Let the beauty we love be what we do. There are hundreds of ways to kneel and kiss the ground.
The Creator's Renaissance
Twelve thousand songs in three years. The Renaissance atelier returning, with cheaper apprentices. What becomes scarce when execution becomes free.
Chapter 10 — The Creator's Renaissance
Y poesía llegó
en busca de mí.And poetry arrived
in search of me.— Pablo Neruda, La Poesía, Memorial de Isla Negra (1964)1
Let the beauty we love be what we do.
There are hundreds of ways to kneel and kiss the ground.— Jalāl ad-Dīn Rūmī, 13th c.2
The night of the thousandth song
I remember the exact week. It was late autumn, the kind of light Amsterdam gets when the canals turn slate, and I was sitting at the same desk in the same apartment where I am writing this chapter, and the catalog hit one thousand.
A thousand songs in just under six months.
I had not planned for this. I had not, in my entire prior life as a music listener and occasional amateur composer, written one finished song. The accumulated total of my musical output, before the night I downloaded Suno on a whim, was a half-finished piano piece from when I was nineteen and a Garageband demo I had abandoned in 2014. Now, six months in, the number on the catalog page said 1,003.
The strange thing was not the number. The strange thing was that, scrolling through the catalog at 11 PM with a glass of wine, I knew which songs were mine. There were twenty-something tracks, scattered across the thousand, that I would have called my work in any honest sense — that had something I had specifically wanted in them, some seen thing or felt thing or remembered thing pressed into a particular arrangement, and that the world would not have if I had not been the one prompting that night, with that question, in that mood, after that conversation, in that grief or that lightness.
The other 980 were exercises, sketches, genre studies, voice tests, technical experiments, the residue of three thousand iterations that had been necessary to find the twenty-something. I do not regret them. They were the apprenticeship. But they were not, in the deep sense, mine.
What I noticed that night was that the structure of being a creator had changed.
Before the models, the bottleneck was making. A song took weeks to compose, days to record, hours to mix, more hours to release. The ratio of inner-events-experienced to outer-artefacts-produced was extreme — a serious composer might finish, in a year, a few completed pieces. Of those, only a fraction would be remarkable. The rest of what passed through the composer's interior was lost.
After the models, the bottleneck is not making. The bottleneck is recognizing.
Ninety-eight percent of what comes back from a model is forgettable. Two percent has the small, hard-to-name quality that makes a piece worth keeping. The skill of the new artisan is the skill of the curator — running the loop fast, recognizing the alive ones, killing the dead ones, refining the prompt, returning. The making is cheap. The seeing is expensive.
This is the Renaissance, and it is the second one in five hundred years. We are still learning what it asks of us.
What the first Renaissance was actually like
We tell the story of the first Renaissance as if it were principally about masters. Michelangelo, Leonardo, Botticelli, Raphael — solitary geniuses producing astonishing works one at a time, with the rest of the population presumably watching from below.
This is not how the first Renaissance worked.
A Renaissance studio of the late fifteenth century in Florence or Venice was an atelier — a workshop with a master, several journeymen, and a constant stream of apprentices. The master might receive a commission and design the composition. The studio executed it. Underdrawings might be by one apprentice; the figure of the second saint by another; the drapery by a journeyman; the master's hand might be visible only in the central faces and the most demanding passages. Major works were almost always collaborations of this kind. Verrocchio's Baptism of Christ, in which the young Leonardo painted one of the angels, is a famous and well-documented example. The norm was atelier production.
What gave the master his title was not that he did all the work himself. It was that he directed the studio, trained the apprentices, designed the compositions, and brought the whole to a level of integrated excellence that no individual hand could have produced alone.
This has the structure that anyone working with agentic AI in 2026 will recognize.
The frontier-model era is, in its present form, a return of the atelier. The operator is the master. The models and tools are the studio. The apprentices are the agents — sub-models, specialised tools, custom assistants, scheduled workflows — who do the journeyman work under direction. The operator's job is the master's job: to design the composition, recognize what is alive, refuse what is dead, raise the integrated whole to a level of excellence that none of the components could produce alone.
The Renaissance studios produced more work, of higher quality, in more variety, at lower individual cost, than any prior model of art-making. They produced the cathedral programs, the great altarpieces, the civic murals, the portrait industry of an entire merchant class. They did so by industrializing the creative process around the figure of the directing master.
We are doing the same thing now, with cheaper apprentices.
What changes when execution becomes cheap
The economic structure of creative life depends, in any era, on the relative price of the inputs. When physical pigment was the constraint, a portrait was a luxury. When photography arrived, the portrait industry transformed in a generation. When recording technology arrived, the music industry transformed in a generation. When digital reproduction arrived, the publishing industry transformed in a generation. Every major shift in the price of an input restructures who can make what.
The model era is the largest such shift in centuries.
The cost of generating a competent piece of writing has fallen, in five years, from hours of human time to seconds of model time. Same for an illustration. Same for a serviceable melody, a serviceable section of code, a serviceable narration, a serviceable summary, a serviceable rough draft of almost anything that can be expressed in text. The price has not fallen to zero — there is still electricity, infrastructure, subscription cost — but it is, as a function of human attention, close enough to zero that it is no longer the binding constraint.
