There’s a recurring pattern in creative history that we keep failing to notice while we’re living through it: every time a machine gets good at producing the expected, humans respond by producing the unthinkable.
We’re in the middle of one of these moments right now. And if history is any guide, what comes next won’t be the death of human creativity, it’ll be a violent, exhilarating explosion of it.
When Photography Killed Painting (And Painting Got Interesting)
In 1839, the daguerreotype arrived and painters collectively panicked. For centuries, a painter’s bread and butter was the faithful reproduction of reality, portraits, landscapes, historical scenes rendered with painstaking accuracy. Photography did all of that in seconds, and it did it better.
The establishment mourned. Paul Delaroche reportedly declared, “From today, painting is dead.” But what actually died wasn’t painting, it was painting’s obligation to be a mirror.
Freed from the burden of documentary realism, painters started asking questions they’d never had permission to ask. If a photograph could capture a haystack perfectly, what was the point of Monet painting one? The answer turned out to be: to show you what a haystack feels like at 4pm in October when the light is dissolving. Impressionism wasn’t a retreat from photography. It was a response to it, an assertion that human perception contains something a lens cannot capture.
Then things got properly weird. Cubism shattered the picture plane entirely. Picasso and Braque stopped trying to show you what something looked like from one angle and instead tried to show you what it looked like from every angle simultaneously. Photography had claimed the monopoly on faithful representation, so the Cubists abandoned faithfulness altogether. Fauvism threw out naturalistic colour. Expressionism threw out emotional restraint. Abstraction threw out the object itself.
None of these movements would have had the urgency they did without the pressure of the camera. Photography didn’t kill painting. It killed mediocre painting. It killed the painter who had nothing to offer beyond technical reproduction. And in doing so, it forced the painters who remained to become more inventive, more radical, more human than they’d ever needed to be before.
The Suno Provocation
Now watch the same pattern unfold in music, in real time.
AI music generation has reached an unsettling level of competence. Suno, the dominant platform in this space, has over 100 million creators and generates complete, polished songs from text prompts. Describe a mood, pick a genre, and you’ll have a studio-quality track in under a minute. It supports over 1,200 musical styles. The output is, and this is the uncomfortable part, pretty good. Good enough that a significant amount of background music, content creation audio, and demo tracks are now AI-generated. The musical equivalent of stock photography.
And just as stock photography didn’t kill fine art photography, AI-generated music isn’t killing music. But it is doing something interesting to the landscape.
Enter Angine de Poitrine. A duo from Saguenay, Québec; Anonymous, performing in papier-mâché alien heads and polka-dot bodysuits, playing microtonal math rock on a custom-built double-neck guitar that the creator literally modified with a saw. Their KEXP session has surpassed ten million views. They have 1.6 million monthly Spotify listeners. Their music is built on quarter-tones borrowed from Indian and Japanese traditions, time signatures that seem designed to make your brain itch, and a level of instrumental virtuosity that is, by definition, inimitable.
And it’s going viral. Microtonal math rock is going viral.
The cynical read is that this is just the internet’s usual appetite for novelty. But I think something deeper is happening. In a world where anyone can prompt a convincingly “good” pop song into existence, audiences are subconsciously recalibrating what they value. Competence is becoming cheap. So the premium shifts to the things AI cannot fake: genuine strangeness, physical virtuosity, idiosyncratic vision, the unmistakable evidence of a specific human behind the work.
One reviewer captured the sentiment perfectly:
in the ever-growing flood of AI-generated content, Angine de Poitrine’s music represents a scream of humanity, a deep thumbprint pressed into the sound by its creators.
The appeal isn’t despite the difficulty and weirdness. It’s because of it. The difficulty is the proof of authenticity.
The Pattern
Zoom out and the pattern crystallises:
Automation commoditizes the middle. Photography commoditised realistic depiction. AI music commoditizes competent songwriting. In both cases, the thing that was previously hard enough to be impressive becomes trivially reproducible.
The extremes become more valuable. When the average is freely available, the only differentiating positions are the extraordinary and the deeply personal. This creates pressure in two directions, toward unprecedented technical mastery and toward radical creative divergence.
The avant-garde accelerates. Movements that would previously have been too niche to sustain themselves find audiences, because audiences are hungry for the un-fakeable. Cubism went from a Parisian curiosity to a dominant art movement in under a decade. Angine de Poitrine went from Quebec bar gigs to millions of streams in months.
This isn’t a coincidence, and it’s not limited to the arts.
Software Engineering’s Impressionist Moment
We are, right now, living through the photography kills painting moment in software engineering. And most of us haven’t realised it yet.
