Paul Graham is Right About Almost Everything

Paul Graham, a co-founder of the influential Y Combinator startup accelerator, just published an essay arguing that universities should not build entrepreneurship centers, should not hire “deans of entrepreneurship,” and should hand any spare money to the people teaching computer science and mechanical engineering and molecular biology (agreed, btw). At one point he asks — as a scoff, and a good one — where the innovation center is.

Well, hello! It’s here.

The Lassonde Entrepreneur Institute, a top-10 ranked program at the University of Utah and creator of Lassonde Studios, is exactly the thing he’s describing: a center, housed inside a business school he footnotes as a source of strategic error, operating a residential building of the kind he says will drag a university down. If the essay has a bullseye, I’m standing in it.

And before anything else — thank you. I mean that. Most criticism of entrepreneurship education is lazy, a sneer at hoodies and pitch decks from people who have never once watched a student ship something at 2 a.m. and then watch nobody use it. This isn’t that. This is a careful public argument from the person with the best data in the world on what a promising founder looks like, aimed at a category of institution that mostly escapes scrutiny (because everybody assumes somebody else is checking, and nobody is). That’s a gift. Slightly uncomfortable gift. Still a gift.

It deserves a real answer. It’s going to get five of them, which is either generous or unhinged depending on your tolerance.

I got my degree at Harvard. I got my education at the University of Utah.

Density has architects

Here’s what I took from both, and everything that follows depends on it.

Everywhere that builders find each other, somebody made a room. Or a schedule. Or a rule that let it happen. From the outside this looks like weather — an ambient culture, a thing a place simply has, the way Seattle has rain. From the inside it looks like design.

Reading period, which Graham correctly identifies as the incubator of two trillion-dollar companies (and I love that he found this — it’s the single best observation in the essay), is not a season. Somebody drew that calendar. Defended it in a meeting. Kept it. Somebody sited those labs. And the tolerance for untidy student projects — the tolerance that let Bill Gates smuggle Paul Allen, a non-student, into the computer lab — isn’t an absence of policy. Somebody chose that policy. Somebody kept it, and I’d bet took heat for it.

That’s not a Harvard point, though, so I won’t spend more of your attention there (Harvard gets plenty of ink; it does not need mine). Bell Labs built corridors long enough that you couldn’t get to lunch without running into somebody working adjacent to your problem. Xerox PARC put beanbags in a common room and made showing up socially mandatory. And the cleanest case in the world is Y Combinator itself. Graham ran a batch, so founders were co-located. He held office hours, so the craft passed from someone who’d done it. He fed them dinner, so the peer group formed. He set a demo day, so the deadline was real.

Which is to say: Paul Graham built a density machine on purpose, refined it for twenty years, and then wrote an essay recommending universities not do that.

I say this with affection. But come on.

Which is also where it costs nothing extra gives the game away. It costs nothing extra at institutions where the infrastructure is already standing and was paid for by somebody else a very long time ago. Everywhere else — which is most places, which is where most students are — somebody has to build it, and the second you build it rather than inherit it, it shows up as a line item with a name and a number attached. Inherited density is invisible. Purchased density has an invoice.

Same thing. Different accounting.

On Y Combinator

Let me say this without a hedge, because I’m about to disagree with the man and I’d rather square the ledger first.

Y Combinator is the singular institution in founder acceleration. Under classical measures — companies started, value created, founders who went back and did it again — it is the standard against which everything else in this field should be judged. Us included. Especially us. And their published material is, in my honest estimation, the best content available anywhere to a person who wants to understand how companies actually get built, and it’s free, and I consume it constantly and by choice. (I have listened to some of those talks more than once, on I-15 in Salt Lake traffic.)

So when I say Graham has a blind spot, understand I’m saying it about the institution I learn the most from.

What he gets right

Nearly all of it. And it’s one argument, not a list.

You cannot teach a person to start a company. It’s learned by doing, which means an honest course on the subject would have to be a lab where students actually start them — and that lab already exists, and it’s his. Cram it into a degree program as one more class among four others and you get a parody: no capital, no time, and an awkward question about what happens if the thing works. (Congratulations on the traction, please drop out or abandon it, those are the options, see you at graduation or don’t.) But that’s a constraint, not a law. Build the program so the venture is the coursework — the student’s hours go to the work, the credit follows the work, and the degree bends around it instead of competing with it — and the trap he’s describing doesn’t close. That can be done at the graduate level and it can be done for undergraduates. We run both. Some cohorts clear his bar. Not all of them do.

He’s right that project work beats coursework as evidence, and right about why: projects are where deep knowledge forms, where cofounders find each other, and where the ideas nobody would have chosen on purpose turn out to be sitting. He’s right that the instruction should be to build things without asking whether they’ll become companies. He’s right that unstructured time is a precondition, not a luxury. And he’s right that institutions resist providing it, because leaving students alone feels, to an administrator, like doing nothing.

And on the spare money, he’s right, and I’d push it one step further than he does. The hard part is the product — knowing what to build and being able to build it — and that capability comes out of a discipline, not out of a center. When one of our students ships something that works, it works because somebody in a lab down the hill taught them how. We are downstream of those departments and we know it.

