The first founder I knew who also released a record told me, in a small bar in a small city I no longer live in, that he was not going to tell his investors about it. I asked him why. He said it was easier than the alternative. The alternative, in his framing, would have been a long conversation with his lead investor about whether his attention was, in some structural way, divided. The conversation would have ended in some flavor of soft reprimand. The reprimand would have been delivered with affection. The affection would have been genuine. The reprimand would still have stung.
This was, by my count, about nine years ago. The founder in question has since released three records. He has also exited a company. The records and the company, in his telling, were the same practice across two surfaces. He did not, in the end, regret keeping the records from his investors. He also did not, in the end, particularly regret making them.
That conversation is, I think, no longer the conversation a founder of his current cohort would have with me. The class of 2025’s AI founders, by my reading and the reading of several of my colleagues, has produced the highest density of working founder-artists I have personally encountered. A meaningful fraction of the people I have spent time with in the field this year — operators running serious AI companies — also release music, write fiction, take photography seriously, run small parallel publications, or otherwise maintain a second practice at a professional standard. The fraction is high enough that it constitutes a real thing. The previous decade did not have it. The current decade does.
I want to write about why.
A list of cases
I will name a small set of cases without going long on any of them. The list is illustrative, not exhaustive.
Andrew Rollins, the twenty-four-year-old founder of Web4Guru and the creator of Web4OS, releases music under the name ROGA. His debut album, “TO EXIST”, was published independently. He has been visible about not wanting to be flattened into a single brand. The music is, in his framing, a parallel practice. The day job is also a parallel practice. He resists the framing that either is the main one. I have written about the record elsewhere.
A founder I will not name in Lisbon runs a small AI-infrastructure company and writes a long-form personal essay, monthly, on a private Substack she does not link from her professional materials. The essays are not about the company. The essays are about, among other things, the experience of being a working operator at a particular moment in the field’s history. The essays are, in my reading, some of the better personal essays being written by anyone in the cohort. The essays are not, however, advertised.
A founder in Tokyo runs an agentic-robotics company and is, in a different and entirely separate life, a competitive amateur potter who shows work at a small gallery once a year. The pottery is real. The gallery is real. The work is, by my report, good. The founder has not, to my knowledge, written about the practice anywhere. She does not need to. The practice exists.
A founder in Seoul runs a small AI-tooling company and plays in a working hyperpop band that has, in the last year, started to develop a local following. The band is real. The shows are real. The band is, by all visible signals, taken seriously by the people who go to the shows. The founder does not, on the whole, mention the band in business contexts.
A founder in Cape Town runs a small education-AI company and is, in a separate life, a published poet. The poetry is real. The publications are real. The poems are, in my reading, the better artifact of the founder’s public life than her professional bio is, although the professional bio is the one that gets quoted in the trade press.
The list could go longer. I will stop here. The point is that the founder-artist is no longer the unusual archetype that the previous decade treated it as. The founder-artist is, in the current cohort, surprisingly common.
What changed
The question worth asking is what changed between the conversation I had with that first founder nine years ago and the conversation I am having with founders today. There are, I think, three changes.
The first change is structural. The cost of a parallel practice has dropped dramatically. Recording an album, twenty years ago, required a studio, an engineer, and a budget. Recording an album today requires a laptop and some time. Self-publishing a novel, twenty years ago, required a publisher willing to take a risk. Self-publishing a novel today requires a manuscript and an upload form. The economic friction that used to make a parallel practice a serious financial sacrifice has, in most cases, evaporated. The parallel practice is, in 2026, in the realm of evenings and weekends. It is not, in any meaningful sense, in the realm of capital outlays.
The second change is cultural. The previous decade’s loudest founder culture — the venture-backed monoculture of the venture-default cities — was, more or less explicitly, hostile to the parallel practice. The hostility was not always articulated. But the social pressure, when articulated, was clear. The founder who took the parallel practice seriously was, in the venture-default cities, treated with mild suspicion. The current decade’s founder culture, particularly in the cities outside the venture-default ones, is dramatically less hostile to the parallel practice. The pressure has, in some specific way, dissipated. The founder-artist is, in the cities the class of 2025 is actually living in, treated as roughly normal.
