The Paintbrush Principle: What AI Songwriting Actually Looks Like (And Why Most People Get It Wrong)

When people hear "AI-generated music" they picture one thing: someone typing "write me a hit song" into a chatbot and pressing enter. A machine spitting out lyrics, melody, production — the whole thing — while a human sits back and takes credit.

That's not what I do. That's not even close. And the gap between what people think AI music-making is and what it actually is — at least when it's done with any real intent — reveals everything about where AI is going and who will matter in that future.

I just finished an album called HALLUCINATE. Nineteen tracks. A concept album about the progression from blame to integration — twenty years of lived experience distilled into sound. It's on Spotify under Rodington Pear. AI was involved in every track. And AI wrote exactly none of it.

Let me explain what I mean.

The Paintbrush Principle

A paintbrush doesn't paint. A painter paints. The brush is the tool that translates what's inside the painter into something visible. Nobody looks at a Rembrandt and says "wow, that brush was talented."

AI is my paintbrush. That's it.

Before AI, I'd spend hours — days — fighting with production. EQ. Compression. Mixing. Sound design. DAW routing. All the technical scaffolding that sits between having something to say and actually getting it recorded. I'm not a trained producer. So that technical layer was a wall. Not an insurmountable wall, but a friction layer that ate my time, my energy, and sometimes killed the creative impulse entirely by the time I'd finished wrestling with the tools.

AI removed that wall.

And here's what happened when the wall came down: I went straight to what actually matters. The lyrics. The meaning. The feel. The content. The thing I was trying to say.

I didn't become a worse musician. I became a more honest one. Because all the energy I used to waste on technical production — energy that had nothing to do with the actual art — was now freed up and redirected into the creative core.

My Actual Process (Raw and Unfiltered)

Here's what it really looks like.

I start with a feeling. An idea. A phrase. Sometimes it hits me in the gym, sometimes at 2am, sometimes mid-conversation. When it comes, I move fast. I open a chat with Claude and I start writing. And I mean writing — stream of consciousness, typos everywhere, half-formed thoughts crashing into each other, spelling mistakes piling up because my fingers can't keep pace with my brain.

It looks like a mess. It looks like someone had a seizure on a keyboard.

And every single word comes through.

That's the thing people don't understand. The typos don't matter. The grammar doesn't matter. What matters is the signal — the raw energy, the emotional truth, the thing I'm actually trying to express. AI understands it perfectly. Not because AI is smart, but because when the signal is strong enough, surface-level noise is irrelevant. You could misspell every other word and the meaning would still punch through.

I discovered this by accident. I was having a conversation about philosophy and AI — typing fast, not correcting anything, raw stream of thought — and realised the ideas were landing harder because I wasn't self-editing. The imperfection was carrying energy that polished text would have killed. I was practicing my own theory without even realising it.

So that became the process. Pour it out raw. Let the energy lead. Don't clean it up. Don't second-guess. Twist the lyrics until they surprise me — take a line and bend it somewhere unexpected, somewhere that feels wrong at first and then clicks into something truer than what I originally intended. Use AI to handle the production, the arrangement, the technical layer. But the words, the meaning, the emotional architecture — that's all mine. Every syllable.

And it's not just the lyrics. It's the taste I bring to the sound itself. My musical DNA is a mess — in the best way. 80s TV theme tunes. Game console soundtracks. Disney scores. Classical music. Hip hop. Drum and bass. Fado. All of it absorbed over decades, all of it feeding into what I hear in my head when I'm writing. No AI trained on genre-specific datasets would ever arrive at the combinations I'm reaching for, because my references are too weird, too personal, too cross-pollinated.

Here's an odd detail that might mean something: a few months before all this creative inspiration hit, I joined a batucada — a Brazilian percussion group. One of the most raw, primitive musical forms there is. No technology. No production. Just humans hitting drums together until something locks in. I'm doing that at the same time as making AI-assisted music at the cutting edge. The most ancient and the most futuristic, running in parallel. I don't think that's a coincidence. I think the primitive fed the futuristic. The raw human energy of batucada is in HALLUCINATE somewhere, even if you can't point to exactly where.

