In 2025, a photographer won an international photography award with an AI-generated image. He returned the prize. The photo was beautiful. The debate it started was more important than any award.
The question wasn't really about one competition. It was about what "creative work" even means. Is it the output — the beautiful image on a wall? Or is it the process — the human struggle, the technical mastery, the choices made through hours of effort? If a machine can produce the output without the process, does the output still mean anything?
I've been sitting with this for a year. No clean answer. But I have an observation that might be more useful: AI didn't kill creativity. It killed a specific kind of creativity — the kind that lives in the middle. What remains on either side is more interesting than what was lost.
The Numbers Are Staggering
Midjourney has generated over 1 billion images since launch. One billion. The entire history of photography, from 1826 to today, produced roughly 12 billion photos total. A single AI tool generated 8% of all photographs ever taken — in two years.
Suno and Udio generate approximately 10 million songs per day. The entire Spotify catalog — everything ever released on every label in every country — is about 100 million tracks. AI music generators produce a new Spotify-equivalent catalog every 10 days.
Adobe's stock photography revenue dropped 35% in 18 months. Shutterstock's contributor earnings fell 40%. Getty Images is suing AI companies rather than competing with them — the litigator's admission of obsolescence.
The floor for creative work has fallen through completely. What used to cost thousands and days of skilled labor now costs pennies and seconds.
The Missing Middle
Here's the pattern, and it maps to what's happening with jobs broadly: AI destroys the middle tier of creative work. It doesn't touch the top. It barely affects the bottom. Everything between gets annihilated.
The bottom: hobbyists, amateurs, people creating for joy. They're fine. They were never competing commercially. AI gives them superpowers. A grandmother who always wanted to paint but never had the skill can now generate exactly the image she sees in her mind. A teenager with melodies in her head but no music theory can produce the song. Democratization of creative tools is real and good.
The top: elite artists, auteurs, cultural icons. More than fine — thriving. When everything is possible at zero cost, scarcity moves upstream from production (anyone can produce) to vision (few can see). The best human artists are MORE valuable because they're the signal in infinite noise.
A Beyonce album in a world of 10 million daily AI songs isn't less valuable — it's more. A hand-painted oil portrait in a world of 1 billion AI images isn't obsolete — it's rare. A Christopher Nolan film in a world of AI-generated video isn't redundant — it's an event.
The middle: this is where the carnage is. Stock photographers. Session musicians. Commercial illustrators. Corporate video producers. Mid-tier graphic designers. People whose value was "competent execution of a creative brief at a reasonable price."
That value proposition is dead. Not dying. Dead. AI execution isn't "reasonable price" — it's free. And it's not just competent — it's good. Eighty-percent-as-good-as-a-professional good. For a corporate blog header or a podcast intro jingle, that's more than enough.
What Creativity Means When Production Costs Zero
For all of human history, creative work required two things: vision (knowing what you want to make) and execution (being able to make it). The marriage of these was what we called "talent."
AI has divorced them.
You no longer need execution ability. You need vision. The ability to describe what you want, evaluate whether what you got matches what you imagined, iterate, refine, direct.
This changes who gets to be "creative."
Under the old model, a film director needed to understand cinematography, lighting, editing, actor management, sound design — or at least speak the language well enough to direct specialists. Under the new model, a director needs to know what a good film looks like. The execution is handled.
Under the old model, a musician needed to play instruments, understand theory, master recording software. Under the new model, a musician needs taste. Knowing what sounds right, what feels right, what serves the story.
This isn't lowering the bar. It's removing the bar entirely from one dimension (execution) and raising it infinitely in another (taste). Taste — the ability to distinguish between good-enough and genuinely great — turns out to be rarer and more valuable than technical execution ever was.
What the Hollywood Writers' Strike Taught Us
When the WGA went on strike in 2023, AI-generated scripts were a central issue. The writers weren't afraid AI would write better scripts. They were afraid studios would use AI to generate first drafts, then hire writers at reduced rates to "polish" them — turning authorship into revision.
They were right to be afraid. That's exactly what was planned.
But here's what emerged: AI scripts are mediocre in a specific way. Structurally sound. They hit plot points on schedule. They're also completely devoid of the idiosyncratic, personal, weird choices that make great writing great.
An AI can write a script that follows the Save the Cat structure perfectly. It cannot write Fleabag. It cannot write the moment in Breaking Bad where Walter White watches Jane die. It cannot generate the specific human truth that makes you gasp — the thing that comes not from pattern recognition but from one person's particular experience of being alive.
