I recently posted something on X about Mo Gawdat’s framing of AI as the next leverage layer in human history. Hunter-gatherer to agriculture to industry to information to AI, with each transition rewarding whoever owned the automation beneath everyone else. I said the framing was good, that staying at the bleeding edge would require a kind of merger with the machines that most people won’t want, and that for the rest of us, life would probably get more prosperous and interesting, the way each previous jump made it.
That last part wasn’t quite right. Or rather, it was right in a way that almost everyone, including me in 280 characters, gets wrong. Let me try to do this properly.
Movement I The Optimism Trap
The standard takes on AI sort into a few familiar buckets. Things will get better: Pinker, Kurzweil, the accelerationists. Things will get worse: the doomers and the displacement-anxious. Things will be mixed, but we’ll adapt: the Mo Gawdats of the world, lamenting the narrowing of the elite path while waving at general prosperity.
All of these takes share a hidden assumption: that the yardstick stays the same across the transition. They argue about whether the numbers go up or down, but the numbers themselves are inherited from the epoch that’s ending: material wealth, productivity, lifespan, professional status, cognitive capability. They’re the measures the Enlightenment taught us to care about, and the ones the whole knowledge economy was built to maximize. Every one of them is a measure of what slow, deliberate, analytical thinking can produce, the mode psychologists call System 2. (More on that term in a moment.)
Measure the AI epoch with those instruments and you get a story about disruption and adaptation, with the elite path narrowing and everyone else experiencing some mix of gains and losses. That’s Mo’s story. It isn’t wrong on its own terms. It’s just stuck inside a frame that I think is about to break.
Movement II The Civic at the Drag Strip
For the last two hundred years, civilization has been a drag strip, and the only car available to race on it was a Honda Civic.
The Enlightenment, the Industrial Revolution and the Information Age selected for and rewarded one particular kind of human cognition: analytical reasoning, symbolic manipulation, the ability to hold complex abstractions in working memory and run logical operations on them. Call it System 2 thinking. Call it Spock cognition, if you grew up on Star Trek. Whatever you call it, this is what universities credentialed, what professions selected for, and what the knowledge economy paid a premium for.
And it worked. System 2 cognition built the modern world. Science, engineering, law, medicine, finance and technology all run on explicit reasoning, and the whole institutional apparatus of credentials, hierarchy and reward was tuned to find the people who were good at it and put them where they had leverage.
But the Civic was never a race car. It was pressed into drag duty because nothing faster existed. Since late 2022, something faster does: a Top Fuel dragster of analytical capacity, on demand, with unlimited fuel and no fatigue. The Civic hasn’t become useless. Its value on the drag strip is collapsing, and the strip itself is ceasing to be where the value lies.
The Civic was never a race car.
It was the only car
available.
The Honda Civic
Universities credentialed it. The knowledge economy paid a premium for it. It built the modern world. It was pressed into drag duty because nothing faster existed.
The Top Fuel Dragster
Unlimited fuel and no fatigue. The Civic hasn’t become useless. Its value on the drag strip is collapsing, and the strip itself is ceasing to be where the value lies.
It’s worth being precise about which System 2 work collapses and which grows. Generic analytical work, the drafting, summarizing, boilerplate, mid-tier code, standard analysis and routine research, is being pushed toward near-zero marginal cost. That was the Civic’s old job, and the dragster now does it nearly for free. Architectural System 2 work grows more valuable, not less: designing the constraints, the multi-agent workflows, the evaluation tests and the systems that direct what the dragster does and check what it produces. The shallow layer commoditizes. The deep layer amplifies.
At least for now. Architectural work is a high-leverage human skill today, and I doubt it stays that way; AI systems are already learning to design and orchestrate other AI systems. What I expect to last is narrower: deciding what is worth building at all, and judging whether a system is doing the real work or only imitating its form. Movement VII returns to this.
What ends is the assumption that analytical altitude is the point. That assumption organized two centuries of credentials, hierarchy and reward. Its ending opens ground that was always more interesting than the strip we paved over it.
Movement III What System 2 Was Actually For
Here’s the move most people haven’t made yet. System 2 cognition was never the goal. It was the balloon.
You build a hot-air balloon to ascend. You ascend in order to see. The view is the point. Nobody confuses the balloon with the view. Nobody chrome-plates the gondola and calls it a destination. The balloon is disposable equipment for reaching a perspective you can’t get from the ground.
For two centuries, civilization mistook the balloon for the destination. We polished the gondola, licensed the pilots and built entire institutions around producing people who could fly. We forgot the ascent was supposed to enable something. We started treating altitude as the point.
AI commoditizes the balloon. Anyone can ascend now; analytical altitude is available to everyone, on demand. The skill of getting up there, which is what the educated class was selected for, is losing most of its market value, and losing it fast, at least in its generic form.
