In the fall of her senior year at Princeton, Wendy Kopp was, by her own account, in a complete funk, with no job lined up and, more pressingly for a Princeton senior, no thesis topic. The thesis she eventually produced, in 1989, was titled "An Argument and Plan for the Creation of the Teachers Corps." She graduated, founded the organization her thesis described, and within a year had raised $2.5 million and recruited 500 teachers into some of the poorest school districts in America: a thesis-to-classroom cycle time of roughly eighteen months, which would flatter most venture-backed startups. She did all of this at twenty-one, having never run anything larger than a student organization.

The plan worked better than anyone had a right to expect. Two decades later, Teach For America was the single largest employer of graduating seniors at Yale, Dartmouth, Duke, Georgetown, and UNC Chapel Hill, and in 2010 18 percent of Harvard's senior class applied, the largest share of any school in the organization's history. TFA received 46,000 applications that year and accepted 12 percent of them, which meant that nearly one in five graduates of the most selective university in the country were competing for a job that paid a first-year teacher's salary in Houston or the Mississippi Delta.

The machine wound down almost as quickly as it had assembled itself. Applications peaked at 57,000 in 2013 and declined nearly every year afterward; by 2022 the incoming corps was less than a third of its former size, and the share of Harvard seniors applying had fallen from 18 percent to under 2.

The convenient explanation is generational: young people became more cynical, or more mercenary, or simply received better offers. The evidence suggests otherwise. Last fall a Dutch nonprofit called The School for Moral Ambition papered Harvard with red posters reading "You didn't fight your way into Harvard to end up in a bullshit job," and 8 percent of the junior class applied for its twelve fellowship spots, in the program's first year on an American campus. Rutger Bregman, the organization's co-founder, likes to cite a study by two Dutch economists in which a quarter of workers privately doubted their job had any social value, and he calls the drift of graduates into consulting, finance, and corporate law the Bermuda Triangle of talent. The appetite for consequential work never went anywhere; what disappeared was the machinery for converting it into anything.

Consider the most capable twenty-two-year-old you know, and imagine handing her one of two ambitions.

Give her the first, building a company, and she inherits an assembly line the world has spent twenty-five years constructing for exactly her. Y Combinator will give $500,000 to two people with a demo and a slide deck. The Thiel Fellowship will pay her six figures to skip her degree entirely. Tyler Cowen's Emergent Ventures will wire money about a week after the paperwork clears, with a reporting requirement of one page, due in a year. Behind the capital sits softer machinery that may matter more: the essays, the demo days, an entire status economy whose message is that building things is what talented people do. Cowen has observed that the scarcest input he supplies through his grants is not money but a vision of what a young person could become, and the startup world mass-produces such visions.

Give her the second ambition, fixing a public problem, and the offer changes character entirely. She can apply to a fellowship and wait for its annual cycle, join a large organization where the meaningful decisions arrive around year ten, or pay for a policy degree, and the most common honest advice she will hear is some version of: go do something else first, build credibility, come back at thirty-five. The difference between the two paths is not really the money. The first is organized around the premise that her initiative is precious and perishable; the second asks her to spend a decade demonstrating that it is warranted.

It would be one thing if the world of doing good were simply under-built, but it is not. Over the past fifteen years it has constructed genuinely sophisticated infrastructure. GiveWell can tell a donor which charity saves a life for the fewest dollars. Open Philanthropy recently rebranded as Coefficient Giving in order to advise donors giving more than $250,000 a year. Renaissance Philanthropy has mobilized $533 million in two years, and its founder, the former White House science advisor Tom Kalil, argues that capital for large problems is now abundant while institutions capable of deploying it are scarce. He is right, and it is striking, once you notice it, that nearly all of this sophistication points in a single direction: toward the allocation of money rather than of people. During the same period, the one prominent organization whose entire job was routing people, 80,000 Hours, announced it would narrow its focus to careers in AI and conduct little or no research into anything else.

Something like a Y Combinator for public problems does in fact exist. It is called Charity Entrepreneurship, it operates out of London, and its record is startling: more than fifty organizations launched, over $68 million raised by them, and roughly 85 percent of participants leaving with $100,000 to $200,000 in seed funding. I mention it regularly to smart, ambitious people, and almost none of them have heard of it, which is roughly the equivalent of Y Combinator operating for two decades without computer science majors ever learning its name.