Vibe coding, the practice of describing what you want in natural language and letting AI generate the code, has gone from a joke coined by Andrej Karpathy to the way 72% of developers work daily. By some estimates, 41% of all code is now AI-generated. Tools like Cursor, GitHub Copilot, and Claude Code have made it possible for a single developer to produce working software at a pace that would have required a team two years ago.
This is photography arriving in the painting studio.
The immediate reaction is the same fear painters had: if AI can write code, what’s left for us? And just like with painting, the answer is: everything that matters.
What AI code generation is commoditizing is the mechanical work of software, the boilerplate, the CRUD operations, the glue code, the standard patterns that experienced developers could write in their sleep. It’s the equivalent of photography capturing accurate likenesses. Necessary, but not where the real craft lives.
What AI cannot do, at least not yet, and perhaps not ever, is the work that requires deep contextual understanding, architectural vision, and the kind of creative problem solving that comes from truly understanding a domain. AI can generate a login form. It cannot decide whether your product should have a login form at all. It can write a caching layer. It cannot feel the operational pain that tells you where your system actually needs one.
This is where it gets exciting. Because if the mechanical work is handled, what does the software engineer of the near future actually do?
They become the Cubists.
What Software Cubism Looks Like
If the pattern holds, and I believe it will, we should expect a wave of avant-garde innovation in software engineering. Not immediately, and not from the places you’d expect. But the conditions are forming.
Radical architecture. When building standard architectures is trivially cheap, there’s less risk in experimenting with unconventional ones. Expect more exploration of exotic paradigms, systems built around event sourcing not because it’s fashionable but because someone had the headroom to try it; applications structured around novel concurrency models; infrastructure that would previously have been too expensive to prototype. The equivalent of painters abandoning perspective: engineers abandoning assumptions about how software “should” be structured.
Deeply personal tools. When anyone can vibe code a standard SaaS app, the value shifts to software that embodies a specific vision of how work should be done. Think of it as the difference between a stock photograph and a Cartier-Bresson. Both capture a street scene. Only one has a point of view. We’ll see more tools that are opinionated to the point of being polarising, and that’s exactly what will make them valuable.
The return of the polymath. When code generation is no longer the bottleneck, the limiting factor becomes what you know about the problem domain. The most valuable engineers will be the ones who combine deep technical skill with deep domain expertise, the developer who also understands genomics, or logistics, or music theory, or urban planning. These are the people who can see solutions that a prompt-driven workflow would never surface, because the prompts emerge from a knowledge base that includes more than just programming.
Craft as signal. Just as Angine de Poitrine’s hand-built microtonal guitar is a statement of authenticity, we’ll see software where the craft becomes the point. Systems where the architecture is so thoughtfully designed, the abstractions so elegant, the performance so carefully tuned that the engineering itself becomes the differentiator. Not because AI couldn’t have produced something functional, but because the human choices embedded in the design create something AI wouldn’t have thought to create.
The Uncomfortable Middle
This transition isn’t painless. Photography didn’t just inspire the Impressionists, it also destroyed the livelihood of thousands of perfectly competent portrait painters. The avant-garde flourished, but the middle hollowed out.
The same is happening now. If you’re a developer whose primary value proposition is reliably producing standard code to well-defined specifications, the economics of your role are changing fast. Not because you’re bad at your job, but because the definition of what constitutes “the job” is shifting underneath you.
But here’s the crucial distinction: the portrait painters who survived weren’t the ones who tried to paint faster than a camera. They were the ones who learned to paint differently. John Singer Sargent thrived in the age of photography not by competing with it on accuracy, but by bringing a psychological intensity to portraiture that no camera could match.
The developers who’ll thrive aren’t the ones who’ll try to code faster than AI. They’re the ones who’ll learn to think in ways AI can’t replicate, who’ll bring the kind of lateral, contextual, aesthetically driven thinking that transforms a working system into an elegant one.
The Provocation
So here’s my challenge to every software engineer reading this: AI is about to hand you the same gift that photography handed the painters of 1860. The drudge work is being automated. The mechanical reproduction is being handled. You are being freed.
The question is: freed to do what?
If your answer is “the same things, but faster,” you’ve missed the point. The Impressionists didn’t paint faster. They painted differently. The Cubists didn’t draw more accurately. They drew impossibly.
The opportunity in front of us isn’t to produce more software. It’s to produce software that couldn’t have existed before, software that’s stranger, more ambitious, more deeply considered, more human than anything we’ve built in the era when most of our energy went into fighting the compiler.
The automation avant-garde is coming. The only question is whether you’ll be in it.