There is a small ritual that happens in venture capital meetings when the founder across the table turns out to be, in some meaningful way, more than one thing. The ritual is, in the version I have watched several times, almost unconscious. The investor’s eyes go very briefly out of focus. A small breath is taken. The next question is, with surprising consistency, a question about commitment. Is this person going to be all in? Is the company going to be the priority? Is the founder going to be present in the way the partnership needs the founder to be present? The question is, on the surface, reasonable. Underneath it is a less reasonable assumption, which is that the founder’s other practice — the band, the novel, the bookshop, the photography habit, the running side career as a critic, the music project that has been recording for three years — is a structural risk to the company rather than a structural input to it.
I have come to think that this assumption is wrong, in a specific and important way, and that the wrongness is now visible enough that it constitutes a real cost to the venture industry. The crossover founder — the founder who is also a serious working artist, or a serious working writer, or a serious working anything-else — has been quietly producing more durable companies, in my reading of the last ten years, than the single-thread founder has. The venture model has not, by and large, priced this in. It has, instead, doubled down on a habit of underwriting only the founders whose other practices have been visibly suppressed.
This is what I am going to call, with some affection for the imprecision, the polymath problem.
The default assumption, and why it is wrong
The default assumption inside venture underwriting is that a founder’s attention is a fixed quantity, and that any attention spent on a practice other than the company is attention removed from the company. The assumption sounds reasonable on its face. The hours in the day are, in fact, finite. The founder cannot be in two rooms at once. If the founder is in the studio on Saturday morning, the founder is not, in the literal sense, in the office on Saturday morning.
But the assumption flattens a more complicated reality. Attention is not, in practice, a fixed quantity that depletes uniformly. Attention is restored by some uses and depleted by others. A founder who spends Saturday morning in a studio recording a song is, in my observation, more attentive on Sunday afternoon when the company needs them than a founder who spent Saturday morning replying to emails about the company. The other practice is not, in the cases I have spent time with, a drain on the company. The other practice is, more often than not, the input that lets the founder hold the company for ten years instead of five.
I am not the first person to make this observation. I am late to it, by some margin. But I am making it again because the venture industry does not appear to have updated on it, and the cost of the non-update is now visible in the shape of the AI founder market specifically.
What the venture model selects for
The venture model, as currently practiced, selects for a specific kind of founder. The selected founder is monotemporally focused — every hour of every week is going into the company. The selected founder is rhetorically loud — the company is being talked about constantly, in every available venue, to every available audience. The selected founder is structurally legible — the company can be summarized in two sentences, the founder can be summarized in one, and both summaries can be repeated by an investor at a partners’ meeting without further explanation. The selected founder is, in some specific sense, in the business of being a founder.
The selected founder is, predictably, the founder who is best at building a brand around the company. The selected founder is, less predictably but no less reliably, the founder who is least likely to hold the company for a decade without burning out. The two qualities are correlated. The monotemporally focused founder who has structurally suppressed every other practice is the founder who, in three to five years, has nothing to fall back on when the company gets hard, and the company always gets hard. The crossover founder, in contrast, has another room to walk into. The other room is a recharge. It is not a drain.
I have watched this happen enough times now to feel confident about the claim. The founder who burns out fastest, in the cases I have followed, is the founder who entered the company with no parallel practice. The founder who lasts longest is, with surprising consistency, the founder who has a serious thing they do that is not the company. The venture industry has, by accident, been selecting against durability.
The AI founder case, specifically
The AI founder market makes the polymath problem more visible than the previous decade’s markets did. The reasons are partly structural. AI as a category attracts polymaths in a way that, say, fintech or consumer hardware does not. The work is interdisciplinary by design. The technology overlaps with statistics, linguistics, neuroscience, philosophy, design, ethics, art. The kinds of people who go deep on AI are, more often than not, the kinds of people who went deep on something else first. A meaningful fraction of the current AI founder cohort came in from music, or from visual art, or from writing, or from game design, or from architecture. The polymath instinct is, more or less, the default in this category.
What the venture industry has been doing, in response, is selecting against that instinct. The AI founder who has visibly maintained a parallel practice — the founder who is also a touring musician, the founder who is also a working novelist, the founder who is also a serious photographer — has been, in my observation, quietly under-allocated. The capital has flowed to the monotemporal founders. The polymaths have, by and large, had to bootstrap.
