A founder essay · May 2026

What the Tool Was For

On making music after she died — and why I built HoldingMusic.

[Her name] died in October 2021. This essay is not about that, and also it is not about anything else.

What happened in the months after is the part I can write about. I could not breathe right. I do not mean this as figure of speech. I mean there was a particular shape my chest used to make when air went into it, and that shape was gone. In its place was a smaller, shallower motion that did not refill anything. I noticed it on a Tuesday afternoon in [month, year], standing in the kitchen, holding a glass of water I did not remember pouring. I had been doing it for months and had not seen it because grief does not announce itself in the way the books tell you it will. It moves into your body and rearranges the furniture and you do not notice you are sleeping on the floor until someone asks why your back hurts.

I tried the things people try. I will not enumerate them; you can guess. None of them brought the breath back. Some of them put a thinner layer between me and the days, which I was grateful for, and which was not the same thing.


One afternoon — and I do not remember the date, and I do not trust the people who do — I opened a website called Suno on my phone and typed eight or nine words into a box.

I am not going to romanticize what happened next. The tool produced a song. It was not a great song. It was not a song anyone else needed to hear. But it was a song that came out of a sentence I had been carrying for a year and had not yet found a way to say. The sentence was about [her name], and it was the kind of small private thing that does not survive being said out loud in any room I had access to. I had no piano. I had no guitar. I had no co-writer to sit across from. I had a sentence and a phone and a model that did not ask me what my qualifications were.

The song played. I sat down. The breath came back. Not all of it. Not for long. But it came back the way it used to come back — through the whole chest, not just the top of it — and I sat in the kitchen and listened to it twice more and then I closed the app and went outside and walked for an hour, and when I came home I was a person who could be in a room again.

I do not know what to call what that tool did. I know what it did not do. It did not write the song; the song was the sentence I had been holding, and the sentence was mine. It did not understand grief; nothing understands grief. It did not replace any human being I had lost or any musician I had not become. What it did was meet me where I was, on the floor, with nothing in my hands, and hand me back a thing I had made.


For the next year I made songs for other people.

A friend whose father had died. A neighbour whose son had moved away. A woman I have known since I was nineteen whose marriage ended on the same December morning my wife had been gone two months. I made songs the way other people bake bread — quietly, in the kitchen, for someone specific. I did not post them anywhere. I did not put my name on them. I sent them by message, one person at a time, with a short note saying this is for you, you do not have to like it, you do not have to thank me.

Some of them did not reply. Most of them did. One of them played the song at her father's wake. Another sent it to her mother, who I never met, and her mother sent back a voice note crying, which I did not listen to for three days because I did not yet know how to receive that. I learned. You learn.

I am telling you this because at no point during any of those exchanges did anyone ask me what tools I had used. They asked who the song was about. They asked what I had been thinking. They asked, sometimes, whether I had written the words — which I had, and which is the part that always matters. No one asked which model. No one asked whether it was real music. The question never came up because the question is not a real question. The question is a thing people who have not been on the receiving end of one of those songs ask about other people's songs.


It came up later, of course.

It came up the first time I mentioned to someone in a public setting that I had used a generative model to make music. The face did the thing faces do. The voice went slightly cooler. Oh — but it's AI music. Said like the gift was less. Said like the breath I had found in my own kitchen was a counterfeit breath, retroactively. Said like the wake song, the one that played while a woman whose father had just died stood next to her mother and held her hand, was a different kind of song than it had been thirty seconds before, because of the tool.

I have thought about that face a great deal in the four years since. I have thought about what it would be like to walk up to the woman at the wake and tell her that the song her father got was not real music because of the way it was made. I have thought about what it would be like to walk up to the widow on the voice note and tell her she had cried at a fake song. I have not, in any of these thought experiments, been able to figure out what the person making that face actually believes is at stake.

So I went to look it up.


In 1839 the French Academy of Sciences declared that photography was not art. The argument: a machine had made the image. The image was mechanical, repeatable, soulless. Within a generation, photography produced Eugène Atget, Henri Cartier-Bresson, Dorothea Lange. The Academy was not wrong about the mechanism. They were wrong about what the mechanism meant.