The strike ended with protections for writers. But the real protection isn't contractual. It's capability. The things humans write that AI cannot are the things that matter most. The catch: "the things that matter most" are also the rarest and hardest to produce on a corporate schedule.
The Taste Economy
Here's my thesis, stated directly: we're moving from a production economy to a taste economy for creative work.
Production economy: value comes from making things. Bottleneck is capacity. Produce more, win.
Taste economy: value comes from choosing things. Bottleneck is judgment. Choose better, win.
The music producer who wins in 2026 isn't the one who can play every instrument or mix a track perfectly. It's the one who can generate 50 AI variations and pick the one that makes people feel something. The one who can direct a machine toward emotional truth.
The visual artist who wins isn't the one with the steadiest hand. It's the one who can describe what doesn't yet exist in terms the machine can translate — and then knows whether the output matches the vision. And can iterate until it does.
This is simultaneously more democratic and more elitist than before. More democratic because anyone can now produce. More elitist because taste isn't teachable the way technique is. You can practice guitar for 10,000 hours and measurably improve. How do you practice taste?
The Emotional Truth Problem
Here's where I get genuinely uncertain.
Great art moves us because it contains emotional truth. A Dorothea Lange photograph captures something about the human condition. A Bob Dylan song articulates an experience you've felt but never named.
This emotional truth comes from lived experience. Suffering. Joy. The particular texture of one human consciousness encountering reality.
AI has no consciousness. No experience. No suffering. It has patterns learned from expressions of human consciousness. It can recombine those patterns convincingly — sometimes startlingly so.
But is a recombination of emotional truth itself emotionally true? When AI generates a poem about grief by pattern-matching across thousands of human grief poems, is the result grief? Or is it a convincing simulation that lacks the actual wound behind the words?
I don't know. But audiences consistently report different emotional responses to art they know was human-made versus AI-generated. The context matters. Knowing a human being poured their pain into something changes how we receive it.
This may be the last moat. Not technical execution. Not even taste. But the simple fact of having lived, suffered, loved, lost — and made something from that irreplaceable experience.
What Survives
Enough philosophy. What creative careers actually survive?
Creative direction. Setting vision, defining tone, maintaining consistency across a project, deciding what's right versus almost-right. Pure taste. AI can't do it because it doesn't have preferences — it has probabilities.
Curation. When infinite content exists, filtering signal from noise becomes essential. DJs, editors, gallery owners, playlist makers — anyone whose job is "from everything that exists, here is what deserves your attention."
Performance. A live concert, theater, comedy — unreplaceable not because of sound or words but because of shared human presence. You're in a room with a person who is alive and risking something in front of you. AI cannot risk. Cannot fail in front of an audience in a way that makes success meaningful.
Personal expression. Your sketchbook, your garage band recordings, your memoir. Value comes not from objective quality but from the fact that it's YOURS. More people expressing themselves more fully is unambiguously good.
The truly exceptional. Transcendent work that changes culture, redefines a medium, captures a moment in civilization. It remains human because it requires the one thing AI lacks: a point of view earned through the irreducible experience of being alive.
The Honest Conclusion
AI has ended the era where creative competence was economically sufficient. Being good at making things is no longer enough if "good" is what the machine does for free.
What remains: being great. Being weird. Being specific in a way no dataset can reproduce because it comes from one unrepeatable human life.
The middle is gone. The floor has fallen away. But look up — the ceiling was never higher. The best creative work of the next decade will be made by humans wielding AI as an instrument, not replaced by it, not competing with it, but using it in pursuit of visions only a human could have.
The photographer who returned his prize understood something. The image was beautiful but it didn't cost him anything. Not time, not struggle, not the hundred failed shots that make the one great one meaningful. And meaning, it turns out, is the thing AI cannot generate. Not because it lacks capability, but because meaning requires a subject — a someone for whom things mean something.
We are that someone. For all the power of these systems, they have no "for whom." They produce beauty with no beautiful, music with no listener, stories with no teller.
In a world drowning in effortless beauty, the thing that will be most valued — maybe the only thing — is effort itself. The human choosing to make something the hard way, not because the machine couldn't, but because the struggle is the point. Because the wound is the art. Because being alive and making things from that aliveness is what we are.
---
Keep Reading
- The AI Job Apocalypse Nobody Wants to Talk About — the "missing middle" pattern hits every industry, not just creative ones
- Vibe Coding Income Guide — when AI handles execution, directing it becomes the skill that pays
- Lovable vs Bolt vs Cursor vs Claude Code — the creative tools reshaping who can build what
- AI YouTube Channels Worth Following — where creators are thriving despite (and because of) AI
