Three projects, one question

I'm a tenured teacher, Area 6 Special Function and Google Workspace for Education administrator at a secondary school in Sardinia, Italy. I'm also the author of the book "AI e DSA" (Italian) and founder of Bachata Vibes Music, the platform that generates and publishes bachata, salsa and merengue tracks with AI every day — including the production of Guerda, an artist blending Haitian Kompa with Italian influences.

For years I treated these two things as separate compartments: the classroom by day, the Raspberry Pi by night. Then, while writing chapter 3 of the book — the one that, in the outline, is simply titled "Structure and rhythm: the cognitive pattern shared by music and teaching" — I realized I was answering the same question in both contexts: how do I turn complex content into something that someone with different constraints from mine can actually receive?

What I learned in the classroom

With a student who has learning disabilities, you never start from the raw content. The first step is always the same: break it down. A text becomes an outline, an outline becomes numbered steps, a step becomes one question at a time. The "Student Profile" I talk about in the book is nothing more than this: a set of explicit rules — how much text per line, what kind of language, what rhythm — that turn a generic input into something usable for a specific person.

It's not simplification in the sense of "less content." It's more structure, applied before the content, not after.

What I brought into music

When I built the Bachata Vibes Music pipeline — and later Guerda's — I made exactly the same move, on a different material. A song generated with Suno isn't "press a button and get a track": it's a step-by-step flow, each step with an explicit constraint.

And just like with the Student Profile, the most important constraint isn't technical: it's knowing what shouldn't be automated. Guerda's voice is always a real recording, never generated — it's the non-negotiable of the project, exactly like for a student with learning disabilities the non-negotiable is that the subject matter doesn't change, only how it arrives changes. Automating everything you can doesn't mean automating everything you technically could.

Structure isn't a constraint that limits content. It's the tool that makes it reachable — for a student with dyslexia or for a listener scrolling through a feed in three seconds, the problem is the same: too much information, not enough form.

Why this isn't just a nice analogy

When I automated social publishing with n8n, the first prototype always failed at the same point: I was trying to make the machine do everything at once, without breaking the problem down into nodes that could be verified one by one. I only got it working when I applied the same principle I use every day with an individualized education plan: one step at a time, each step verified before moving to the next, no shortcut that skips a step to "save time."

I'm not saying teaching and building AI pipelines are the same activity. I'm saying the cognitive pattern underneath both — structure before content, verify at every step, automate only where it's safe to do so — is identical, and I refined it in the classroom before ever applying it to code.

The real product

The book, Guerda's tracks and the n8n workflows aren't three outputs from three separate hobbies. They're three applications of the same method to three different materials: a school text, an audio track, a post to publish. What I've built over these years isn't a catalog of projects — it's a repeatable way of breaking a complex problem down into steps that someone (a student, a listener, an automated system) can actually receive.