In 1965 Bob Dylan plugged in an electric guitar at the Newport Folk Festival and was booed off the stage by people who had loved his acoustic work the year before. The argument: the instrument was not authentic. The music was no longer real. Within ten years the electric guitar had given us Hendrix, Stevie Ray Vaughan, Prince, the Edge. The folk purists were not wrong that the instrument had changed. They were wrong about what the change meant.

In the early 1980s, drum machines and samplers were dismissed as the end of musicianship, a betrayal of the discipline. The argument: these tools cheat. They did not cheat. They produced hip-hop, which became the dominant musical form of the next forty years, which is now the soundtrack of every wedding and every funeral and every commercial and every revolution. The critics were not wrong about the tools. They were wrong about what the tools were for.

The pattern is unbroken. It is two centuries long now. Every tool that has entered music since the player piano has been met by the same face, in roughly the same words, with roughly the same certainty. The face has been wrong every single time. Not most of the time. Every single time. The historical record on this is not contested.

The face is wrong about AI music for the same reason it was wrong about the electric guitar. It is confusing the tool with the music. It is confusing the instrument with the artist. It is forgetting that the woman at the wake did not ask which tools, and the widow on the voice note did not ask which tools, and the person sitting in a kitchen who could not breathe did not ask which tools — because the tool is never the thing. The tool was never the thing. The tool is what lets the thing through.

A guitar is a tool. A pen is a tool. A camera is a tool. A drum machine is a tool. A model is a tool. The tool produces possibility. The artist produces taste. The artist chooses which of the thousand things the tool offered is the one that means something — the one with the right pacing, the right weight, the right hook, the right truth. Selection is composition. The job has not changed. The job was always selection. The job was always taste. The job was always knowing when to stop. Hendrix did not invent the guitar. He chose, out of every possible note that guitar could make, the ones that he made. That is what an artist is. That is what an artist has always been.


There are, by the most recent estimate I have seen, fifty thousand AI tracks uploaded to Deezer every day. Most of them are nothing. A small fraction of them are something. A smaller fraction of them are the song that lets a person breathe again in a kitchen on a Tuesday afternoon. None of them have anywhere to go. There is no venue for them. There is no leaderboard that means anything. There is no producer scouting them. There is no fund that pays the artists who made them anything when they are heard. The music exists, and the people who made it exist, and the listeners who would care about it exist, and there is no infrastructure between any of these three groups.

That is what I am building. I am building it because it should exist. I am building it because the woman at the wake should not have to wonder, four years from now, whether the song her father got was real music. I am building it because the next person sitting in a kitchen unable to breathe will need the tool that I needed, and that person will deserve, on the other side of having made something, a place to put it where it might be heard by someone who will be glad it exists.

I am building it because the tool was never the music. The tool is what let me back into the room.


HoldingMusic is the discovery and attribution layer for AI music in real-world venues. Real cafés and hotels and yoga studios and gyms across thirty countries will stream tracks made by artists like me and like the thousand artists who will be the founding cohort. City, country, and global leaderboards will rank what is rising. Producers will scout the Rising Star list. Fifteen percent of platform profit will be paid out, every quarter, to the artists who built the platform — leader rewards, rising star grants, new voices lottery. We are a discovery engine, not a label; we own no catalogs and we take no rights. If an artist is signed because of us, we take a small finder's fee. If the artist removes the track later, we keep a small ongoing share. The numbers are low because the numbers should be low. The arithmetic should not be the obstacle.

I think this is what is going to happen, and I think it is going to happen whether I build it or someone else does. The volume of AI music being made is too large and too good and too consequential for there to not be infrastructure underneath it within twenty-four months. The only question is who is going to do it and what their reasons are going to be.

My reasons are above. The breath came back when I made the song. The breath came back for other people when I sent them a song. The face that said but it's AI music did not change that. The face has been wrong for two hundred years. I am not going to argue with it any longer. I am just going to build the room where the music plays, and let the world listen, and let the listeners decide what is real, the way the listeners have always decided.

[Her name], this is for you. October is coming. I am writing this in the kitchen.

Let the world listen.

— [Your name], Palawan, May 2026

If you make AI music, this is your platform.

Open now to the first 1,000 founding artists from the Suno, Udio, ElevenLabs Music and wider AI music communities. The cap is real. Permanent founding artist status, priority in the Artist Fund, named in the launch press release.

Claim my founding spot →