What becomes scarce when execution is cheap?
Taste becomes scarce. The capacity to recognize, in a flood of generated outputs, which two percent are alive — this is not automatable, it does not scale linearly with money, and it is the rate-limiting input of the entire creative pipeline now. The operator with strong taste produces vastly more value than the operator with weak taste, even when both have access to the same models.
Specificity becomes scarce. The capacity to bring particular seen things into the prompt — the writer's eye, again — is the difference between art and the average. The model's outputs reflect the prompter's specificity. Generic prompts produce generic outputs and the world is already drowning in them.
Persistence becomes scarce. The willingness to run the loop a thousand times, to throw away nine hundred and seventy of the outputs, to keep coming back until the right one lands — this is the apprenticeship the new tools demand. Most people, faced with a fast generative loop, run it ten or twenty times and decide it doesn't work. The serious operator runs it a thousand.
Voice becomes scarce. The accumulated, particular, recognizable signature of a single creator — built from years of choices about what to make and what to refuse — is the one thing the models structurally cannot manufacture. They can recognize voice; they cannot have it. Voice is the residue of a specific life, lived deliberately, expressed across a body of work. It cannot be averaged. It cannot be borrowed. It can only be cultivated.
Curation becomes scarce. With the price of generation collapsed, the value chain shifts toward whoever can sift the river. A great curator, in the new economy, is closer to a great editor than a great producer. They are the people whose taste decides what a community sees. The economic and cultural premium on great curators is going to be one of the defining features of the next decade.
The aggregate effect is paradoxical. The cheaper execution becomes, the more valuable the specifically human becomes. The cheaper outputs become, the more valuable the irreducibly personal substrate that prompts them is. The cheaper images and texts and tracks become, the more valuable the originating sensibility — the eye, the ear, the conscience — that shaped them.
This is the great open secret of the Renaissance currently underway. The dread of being replaced is being reported most loudly by people who have not yet noticed that the replacement is going the other way: the lazy generic is being replaced by the cheap generic; the specific human is being elevated by the new amplification. The first group is loud and frightened. The second group is quietly making things.
Music as the leading edge
I have spent more time on music than on any other generative domain, and I want to be specific about what the experience has actually been, because most public commentary on AI music is either fearmongering or breathless hype, and the reality is more interesting than either.
Music as a medium has properties that make it the leading edge of the generative wave in some respects. Music is short — a typical piece runs three to five minutes. Music is iterable — generating a hundred drafts takes a day, not a year. Music is felt — the difference between alive and dead, in a song, is registered in the body within seconds, which makes the curation loop fast. Music has a long tradition of micro-genre specialization, which gives the prompter a vocabulary of particulars to draw on. And music is consumed casually — most listening happens in the background, which means that even imperfect generative work has a use case (workouts, focus, reflection) that text or video does not have to the same degree.
What I have learned in three years and twelve thousand tracks is, roughly, the following.
The model is most powerful when the prompter has a specific emotional or sensory target. A vague request for "a chill song" returns the average. A request for "the feeling of returning to your old neighbourhood after fifteen years away, on a winter afternoon, the streets emptier than you remember, an old jazz radio playing through someone's open window" returns something with a chance of being alive. The prompt is the seen thing.
The catalog is more important than any single track. A creator with one viral hit and no body of work is a fragile creature in the new economy. A creator with twelve thousand tracks, of which five hundred are good and fifty are excellent, has built an instrument no individual track could replace. The catalog is the artist. The tracks are its facets.
State-engineered music is its own emerging genre. The frequency entrainment work I described in Chapter 5 is a real, durable use case, and it cannot be served by any pre-existing music industry — there is no Spotify catalog of "a forty-minute 40 Hz binaural focus piece in the harmonic register I find pleasant in late afternoon". Generative tools make this kind of bespoke composition trivial, and a meaningful subset of my catalog now serves this function for me and for friends who use it. The genre will scale.
The hardest skill is restraint. With the price of generation near zero, the temptation is to make more. The discipline is to make less, choose better, and refine the small set that matters. Most catalogs are too large to be useful. Mine, at twelve thousand, is past the point where any human listener could go through it linearly. The next phase of work is curation — building views, playlists, sequences that surface the alive ones. The making is done. The seeing has just begun.
I share these specifics so the reader has a concrete picture of one operator's three years inside the wave. The patterns generalize beyond music. They will become the common patterns of every creative domain over the next half decade.
The new artisans
What kind of person flourishes in this Renaissance?
I have come to believe it is a particular type, with a recognisable set of traits, and the type has existed in every era. The Renaissance does not invent the type; it amplifies it.
The new artisan has patience for the loop. They will run the prompt a thousand times. They are not bored by repetition because they understand that the iteration is the practice. They have made peace with the fact that ninety-eight percent of any given session is not the work; the work is the two percent that survives the curation.