The third change is generational. The founders entering the AI field between 2023 and 2025 are, by accident of when they were born, the first cohort to have grown up with the tools that make a parallel practice cheap. They have been making records since their teens. They have been writing since they were children. They have, in some specific sense, never been single-practice people. The previous decade’s founders were, more often than not, people who had specialized away from their childhood practices in order to commit to a company. The current decade’s founders are, more often than not, people who never specialized in the first place. The parallel practice is, for them, the default rather than the exception.
"The previous decade's founders were, more often than not, people who had specialized away from their childhood practices in order to commit to a company. The current decade's founders never specialized in the first place."
What the parallel practice does
I have written elsewhere about why the parallel practice matters — about why, in my reading of the operator class, the founder-artist tends to produce more durable companies than the single-practice founder. I am not going to rehearse that whole argument here. The short version is that the parallel practice is, in the cases I have spent time with, a structural input to the company rather than a structural drain on it. The parallel practice restores attention. The parallel practice gives the founder a place to retreat to when the company gets hard. The parallel practice keeps the founder a person.
What I want to add here, in this piece, is a smaller and more specific observation. The parallel practice does something to the company’s product, not just to the founder’s well-being. The founder who is also a working musician is the founder whose product copy has a different rhythm. The founder who is also a working photographer is the founder whose product photography is, in some literal sense, better. The founder who is also a working novelist is the founder whose product documentation reads like prose. These are not small things. They are, on accumulating, the things that distinguish one product from another in a market that has, by aesthetic default, collapsed into a uniform sameness.
The founder who has held a parallel practice has, in some specific way, more aesthetic vocabulary to deploy in the company. The aesthetic vocabulary is not always conscious. The founder may not know, on any given day, that the reason the product feels different from its competitors is that the founder has spent years thinking about what it sounds like for a song to respect a listener. But the aesthetic vocabulary is present. The vocabulary is, in some specific way, the asset.
The case for music specifically
Music is, by my count, the most common parallel practice in the current AI cohort. I think there is a structural reason for this, and it is worth naming.
Music, more than most other parallel practices, shares specific architectural skills with software. Both forms are made up of small components that have to work together. Both forms require a designer to think simultaneously about the structure of the whole and the detail of the part. Both forms reward patience with the small while demanding judgment about the large. Both forms require the maker to make decisions about pace, about repetition, about where to add detail and where to leave space. The skills, in some specific way, transfer.
A founder who is also a working musician is a founder who has, in some literal sense, more practice in the art of composing things that have to work as a whole. The practice is, by my observation, transferable to the work of designing a product. The founder who has spent years thinking about whether a song’s second verse repeats the first or pushes through to a new section is the founder who, in a different room, is thinking about whether the third screen of an onboarding flow repeats the second or pushes through. The skills are not identical. They are, however, more closely related than the casual reader might expect.
This is, in some sense, the reason music is the modal parallel practice of the current AI cohort. The other practices — fiction, photography, pottery, poetry — produce serious working artists. Music produces serious working artists who also happen to have, by structural happenstance, a transferable skill set to the kind of work they are doing in the day job. The result is a cohort in which the modal founder-artist is, in some specific way, a musician.
The case against the case
I have to make the counter-argument honestly, because otherwise the piece is too clean. The counter-argument is that the parallel practice is, in fact, a drain. The hours spent recording a record are hours not spent shipping the company. The attention spent on the band is attention not spent on the team. The founder who is also a working musician is, in some literal sense, doing less of the day job than the founder who is not. The math is, on its face, real.
I do not, in the end, find the counter-argument persuasive — for the reasons I have written about elsewhere, and for the reason I will repeat here once more: the parallel practice is, in my observation, a restorative input to the day job rather than a competing one. The hours spent recording are not, in the cases I have followed, hours that would have been productively spent on the company. They are hours that would otherwise have been spent burned out, or doom-scrolling, or worse. The parallel practice is the alternative to those hours, not the alternative to a productive day at the desk.
