Prompts Are the Scales

As a teenager I produced hip hop beats. And I was always hunting for scales.

I was convinced that my beats would sound great the moment I found the right one. The secret ingredient. I spent months living inside the Lydian scale. At some point I asked a friend, a real musician, which scales the professionals use. I was almost obsessed with the question, because I was sure this was the missing piece: they knew something about scales that I did not, and that was why their tracks knocked and mine did not.

He answered with contempt. You cannot read groove from recipes.

Today I watch Instagram videos of people using Claude, trading prompts with each other the way I traded scales. The perfect prompt for a landing page. The secret system prompt that makes the code clean. Copy it, and your software will knock.

I recognize the pattern immediately, because I was sixteen once.

Here is what actually happens. A new tool hands you 80 percent of a craft for free. No experience required. You type, something real comes out, and it runs. This is intoxicating, and it is genuinely new: two years ago, those 80 percent took years to reach.

But the last 20 percent still cost 80 percent of the experience. Programmers have known a version of this for fifty years as the ninety-ninety rule: the first 90 percent of the code takes 90 percent of the time, and the remaining 10 percent takes the other 90. The AI version is harsher, because it does not measure time. It measures experience. The tool compresses the first 80 percent into an afternoon and leaves the last 20 exactly where they have always been: behind years of reps.

The recipe hunt is the tell. The person collecting prompts believes the gap between them and the result is information. A secret. A scale. But the gap is not information. It is groove: the thousand small judgments you only earn by shipping a thousand bad beats. When something breaks at the 80 percent line, the prompt collector stands in front of it the way I stood in front of my Lydian loops, wondering why the magic ingredient did not work.

I am not laughing at these people. I was these people. The scale hunt is not stupidity. It is a phase, the first honest phase of wanting to be good at something before you know what being good consists of.

And here is the part my younger self deserves to hear, and the part that should worry my current self.

The kids on Fruity Loops won.

FL Studio was the toy. Real producers had studios, consoles, training. The industry laughed at bedroom kids with cracked copies and stock plugins. Then those kids became the biggest producers in the world. Trap runs on FL Studio to this day; Metro Boomin never switched. And the detail almost everyone misses: the limitations of the toy became the sound of the genre. The industry did not absorb the kids. The industry ended up sounding like their software.

So no, I do not think there is a case for gatekeeping. The gate never guarded quality. It guarded the door.

Our skills as software developers are still valuable. The last 20 percent are real, and depth still compounds there. But the skills are no longer the entry ticket. There will be people with almost no experience who get further than the music student. Some because they carry a different kind of experience, a domain, a taste, an audience. Some because of timing and luck.

That is not the meritocracy we like to tell ourselves about. But the meritocracy was always a story told by the people who won under the old rules. The degree called itself merit because the degree holders wrote the definition.

Claude is the Fruity Loops of programming.

In ten years, software will sound like the kids.