The Levels (And Why Most People Are Stuck on Level One)

Not all AI-assisted creation is equal. There are distinct levels, and most people can't even see past the first one.

Level One: The Prompt Merchant. "Write me a song about love." "Make it sound like The Weeknd." This is what 99% of people think AI music is. You give the machine a brief, it generates output, you publish it. The human contribution is a sentence. The result sounds fine. It sounds like everything else. It has no heartbeat. This is the default-accept mode I talked about in my other writing — letting AI do its generic thing and calling it yours.

Level Two: The Editor. Better. You generate something, then you shape it. You push back. You say "no, darker" or "change the second verse" or "that chorus is too clean." You're curating. You have taste. But the raw material is still AI's. You're sculpting someone else's clay.

Level Three: The Amplifier. This is where it shifts. The raw material is yours. The words, the ideas, the emotional core — they come from your lived experience, your specific perspective, your accumulated years of actually being alive and paying attention. AI amplifies what's already inside you. It handles the execution layer so you can focus entirely on the creative layer. The paintbrush principle in action.

The difference between Level Two and Level Three might sound subtle. It's not. It's the difference between a cover band and a songwriter. Between someone who can arrange existing ingredients and someone who grows their own.

Level Four: The Synthesiser. This is where it gets rare. You're not just bringing personal experience to one domain. You're bringing cross-domain pattern recognition — seeing connections between architecture and music, between philosophy and technology, between crypto markets and creative process. The AI becomes a tool for expressing a web of understanding that no specialist in any single field would generate. This is where the truly original work comes from. Not because the tools are better, but because what's being fed into them is richer, stranger, more interconnected.

Most people will never get past Level One. Not because they can't, but because Level Three and Four require something AI can't provide: a life actually lived. Experiences accumulated. Mistakes made. Domains crossed. Wisdom earned the slow way.

The Energy Transfer

There's something else happening in my process that I think matters, and I haven't heard anyone else talk about it.

When I type raw — fast, messy, typo-riddled, unfiltered — there's an energy in that text that polished writing doesn't carry. It's the written equivalent of a live recording versus a studio recording. A live recording has mistakes, feedback, crowd noise, the singer slightly out of breath. And it hits different. Because you can feel the liveness. The presence. The fact that a human being was right there, in that moment, fully committed.

My prompts read like that. They read like someone mid-thought, mid-feeling, writing at the speed of emotion rather than the speed of grammar. And that energy transfers into the output. Not literally — AI doesn't "feel" my urgency. But the specificity that comes from writing in that state, the unexpected word choices, the raw phrasing that a careful writer would edit out — all of that feeds into what AI generates. The rough edges in the input produce rough edges in the output. And those rough edges are where the humanity lives.

If I sat down and typed a perfectly formatted, correctly spelled, carefully considered prompt, I'd get a perfectly formatted, carefully considered, utterly forgettable result. The mess is the point. The mess is the signal.

Why This Matters Beyond Music

Everything I've described about songwriting applies to every other creative and professional domain.

The architect who sketches rough ideas and uses AI to render them is working at Level Three. The one who just asks AI to "design a house" is at Level One. The business strategist who dumps twenty years of market intuition into a raw, unpolished brief and lets AI structure it is operating differently from the one who asks for "a strategy for entering the European market."

The tool is the same. The input is everything.

And the input can't be faked. You can't prompt your way to twenty years of experience. You can't type your way to wisdom. You can't generate a life you haven't lived.

That's the paintbrush principle. The brush is available to everyone now. What you paint with it depends entirely on what's inside you. And what's inside you is the one thing in this whole equation that AI cannot produce, replicate, or replace.

So the question isn't "can AI make music?" or "can AI write?" or "can AI design?"

The question is: what have you got that's worth amplifying?

Dave Mattu is a Welsh-British technologist, musician, and reluctant philosopher currently based in Huelva, Spain. He builds AI-powered tools for architects and writes songs about the space between humans and machines under the name Rodington Pear on his label Threshold Artifacts.