What survives is what you do with the view. Whether you can actually see. Whether you can hold what you saw, bring it back down, commit to it and build with it. That isn’t balloon work. That’s what the balloon was always for.
Movement IV The Activities That Actually Matter
Ask what gives a human life meaning and you get answers like culture, exploration and sense-making: activities that connect us to something larger, engage our capacity for wonder, and let us take part in the long project of figuring out what it means to be here.
The modern yardstick translated these into System 2 terms. Culture became cultural production: books written, films made, papers published. Exploration became the discovery of new facts, territories and technologies. Sense-making became the construction of complete systems, theological or scientific. But these are translations. They were never the things themselves.
Real culture is the lived experience of belonging to a tradition: committing to local meaning, taking part in rituals and practices that give a life its shape. A person embedded in a demanding local tradition may experience more real belonging than a modern consumer with unlimited content, even when the embedded life is harder, less free and more constrained in every other way. The Marines I served with shared a thicker culture than most civilians experience in their adult lives, not because military life was better, but because it was bound. Culture is commitment, not content.
Real exploration is the embodied act of attending to what’s actually there. The first time you really see a sunset. The moment a piece of music opens something in you. The face of someone you love. AI can map the space of protein structures faster than any human lab. Whether it has experiences of its own is an open question I don’t need to settle here, because the argument doesn’t turn on it. The exploration that survives the AI epoch is the exploration of lived experience itself, the territory Rick Rubin and David Lynch spent their careers working.
This is where the easy reading of the argument fails. AI can already generate the artifacts of that territory: Rubin-style production, Lynch-style imagery, wabi-sabi-style poetry, Rothko-style color fields, the painted sunset. So retreating to feeling and intuition is no safe harbor. The machine produces things that move us faster than any human can.
The human edge isn’t some mystical residue of suffering that a listener’s nervous system detects in a human-made song. Sound waves don’t carry a certificate proving someone bled to make them. The edge lies upstream and downstream of the artifact.
Upstream is choosing. The machine can generate any piece of work on demand. It can’t decide that this piece is what this moment, in this culture or this situation, actually calls for. That choice runs on lived stakes: a person’s accumulated history of caring about specific things in specific ways.
Downstream is discernment. It is telling whether a piece of work is doing the real thing or only imitating its form, and refining that judgment over years by noticing what didn’t quite land. An AI-generated wabi-sabi poem may move you, but someone steeped in the tradition can tell whether it is true to the tradition or merely has its surface. AI can improve against a metric. It can’t revise what it cared about, because it has nothing riding on the answer.
AI can paint the sunset. It can’t care whether the sunset gets painted. It can’t tell whether this painted sunset matters in this moment, and it can’t refine its practice across decades according to what mattered and what didn’t. That’s the human edge. Not magic in the artifact. Stakes in the loop.
What changes in the AI epoch isn’t that System 2 becomes worthless. It’s that System 2 stops being the center of prestige and becomes support for taste, judgment, commitment and embodied meaning. The balloon stays in the budget. It just stops being mistaken for the destination.
Movement V Rubin, Lynch, Marcus
None of this is a new way of working. It’s an older arrangement, briefly forgotten. Three figures spread across nineteen centuries show the pattern.
Rick Rubin (born 1963) doesn’t write the songs and doesn’t play at the level of the artists he produces. He sits with the work and notices what is alive in it. He is known for steering by feel, asking for more of what is working without needing to explain why. Behind that apparent simplicity sit decades of pattern recognition operating below the level of words, backed by ruthless editorial discipline.
David Lynch (1946–2025) compared ideas to fish: the little ones swim near the surface, the big ones deeper down, and you don’t manufacture them, you get quiet enough to catch them. Transcendental Meditation wasn’t an aesthetic affectation for him; it was professional equipment, because the receiver is the instrument. Then came the scripts, the edits, the financing and the brutal mechanics of actually making a film.
Marcus Aurelius (121–180 CE) wrote the Meditations as a private notebook, addressed to himself, while running an empire and managing military campaigns. It reads less like a philosophical treatise than a practice log: the same few insights restated hundreds of times, in slightly different forms, across years. The historian Pierre Hadot argued that the repetition is the point. These were spiritual exercises, and repetition is how a principle moves from something you know into something you are. My own guess is that this is also why the book still reaches people: it was never written to be performed.
The pattern holds across cultures and centuries. The activities that make human life meaningful begin below explicit reasoning, and then require discipline, repetition and form to be realized. The modern project of running them through analytical frameworks alone was a productive detour that is now ending. What replaces it isn’t pure intuition. It’s the older order, restored: lived judgment on top, analytical reasoning as its instrument, and AI as a cheap, tireless source of volume underneath. The old knowledge economy put explicit reasoning at the top and treated intuition as raw material to be disciplined. AI makes the instrument cheap, which makes the deciding human commitment that directs it more important, not less.