To a first approximation, then, the sector has solved capital allocation and has not begun on talent allocation. We have industrialized ambition for private problems while leaving public ambition artisanal: if you are twenty-two and want to build a company, institutions compete for you; if you are twenty-two and want to work on lead poisoning or literacy or the courts, you must compete for institutions that are not hiring. In any financial market, an inefficiency this large would be arbitraged away within a decade. In the market for talent it has persisted for a generation, mostly, I suspect, because no one captures the returns to correcting it.

I want to be careful about the conclusion, because there is a lazy version of this essay that ends with a call for more nonprofits, and I do not believe it. Through my work at Stripe I talk with nonprofit leaders across nearly every cause area, and many of the strongest organizations I meet are not short on ideas or talent but on money, in which case the highest-leverage move is to strengthen the institution that already works rather than incorporate a rival next door with a fresh logo.

The argument is for founders, and the distinction matters, with TFA itself as the best illustration. Its most durable output was never really the teachers, most of whom left the classroom after their two-year commitment, but a founder class: two young corps members started KIPP, which grew into one of the largest charter networks in the country, and alumni went on to run school systems, found education companies, write policy, and lead the very district schools TFA once parachuted them into. One institution, designed by a twenty-one-year-old, spent three decades producing the people who would build, strengthen, and occasionally replace the institutions of an entire field. None of this is to sanctify TFA itself, whose critics were frequently right: five weeks of training is thin, attrition was high, and the charge that it undercut career teachers deserves to be taken seriously. But the pedagogical debate is separate from the institutional lesson, which is that when someone constructs a prestigious, legible door into a public problem, a generation walks through it.

This is the sense in which the standard startup analogy actually applies. Google was not founded because search did not exist, and Stripe was not founded because payments did not exist; both were founded because an increasingly important problem was being solved badly by the institutions that owned it. The same test should govern the nonprofit world, and it cuts in both directions. Sometimes it says to join rather than found. Sometimes it identifies a problem that is institutionally homeless, understood by many people, funded in fits and starts, owned by no organization capable of sustained execution, and waiting for its founder. The question is never could a new nonprofit do some good here, but does this problem require a new institution, or an exceptional founder working inside, alongside, or through one that already exists.

That question, however, only gets asked if the founder shows up, which the rest of the ecosystem quietly assumes and nothing actually ensures. Effective altruism trained a generation to think counterfactually about money, to ask what would have happened to a dollar anyway; apply the same discipline to people and the picture darkens. For most of institutional history, this was ordinary work for the young: Jane Addams opened Hull House at twenty-nine, and George Williams started the YMCA at twenty-two. Every Wendy Kopp we can name is, in an important sense, a survivor, someone who happened to build the pipeline and walk through it at the same time, with a thesis requirement as her only forcing function and no institution behind her. For each of them there is some unknowable number of counterfactual founders, people who had the plan and the funk but no first check, no co-founder, and no one to hand them a vision of what they might become, and who presumably accepted the return offer from the bank or the consultancy and are by every conventional measure doing fine. The problems they would have taken responsibility for are the ones left worse off.

There is a recursive quality here, because the pipeline is itself the most institutionally homeless problem in the sector: everyone agrees the talent exists and the problems matter, while the conversion layer between them, the fast capital, the playbooks, the co-founder matching, the prestige, belongs to no one. I do not have a good account of why. The components are individually unexotic, and the startup world assembled every one of them a generation ago, in the process changing what ambitious young people believe is possible with their lives. My best guess is the unglamorous one: that building talent infrastructure is nobody's job, returns nothing to whoever builds it, and pays out a decade later in someone else's annual report.

So instead of a conclusion, three questions I would like better answers to. First, where do public founders actually come from? There are entire libraries on the origins of startup founders and, as far as I can tell, no serious study of this at all; we know surprisingly little about the single most leveraged input into the institutions we depend on. Second, how contingent are the institutions we did get? If Kopp had gone the way of most of her classmates, how many years would education have waited for its TFA, and for KIPP downstream of it? When a company fails to get founded, a competitor usually captures the opportunity; when a public institution fails to get founded, nothing does. And third, the one I find hardest to shake: somewhere this fall there is a senior in a funk, carrying a half-written plan the world has not asked for. What is our current mechanism for finding her, and would we know if it were failing?

In 1989 the mechanism was, as far as I can tell, nothing at all, and we got Teach For America anyway. Base rates suggest we should not count on that twice.