, already determined best practices, already established what "good" looks like by training on millions of examples, what is there to refuse? The lion's "I will" requires a "Thou shalt" to push against. A system that presents optimised solutions as defaults gives the lion nothing to fight.
The child requires the capacity for genuine creation, which presupposes having passed through discipline and refusal. If the first two stages are truncated, the third cannot arrive. What looks like creativity (generating novel outputs, recombining existing elements, producing surprising variations) may be something else entirely: pattern recombination without the inner transformation that Nietzsche considered essential. The machine can produce the appearance of the child's play. It cannot undergo the metamorphosis.
The Eternal Recurrence as a Design Test
There is a moment in Die fröhliche Wissenschaft, which Nietzsche later wove into Zarathustra's teaching, where a thought experiment is proposed. A demon visits you in your loneliest hour and tells you: this life, as you have lived it, you will have to live again and again, eternally, with nothing changed. Every pain, every joy, every thought, every sigh. Would you curse the demon, or would you call it the most divine thing you had ever heard?
The eternal recurrence is not a cosmological claim. It is a decision test, perhaps the most rigorous one ever devised. It asks: do you affirm your life so completely that you would choose it again, unchanged, forever?
Apply this to the design of human-AI systems. Would you will the current arrangement to recur eternally? Not just the productivity gains and the conveniences, but the atrophy, the dependency, the narrowing of human capability that accompanies them. Would you will a world in which each generation is slightly more supervised and slightly less self-directed than the last, forever? Would you will the Last Man's comfort, repeated without end?
If the answer is no, then something in the arrangement needs to change. The eternal recurrence does not tell you what to change. It tells you that you have not yet arrived at a configuration you can fully affirm. And the inability to affirm is, in Nietzsche's framework, a signal that the will to self-overcoming has not yet done its work.
This is a more demanding test than Kant's universalisability (can you will this as a universal law?) or Schopenhauer's compassion (does this recognise the suffering of others?). It asks not whether the arrangement is permissible or humane but whether it is worthy of being willed eternally. It is a test of affirmation, and most current configurations of human-AI interaction would fail it, not because they are evil but because they are settling for less than what human beings could become.
Self-Overcoming and the Refusal of Comfort
The thread that runs through all of Zarathustra's teaching is Selbstüberwindung: self-overcoming. Not self-improvement in the modern therapeutic sense, which typically means becoming a more comfortable, better-adjusted version of who you already are. Self-overcoming means becoming something you are not yet, and it requires the deliberate embrace of difficulty, resistance, and even suffering as the materials from which new forms of life are made.
This is the concept that puts Nietzsche most directly at odds with the trajectory of artificial intelligence. The logic of AI development is the logic of friction removal: identify what is difficult and automate it, identify what is slow and accelerate it, identify what is uncertain and predict it. Each step is rational. Each step, taken in isolation, is beneficial. And each step removes a piece of the resistance against which human beings have historically formed themselves.
Consider education. A system that tutors a student by identifying her weaknesses and providing personalised exercises is, by every standard metric, more efficient than a human teacher working with thirty students at once. But the human teacher does something the system cannot: she models the struggle of understanding. She shows the student what it looks like when an intelligent person encounters something difficult and works through it, not around it. The student who learns from the system learns the material. The student who learns from the teacher learns the material and a way of being in relation to difficulty.
Or consider creative work. A writer who uses a generative system to produce first drafts saves time. But the first draft is, for many writers, where thinking happens. The struggle to find the right word, the right structure, the right way into a problem is not an obstacle to the finished product. It is the process through which the writer's understanding develops. Automate the struggle and you may get the text faster. You will not get the writer.
Nietzsche would not have opposed artificial intelligence any more than he opposed railways or telegraphs. He was not a Luddite. But he would have asked, with characteristic sharpness: are you using this tool in the service of self-overcoming, or are you using it to avoid the very difficulty through which overcoming occurs? The question is not whether the machine is powerful. The question is whether you are becoming more or less through your use of it.
What Zarathustra Would Ask
Zarathustra came down from the mountain because he had something to give. What he found in the marketplace was a crowd that wanted the opposite: not challenge but comfort, not becoming but arrival, not the Übermensch but the Last Man. He spent the rest of the book looking for the rare individuals who were still capable of hearing what he had to say.
The parallel is not subtle. Artificial intelligence arrives in a culture that already values convenience, efficiency, and the removal of friction above nearly everything else. It arrives, in other words, in a culture already inclined toward the Last Man. The question is not whether machines will make us into Last Men. The question is whether we were already heading there, and whether the machines are accelerating a trajectory we chose long before they arrived.
Nietzsche offers no programme, no policy, no design framework. What he offers is a question, and it is more uncomfortable than anything Kant or Schopenhauer asks: not "who is responsible?" and not "what does the machine lack?" but "what are you becoming, and is it worthy of you?"
The developer in Stockholm felt the question before she could articulate it. She was not losing her job. She was losing the difficulty that had made her who she was. Whether she finds a way to reclaim it, whether any of us do, is not something a philosopher who died in 1900 can answer.
But he saw it coming. That counts for something.
Suggested references: Nietzsche, Also sprach Zarathustra (1883–1885); Nietzsche, Die fröhliche Wissenschaft (1882/1887).
This is Part III of a three-part series on what artificial intelligence cannot replace. Part I: Kant And The Exam That Changed Nothing. Part II: Schopenhauer And The Shadow.