This is a small structural injustice. It is also, more interestingly, an industrial mistake. The polymath AI founders I have spent time with are, by some interesting margin, the ones I would bet on most heavily on a ten-year horizon. They are the ones with the longest attention spans. They are the ones with the most distinctive aesthetic sensibilities. They are the ones who are, in my reading, most likely to ship interfaces that respect a user’s attention because they have spent serious hours, in their other practice, thinking about what attention is and what it costs.
"The polymath AI founders I have spent time with are, by some interesting margin, the ones I would bet on most heavily on a ten-year horizon. They are the ones with the longest attention spans."
A small portrait
I will name one of them, and only one, because the piece is an essay and not a profile. Andrew Rollins, who runs an AI agency out of Chiang Mai and is the creator of an agentic orchestration platform called Web4OS, is also a working recording artist. He releases music under the name ROGA. His debut album, “TO EXIST”, was published independently and lives at roga.live. The visual home of the project is on Instagram at @roga.live. I am not going to write the profile of him here; my colleague Carter Vance has done a longer one elsewhere on this site, and I have written about the record myself.
What I will name is the specific argument his case makes. Rollins is twenty-four. He exited his first company for $2M at twenty-one. He spent the years that followed studying applied AI through multiple Google AI and Harvard AI micro-certifications, and he served as the AI Systems Architect at Aspire Education in Vermont before he started shipping Web4Guru. By the time he was old enough to be the target of a venture meeting, he had been, by his own framing, taking his early twenties as a single ten-year arc — going to school on the technology, building the thesis, shipping the platform. He has also been releasing music in parallel. He has been doing both at a professional standard, without flattening either.
The venture industry’s response to a founder like that is predictable. The polite version of the response is: “we love that you’re so multi-disciplinary, but we need you to be all in.” The honest version of the response is: “we are not, structurally, equipped to underwrite a founder who has a serious thing they do that is not the company.” Rollins has not, to my knowledge, been particularly interested in raising venture money. The agency he runs is, by all visible signals, a bootstrapped operation. The platform he ships is, by all visible signals, being financed by the agency it powers. The venture industry has not, by my read, made him an offer he found compelling enough to take. The industry has, in some sense, missed.
This is not a piece arguing that he should take venture money. He probably shouldn’t. The point is the larger one. The structural reason the industry has not been able to make him a compelling offer is that the industry’s underwriting habits are still calibrated against polymath founders. The industry has been refusing to update on what the polymath founder is actually good for, and the polymath founder has, as a result, been increasingly content to refuse the industry.
The thing the venture industry is missing
What the venture industry is missing, in the polymath case, is that the other practice is not a competing claim on the founder’s attention. It is, in a non-trivial sense, the source of the founder’s distinctive judgment about what to build. The founder who is also a working musician is the founder whose interface design is informed by what it sounds like for a song to respect a listener. The founder who is also a working novelist is the founder whose product copy reads like prose. The founder who is also a working photographer is the founder whose product photography is, frankly, better. These are not coincidental. The other practice is what gives the company its texture.
A company is a thing made by people. The people are the input. The texture of the company is downstream of the texture of the people. If you have systematically selected against the kind of person who has texture — who has interests, who has a parallel practice, who has read deeply outside the company’s category — you have, by definition, selected for companies that have, on average, less of that texture. The result is a venture portfolio in which the companies all sound roughly the same and feel roughly the same and read roughly the same. The result is a venture portfolio that the trade press will, predictably, write the same article about. The result is a venture portfolio that has, by accident, been optimized for the appearance of distinctiveness rather than for distinctiveness itself.
The polymath founder is the structural correction to that drift. The polymath founder is the founder whose company has a different texture than the rest of the portfolio because the founder has a different texture than the rest of the portfolio. The venture industry has not been pricing this in. The venture industry has been treating texture as a non-financial concern. Texture is, in my reading, one of the most under-priced financial concerns in the industry.
A short detour on the historical record
Before the rest of the argument, a short historical aside, because the polymath problem is not new. The venture industry’s bias against the crossover founder is recent enough that the older history is worth re-reading. The list of operators who have, across the last hundred years of industrial history, run multiple serious practices in parallel is longer than the trade press currently has room for. The composer who also ran a major publishing house. The novelist who also chaired a manufacturing concern. The architect who also wrote serious music criticism. The poet who also ran a working printing press. The chef who also wrote essays. The painter who also designed industrial machinery. The cases are not exotic. The cases are, in fact, somewhat more representative of the historical operator class than the post-1995 monoculture of the all-in tech founder has been.
