After I retired, I decided to try something I had wanted to do for a long time: make music.
I began with early AI-generated loops and used Logic Pro to put the pieces together. I also made a few videos using my own voice. They came out horribly.
That taught me something useful. I was probably not meant to be a performer. But it also taught me something more important: I loved the process.
For more than thirty years, I had worked as a software developer. Much of that work involved creating applications and processes for other people. Songwriting felt strangely familiar. A song had structure. It had requirements. Every part affected the parts around it. The difference was that, for once, I was building something for myself.
Music Was Already in the Family
My cousin, uncle, and brothers are talented musicians in their own right. Some have sung professionally, while one of my brothers is a self-taught guitarist. I did not have their experience, but retirement gave me the opportunity to begin a musical quest of my own.
I learned enough piano to understand what I was hearing and doing. Then I began learning guitar. I studied music theory more deeply through books and YouTube. I was not trying to become a virtuoso. I wanted to understand how songs worked and become more capable of expressing my own ideas.
I continue learning to this day.
The Song That Changed Everything
Eventually, my siblings and I came together to write a song for our mother on her birthday. The song was called Si no Fueras Tú.
I used Suno to help turn it into a recording, and the result made her very happy. At the time, that was enough. We had made something for her, and she heard the love behind it.
She died exactly one year and one month later.
On her deathbed, we sang the song to her.
That experience permanently changed what songwriting meant to me. The value of the song was not determined by the tools used to produce it. It came from my brothers, my mother, our history, and the reason we wrote it. Technology helped us give the idea a form she could hear, but the song belonged to a real moment in our lives.
You can listen to Si no Fueras Tú on YouTube.
Why I Always Begin With the Words
I never asked AI to write an entire song for me. That did not make sense. If the system wrote everything, I would be absent from my own song.
I made the decision early that I had to write the words myself. Part of that was about ownership, but mostly I simply loved writing. Lyrics gave me a place to say something personal before any instrument played a note. Heck, I had a lot to say—and I still do!
Over time, I began hearing more music inside the words. First it might be a rhythm. Later it might become a small riff or melodic motif. I slowly learned how stress, phrasing, syllables, and emotion can influence the direction of a melody.
This lyrics-first approach became the foundation of everything I built afterward.
From Songwriting to Tool Building
Working with Suno was useful, but I began noticing a problem. A generated song could sound impressive at first. Then I would listen to more songs and hear familiar patterns. Different tracks could begin to feel too similar.
I did not want to depend on repeated generation until something happened to sound right. I wanted to bring more of myself into the process.
I started supplying Suno with motifs instead of asking it to make every musical decision. I used digital audio workstations to edit and clean up the results. Later, because I am a developer, I began building software that could generate a deterministic melody—one guided by repeatable rules and the natural prosody of my lyrics. I added ways to customize the result rather than accepting a musical black box.
The objective was not to remove creativity or surprise. It was to make sure that I remained present. In MelodyCurve, users can shape the melody with a draggable curve and other fine-tuning controls, giving them direct control over the result.
The Work People Do Not See
I eventually became an advocate for thoughtful AI-assisted music. But as low-effort generated music became more common, people became skeptical. Some dismissed the entire field without asking how a particular song had been made.
That could be discouraging. I sometimes worked for weeks on a song—writing, revising, developing a motif, testing arrangements, editing audio, and trying to make the result genuine. From the outside, someone could still assume that I had pressed a button and accepted whatever came out. From the finished audio alone, there was no easy way to tell how much work had gone into it.
I understand the concern. There is a great deal of disposable AI content. But not every AI-assisted creative process is the same.
For me, the important question has never been whether a tool was involved. Every recording uses tools. The important question is whether the human being remained responsible for the ideas, choices, meaning, and final direction of the work.
Still Learning, Still Writing
I do not present myself as a professional performer. I am a retired developer who began learning music, discovered that I loved writing songs, and then applied a lifetime of process design to a new kind of problem.
In software, I built for others. In songs, I write for myself.
That does not mean the songs are meant only for me. I hope other people hear something honest in them. But they begin with my experiences, my questions, and the people I love. That is why I keep writing.
In the next article, I will explain why I never let AI write my songs, what I learned about musical sameness, and why I began developing my own tools for signature melodies and human-directed authorship.
Related: Rethinking AI Music