The new artisan has opinions. They like specific things, refuse specific things, and have come to know themselves well enough to recognise their own voice when it appears in the river. The most common failure mode for newcomers to generative tools is the absence of strong opinions — without them, every output looks fine and nothing is alive.
The new artisan reads broadly and deeply. They have a stocked imagination — references, images, sentences, songs, films, paintings, ideas — that the models can draw on through the prompts. The barren imagination produces barren prompts. The stocked imagination produces prompts that are themselves small works.
The new artisan publishes. The catalog is built only in public. Work that lives in the drafts folder does not refine the artisan's taste, because it never meets the world. The new artisan releases regularly, accepts the small rejections that follow, and lets the slow accumulation of made things shape who they are becoming.
The new artisan teaches. The most generative creators of the new era are those who teach what they are learning, in real time, to anyone who is interested. The teaching deepens the practice. It also builds the audience that, eventually, becomes the community that supports the work. The artisan who hides their methods is, in this era, leaving on the table an order of magnitude of compounding effect.
The new artisan honours the lineage. The capacity to make in the new way is built on the foundations laid by every creator who came before. The artisan who scorns the past has nothing to draw from when the model returns the average. The artisan who has read Rilke prompts differently than the artisan who has read only the timeline. The lineage is not a museum — it is the substrate from which the new work emerges.
These traits are unequally distributed. They were unequally distributed in the first Renaissance. They will be unequally distributed in this one. But — and this is the difference that matters — the entry cost has fallen so far that the type itself is no longer constrained by class, geography, or apprenticeship. A teenager with a Lenovo and a curiosity can, today, begin the practice that previously required a decade of training and the patronage of a wealthy court.
The Golden Age of Intelligence is, among other things, a vast widening of who gets to make. The democratization is real. The amplification is real. The cost is the patience to do the apprenticeship — which is, and always has been, the only price the work asks.
What you make is what you become
There is one more dimension of this Renaissance that the productivity discourse around AI almost universally misses, and it is the dimension I find most worth holding onto.
What you make shapes who you are.
The making is not separable from the becoming. The athlete who trains every morning becomes a different person across the years, in ways that are not reducible to the medals. The musician who practices three hours a day becomes a different person, regardless of what audiences ever hear. The writer who writes a thousand words a day becomes a different person, even if the manuscripts stay private. The substrate of who you are is built, in part, by the substrate of what you make.
When the cost of making collapses, this dimension does not disappear. It changes shape. The people who use the new tools to make things they care about will be shaped by the making, in the way the Renaissance studios shaped their masters. The people who use the new tools to make a high volume of generic content will also be shaped — but in a different direction, and not for the better.
The choice of what to make matters more, not less, in the era of cheap making. The Renaissance studios chose ambitious works because the directing masters had specific things they wanted to bring into the world. The model era will be defined by the same choice: do you use this leverage to bring something specific into the world, or do you use it to pile up the generic faster?
The tools do not care. The aggregate of your making is, however, who you become.
I keep returning, in my own life, to this observation. The twelve thousand songs are not just a catalog. They are three years of who I have been becoming. The decisions about which prompts to write, which iterations to keep, which directions to refuse — these decisions are not just craft decisions. They are decisions about character. The catalog has shaped me as much as I have shaped the catalog.
This is the part the discourse about AI as a productivity tool gets badly wrong. AI is not a productivity tool. It is a making tool, and what we make returns into us. The Renaissance is not just an economic event. It is a substrate event. We are about to become a different species — gradually, partially, unevenly — through what the new instruments allow us to make and what the making does to us.
The question is not, what will AI do to us? The question is, what will we make with it, and who will that make us?
A practice for this season
If you have not yet made anything with these tools, make one thing.
Pick the medium that calls you most. Writing, image, song, code, video, voice. Choose the simplest possible project — a single artefact, with a clear shape, that you would be willing to put in front of one trusted person.
Open the model. Write a specific prompt. Iterate. Throw away nine outputs of every ten. Keep going until you have one piece you would sign your name to.
Sign your name to it. Show it to one person.
Then make another. And another. And don't stop.
The catalog is built one piece at a time. The artisan is built one piece at a time. The Renaissance is built, in this generation as in the last, by the people who decide to receive it and start.
Welcome to the workshop.
The light is good.
There is a place at the table for you.
Hand-off
The making is one half of the operator's life. The other half — quieter, less photographed, more demanding — is the discipline of being someone whose making is worth amplifying.
In Chapter 11 we walk into the room of personal governance. The Stoic interiority that the generative era is going to need at scale. Attention as a finite, ethically charged resource. The cost of unmanaged attention in the era of infinite content. The simple disciplines that keep an operator human while the instruments multiply.
Without this room, the studio breaks.
We are walking in.
Footnotes
- Pablo Neruda, La Poesía, opening of Memorial de Isla Negra, 1964. The full opening: Y fue a esa edad… Llegó la poesía / a buscarme. Translation mine. ↩
- From Coleman Barks's translation of Rumi, The Essential Rumi, 1995, p. 36 (the famous Hundreds of Ways). The Persian original is from the Divan-i Shams. ↩