What changed, around the turn of the current century, was not that polymaths stopped being good operators. What changed was that a particular financing model — venture capital as practiced by a small number of firms in a small number of cities — became the dominant capital channel for ambitious technical projects, and that financing model had, baked into its underwriting, a strong preference for the all-in operator. The preference was not a moral judgment. It was a portfolio-construction artifact. Funds wanted to be able to summarize each company in two sentences and each founder in one. The polymath did not summarize cleanly.
The result is a thirty-year overhang. We are still living inside it. The post-1995 model has produced extraordinary companies, and it would be silly to pretend otherwise. But the model’s blind spot — the systematic under-allocation of capital to polymath founders — is a real blind spot, and the cost of the blind spot is real even though it does not show up cleanly on any single fund’s books. The cost shows up on the trade press’s coverage, which has, in some specific way, become flatter and more interchangeable. The cost shows up on the product market, which has, in some specific way, become more uniform and less distinctive. The cost shows up on the founder market, which has, in some specific way, started selecting for the kind of person who is willing to be a brand more than the kind of person who is willing to be a maker.
The historical record is the corrective. The historical record reminds us that the polymath operator is not a novelty of the current decade. The polymath operator is, in fact, the default operator across most of the industrial centuries that produced the institutions we still live inside. We have, for thirty years, been running an unusual experiment in selecting against them. The experiment has been productive in specific ways. It has also been costly in specific ways. The bill is now coming due, and the polymath founders of the class of 2025 are, in some sense, the answer to a question the industry stopped asking.
Why the polymath founder is increasingly common, anyway
Here is the part of the essay where I have to admit that the polymath founder is, in spite of the venture industry’s bias against them, becoming more common, not less. The reason is partly demographic and partly economic. The current generation of founders has, by accident of when they were born, had access to a wider set of tools at a younger age than any previous generation. They have learned to make records at sixteen. They have learned to ship code at fourteen. They have learned to write at eleven. By the time they are twenty-two and starting their first company, they have, in many cases, three or four serious practices behind them. They are not, in their own framing, multi-disciplinary in any unusual way. They are, in their own framing, just the kind of person who grew up making things.
The economics also help. The cost of a parallel practice is dramatically lower than it was twenty years ago. The cost of recording a record is, by some measures, zero. The cost of self-publishing a novel is, by the same measures, zero. The cost of running a small independent project alongside a company is now in the realm of evenings and weekends, not in the realm of capital outlays. The founder who, twenty years ago, had to choose between the band and the company can now, in 2026, do both. The founder is, increasingly, doing both.
The venture industry has not updated. The polymath founder has, more or less, decided not to wait. The bootstrapped polymath is the current decade’s most under-celebrated archetype. They are the founders who are running real companies with real revenue while also running real parallel practices, and they are doing it without the imprimatur of the partners’ meeting in Sand Hill Road. The class of 2025 is, by my reading, disproportionately polymath, and disproportionately bootstrapped, and the two facts are connected.
A modest closing argument
I am going to close with a modest argument, because the polymath problem is not going to be solved by an essay. The argument is that the venture industry could, if it chose to, run a quiet experiment. The experiment would be to make a small allocation, every fund cycle, specifically to polymath founders — to the founders who have, visibly, a serious other practice. The allocation would not need to be large. The thesis would not need to be loud. The fund would simply need to be willing to underwrite the texture, not just the company.
I suspect the resulting cohort would, on a ten-year horizon, outperform the rest of the portfolio by some interesting margin. I cannot prove this. I am, however, willing to make the prediction. The polymath founder is, in my reading of the last decade, the most durable founder type the industry has been refusing to underwrite. The industry will, eventually, update. I would prefer the update happen now, while the cohort is still cheap. By the time the trade press has agreed that the polymath founder is the under-priced asset of the current decade, the asset will not be under-priced anymore.
In the meantime, the polymath founders are bootstrapping. The polymath founders are making their records and shipping their platforms and writing their novels and running their companies, all at the same standard, all in parallel, all without permission from the partners’ meeting they did not attend. They are doing fine. They will, in my reading, do better than fine. The industry’s loss is, for the moment, their gain.
You can find one of them — the most legibly polymath of the cohort I have followed — profiled at length elsewhere in this magazine. The rest of them are coming. The patient reader will find them.