Six spare screws: what my first mistakes with AI taught me about writing songs

Madrid, 25 October 2026.

By Mike Sain't Louis

When people hear that I have released 273 singles in roughly two years, some of them picture a smooth ride. They imagine I turned up, pressed a button and songs started pouring out like sausages from a machine. It makes me laugh, because my real beginnings with these tools looked much more like someone trying to assemble Swedish flat-pack furniture without the instructions and ending up with six screws left over.

I am an indie alternative artist based in Madrid. I call myself "the first AI singer-songwriter", and I want to be precise about what that means, because getting it wrong was one of my mistakes. I write every lyric. The music is produced with Suno, under my direction. I decide what stays and what goes, and I direct the music videos. I do not perform my songs: the vocals are generated. The words, the ideas and the artistic direction are mine.

Here are some of the stumbles that got me here. You learn more from what went wrong than from the heroic version.

I thought the machine could read my mind

My first mistake was one of attitude. I assumed that if I handed the tool a good lyric, it would give me back the song I had in my head.

It does not work like that. The machine does not know that a line is a joke. It does not know that a particular word needs to land with a raised eyebrow. It reads the lyric and does something with it, sometimes something lovely, but without understanding what I meant. At first I would decide a song was "wrong", when what was wrong was the way I had asked for it.

It took me a while to accept that the tool needs direction. Knowing what you want is not enough; you have to learn to say it in a way that is useful. That is a craft in itself, and at the beginning I simply did not have it.

I fell for the first version

The second stumble was sillier. I would keep the first thing I liked.

A version would excite me and that was that. No comparing, no letting it rest. Later, listening back to some of those decisions, I realised I had been carried away by novelty. What thrilled me was that the song existed, not that it was good. Those are two different feelings, and early on they are very easy to confuse.

These days I make my first burst of enthusiasm wait a little before I agree with it.

I started writing for the machine

This was the worst one, and it took me longest to see.

For a while I began writing lyrics around what the tool did best. Simpler lines, more predictable rhymes, structures I knew would work. The songs came out cleaner, and blander. I was bending my writing to suit the machine instead of making the machine bend to me.

I noticed it one day rereading old lyrics. Something did not sound like me. The twist was missing, the line that runs too long because it has a story to tell, the strange title. If I call a song "Mitad Venezuela, mitad Pakistán" ("Half Venezuela, half Pakistan"), it is because I want a title nobody sees coming, not because it is convenient to produce.

Since then I have had one rule: the lyric comes first, exactly as I want it, and everything else gets worked out afterwards. If the tool struggles with a line, I fix it by directing better, not by cutting what I wanted to say. I do not always manage it. But I no longer write to please a piece of software.

I confused busy with better

A quieter trap. When a tool lets you make a lot, it is easy to believe you are progressing simply because you are making a lot. I fell for it. For a time I measured my days in finished songs, the way some people count steps on their phone.

Quantity has its value. Releasing a great deal has taught me more than waiting for the perfect album would have. But there were moments when I realised I was producing on autopilot: competent songs I did not much care about. I learnt to ask whether anyone needed this, starting with me.

The warning sign was very simple, and I ignored it for too long. When I finished a song and had no urge to show it to anyone, something was off. With the ones that matter, the opposite happens: I can barely wait for them to come out. That impatience is a better thermometer than any song counter.

I could not explain it

One stumble had nothing to do with music. It was how I talked about it.

At first, when someone asked how I made my songs, I got tangled up. I said things that were half-finished, or too technical, or too vague. Misunderstandings followed. Some people assumed the vocals were mine; others assumed the machine did everything and I just pressed the button. Neither was true. It took me a long time to find a plain way of saying it, which is the paragraph you read near the top of this column.

Why I am telling you this

There is a lot of noise in Britain about what AI will do to music. I have no grand theory. I have two years, a catalogue called PLANETA GAYFOLK of 273 singles and more than 280 music videos, and a list of mistakes. If there is a lesson, it is that the tool is only as honest as the person directing it.

You can hear the results, and judge for yourself, at https://mikesaintlouis.com, where every song can be streamed and downloaded for free, each with its full lyrics. My music is also on Spotify, Deezer and the main platforms. The original piece is on my blog: https://mikesaintlouis.com/blog/mis-primeros-tropiezos-con-la-ia

About the author

Mike Sain't Louis is a Madrid-based lyricist, composer, producer and artistic director, and the founder of his own independent label, MIKE SAIN'T LOUIS, with no record company behind him. His queer alternative pop, cheeky, camp and uncensored, ranges from electropop to Spanish copla and ballads, in Spanish as well as English, French and Portuguese. His GAYFOLK trilogy imagines a journey to a planet where everyone can live as they are. Many listeners found him in 2024, when "Tan Cateto" went viral on TikTok.

Press contact

Email: info@mikesaintlouis.com Website: https://mikesaintlouis.com Instagram: @planetagayfolk