Movement VI The Meta-Modern Operator
There’s a name for the stance that fits the epoch we’re entering: meta-modernism. Scholars have mapped it, most influentially the cultural theorists Timotheus Vermeulen and Robin van den Akker in their 2010 “Notes on Metamodernism,” but it hasn’t yet spread as a way of actually living. So let me give the working version.
It helps to see it as the third of three stances toward the big questions.
Modernism, roughly from the seventeenth century to 1945, believed that enough data and enough reason would eventually solve the human condition. Two world wars broke that faith. Gödel’s incompleteness theorems, often read far beyond what they actually prove, became the emblem of its limits.
Postmodernism, roughly from 1945 into our own century, answered that if certainty isn’t available, nothing is real. Its cultural mood kept company with nihilism and atomization.
Meta-modernism is the third move. It says the game is rigged, certainty isn’t available, and we’re going to play beautifully anyway. Build local truths: bounded, contingent and chosen rather than universal and proven, whether they operate at the scale of a tea ceremony or an empire. Commit to meanings while knowing they’re local in that sense. Use analytical sophistication in service of commitment rather than as a permanent escape from it. Ascend, get the view, come back down, and build with what you saw.
In practice it runs as a loop:
-
01Scan
Gather information until more gathering stops paying, or until time runs out, whichever comes first. The point isn’t completeness. It’s sufficiency.
-
02Acknowledge
Admit you don’t have a God’s-eye view and never will. The honest practice is to stop pretending you do, and to stop treating that as a reason not to act.
-
03Commit
Choose a value, a meaning: bounded, local, owned. Not proven. Chosen. Marcus would have called it serving Rome anyway.
-
04Execute
Act with conviction. The conviction is real even when the certainty isn’t. Hesitation contaminates everything downstream of it.
-
05Iterate
When reality pushes back, revise the commitment, with humility rather than ego. The discipline isn’t certainty. It’s correction without collapse.
This isn’t optimism, and it isn’t pessimism. It’s a third stance that most public argument hasn’t made room for yet. That is partly why arguments made from it can sound confused or contradictory to readers still working from modernist or postmodernist assumptions. A compressed version of the stance doesn’t survive the trip to a reader who has nowhere to put it.
Movement VII The Real Bet on AI
So here’s the take I should have written instead of the gloss I posted.
The AI epoch isn’t going to make life better in Pinker’s terms or worse in the doomers’ terms. It’s going to change which question matters.
The merger path Mo describes, staying at the bleeding edge by augmenting yourself with AI, is powerful but spiritually incomplete. It can raise your altitude. It can’t tell you where to land. People who take that path become faster Civics, optimizing for exactly the kind of cognition that just became cheap. Most people sense that this path isn’t satisfying, and the deeper reason isn’t lifestyle preference. The path is pointed along the wrong axis.
The real bet is the meta-modern one. Let AI commoditize the balloon. Let analytical altitude become cheap. Then come down, and do the work altitude was always supposed to make possible.
That calls for a particular role. I call it the Orchestrator. It isn’t the analyst trying to out-compute the machine; that path is closed. It isn’t the consumer passively taking in machine-made experiences; that is consumption, not creation. It’s a third position, and it has three layers.
The human owns the why and the frame. The machine handles the volume. The architectural work in the middle is where most of the leverage sits today, and it is the part most likely to be missed, both by accelerationists who think AI will handle everything and by doomers who think humans will be left with nothing. Over time I expect that middle layer to thin, which leaves the top layer as the durable human job.
Four practices follow.
Build things that transmit rather than explain. Stand in front of a Rothko long enough to feel what it’s doing; that’s the model. The painting doesn’t argue for anything. It puts you in a state. Good art goes around the analytical channel and delivers the experience directly. Much of the important work of this era will be building more things like that, games, rituals, communities, practices and places, that teach a way of living by letting people live it, not only by arguing for it in prose.
Commit to local meaning while knowing it’s local. Marcus knew Rome was mortal. He served it anyway. He wrote the Meditations with no thought of an audience, and they are still being read nineteen centuries later. That’s the Marcus move, and it scales: do the work as if no one is watching, including any future AI that might someday read what you wrote, because performing for an imagined audience corrupts the work.
Become a better receiver. Sensitivity to aesthetic, emotional and ethical resonance is the real instrument of the next epoch, and it is the one the analytical economy spent two centuries dulling. Most of us have spent decades blunting it through institutional, goal-driven work. Sharpening it again is the new skill. Then point the analytical power AI has made cheap at whatever that sharpened sense tells you actually matters.
Trust the descent. The view from altitude is real, but altitude doesn’t give it weight. Weight comes from coming back down and committing to specific, bounded realities, with honest humility about what we can know. The builders, the artists and the early adopters who learn to come down will quietly build the scaffolding for what comes next, in the gaps the old institutions leave as they lose their grip.