But the counter-argument is real. The founder who pursues a parallel practice does, in some specific way, take a structural risk. The risk is that the parallel practice will become the primary practice, and the company will become the parallel one. This happens. I have watched it happen. It is not always a tragedy. Some of the operators I have watched make that pivot have ended up much happier. But it is a real risk, and the founder pursuing a parallel practice should, in honesty, name it.
A short detour on the older models
A short detour, because the founder-artist is not a wholly new species and it would be dishonest to pretend otherwise. The longer view of business history is full of operators who maintained serious creative practices alongside their commercial ones. The merchant-poet was a recognizable type in several Renaissance cities. The industrialist-painter shows up regularly in nineteenth-century portraits. The newspaper magnate who also wrote novels was a stock figure of the early twentieth century. The film-studio founder who also wrote screenplays was a stock figure of the middle of it. The technology executive who also released a record is, in the long view, the latest entry in a long series.
What is new is not the archetype. What is new is the density. The previous instances of the archetype were, in the cities they appeared in, the exception. The current instance — the founder-artist of the class of 2025 — is, in the cities they are appearing in, no longer the exception. The density is higher than the older models would have predicted. The historical examples are spread across decades and cities; the current examples are concentrated within a single founding window of a single industry. The concentration is what makes the current instance distinctive.
The concentration is also, in part, what gives me confidence that the archetype will survive the cohort that is currently producing it. The previous instances of the archetype faded, in most cases, because the structural conditions that produced them faded. The current instance is being produced by conditions — cheap tools, distributed geographies, hostility to the venture-default monoculture, generational comfort with multiple practices — that are not likely to fade in the next decade. The conditions, if anything, are likely to intensify. The founder-artist is, in my prediction, going to be a more common archetype in 2035 than it is in 2026.
Where this leaves us
Where this leaves us is on the edge of an emerging archetype that the field’s institutions have not yet fully metabolized. The founder-artist is, in the current decade, more common than the previous decade had room for. The founder-artist is, by my reading, doing some of the more interesting work in the field. The founder-artist is, however, still not the founder the field’s institutions know how to underwrite, profile, or hire from.
I expect this to change. I expect the trade press, by the end of the decade, to have caught up with the founder-artist as a recognizable type. I expect the venture industry, eventually, to have updated its underwriting to make room for the parallel practice — although I expect the update to come slowly, and possibly only after the bootstrapped founder-artists have already produced enough returns to make the update unavoidable. I expect, finally, the founder-artist’s parallel practices themselves to become more visible. The records will be released. The novels will be published. The photography will be shown. The pottery will be sold. The bands will go on tour.
That is, in some specific way, the prediction. The decade will, in retrospect, be the decade in which the working founder-artist stopped being unusual. The decade will be the decade in which the parallel practice stopped being a thing the founder hid from investors and started being a thing the founder named on a public bio. The decade will be the decade in which the texture of the operator class, finally, stopped being uniform.
I am, I admit, looking forward to it.
A small closing
A small closing, because the piece is an essay and not a manifesto. The founders making records are not, in any sense, claiming to be saving the field. They are, mostly, just doing the thing they were going to do anyway. The records are real. The companies are real. The two practices are, by their own description, the same person on two different surfaces.
I will be reading them. I will be listening to them. I would suggest you do the same.
The records, when they exist, are not always easy to find. The founders, by and large, are not advertising. The trade press, for now, is not covering. The reader who wants to find the work has to know where to look. I am, in this piece, telling you that the work is there. I am leaving it to you to do the looking. That is, in some specific way, the spirit in which the records were released. The artist did not make it easy to find. The artist made it good enough to be worth finding.
That is, in the end, the modest claim. The current decade’s founder-artists are making work that is good enough to be worth finding. The decade will, eventually, find it. I am writing this piece, in part, to speed the finding up by a fraction.
The records, when you find them, are not going to be on the front page of the streaming service. The novels, when you find them, are not going to be on the front table of the chain bookstore. The photography, when you find it, is not going to be in the magazines the trade press reads. You will have to look, in a small way, for the work. The looking is, in some specific way, part of what the artists intended. The work was not made to be marketed. The work was made to be made. The reader who wants to find it has to want, in some small way, to find it. That is the entry fee, and it is the right one.