But I wouldn’t stop at the departments. Money behaves like water in a ditch — it soaks in where the ground is already wet, and less of it reaches the end of the row than you planned. So give it to the students. Directly. A thousand dollars in a sophomore’s hands buys hours. Ten thousand to a program about mechanical engineering buys a program. And if the choice is a new initiative in my building or that dollar landing in hers, hers wins every time.

And he’s right about the last thing, which is the one I’d have written myself: done properly, this work looks like the university isn’t doing anything, and the correct response to that appearance is to hold the line and let results speak — rather than manufacture visible activity to reassure parents, prospective students, and accreditors.

That last audience deserves a word, because Graham lays the preference for visible activity at the feet of administrators, and I think he’s being too generous to the system and too hard on the people inside it. Accreditation measures what can be documented. Formation resists documentation on principle. Any learning outcome that can be written cleanly into a standard is, by construction, one an institution can demonstrate without having changed a single student — contact hours, rubrics, artifacts of instruction, all of it verifiable, none of it evidence. So the incentive runs hard toward legible activity, which is exactly the activity Graham says produces nothing.

Half the time this isn’t timidity. It’s a reporting requirement with a review cycle attached. And the people filing it know better.

The adversary we share

Read the essay closely and you find that Graham and I are swinging at the same thing.

It’s the credit-hour model — the assumption that instruction produces capability, that coverage produces competence, that a grade is a consequence, and that a student who has been told about something has been changed by it. His footnote on business schools is the sharpest version of this I’ve read from outside the academy: they were built to train the managerial class of the large industrial firms of the early twentieth century, which is why they’re so often called schools of management. As history, he’s right. As prophecy, he’s wrong, and the counterexample is the school I work in. This institute exists at the David Eccles School of Business because a dean decided formation was worth funding and then funded it — which is not what an institution built for the industrial managerial class does when nobody is looking.

I take his footnote as a description of the problem I came here to work on.

If my objection to Graham were that he’s too hard on universities, I’d have no argument at all. Our practice sits closer to his essay than it does to the academy that houses us. (Which is an awkward sentence to publish on a university website. I’ve reread it four times. It stays.)

What his critique costs me

So let me put the expensive things on the table myself, rather than wait around for somebody to pull them out of me.

He’s right about pitch competitions. A competition judged by simulated investors on whether a venture is investment-ready teaches students that the central act of building a company is constructing a story that appeals to capital. It isn’t. The people you have to convince are users, and you convince them with a working thing, not a slide. Every program running that format — including formats we have run — owes an accounting of what it’s actually training.

And the number we like to quote — total capital raised by ventures that came through here, set against what we’ve awarded in prizes. It’s a good number! I have said it out loud in rooms where it did exactly what I wanted it to do. It is also a capital-formation statistic. It measures how much money followed our students. It does not measure whether our students were formed. Those are two different claims, and I have watched the first do the work of the second — which is precisely the substitution Graham is warning about. If the proof of a center is the money its alumni raised, the center has accepted somebody else’s scoreboard and shouldn’t act surprised when the score gets used against it.

The building deserves the same honesty. A facility built for this work can become a monument to activity instead of a site of it — the appearance of a founder culture, beautifully lit, extremely photographable (I know, because I have approved the photographs), with no real consequence inside it. That isn’t a hypothetical risk I’m raising to seem balanced. It’s the live one. Any director who tells you otherwise isn’t watching closely enough.

The one place he stops short

Now the disagreement. And I want to make it on his evidence, not my preferences.

Graham reports that Harvard alumni apply to YC at roughly twice the rate of Yale and Princeton alumni. He doesn’t conclude that Harvard students are better — he says they’re not meaningfully different, and that the gap is custom. Then he names the mechanism with real precision: until you’ve seen founders in real life, founding is a thing other people do, and seeing them up close makes it look desirable and achievable at the same time. He even gets the dosage right, noting that a founder in their mid-twenties, three years into something promising, moves a student more than a billionaire does.

That is a claim about proximity. Proximity is physical. And his own argument therefore implies that where proximity wasn’t inherited, it can be built.

That’s the sentence he doesn’t write.

I think I know why. Y Combinator meets people for whom two questions were settled long before the application form. Whether they could do this. And how they’d eat while they tried. His model of preparation is optimized for a population holding both answers already, whose only remaining constraint is time. For those students his advice isn’t just correct, it’s complete: give them room, get out of the way, resist the urge to helpfully schedule something.

It’s the other population I’m describing. And it is so much larger.

And here’s what I’d want him to sit with. Those application numbers are his own. If the gap between Harvard and Princeton is culture rather than talent, and culture gets built, then the centers he’s telling universities to defund are part of what builds it — at every school that didn’t inherit one, which is nearly all of them. Follow the advice and you don’t get more applications from better-prepared students. You get fewer applications from the same students, minus the ones who never worked out that this was available to them.

He is proposing to dismantle a pipeline he doesn’t know he has.

What comes next

Four claims, one a week. They’re the four things I wish he’d asked me before he published.

Belief and survival, not time, are the binding constraints for most students who could do this and don’t — and the second is why free time isn’t free. A student working twenty hours a week to make rent doesn’t have hours to hand a project, and no amount of inspiration manufactures them.

Unstructured is not the same as unsupervised, and the space between a class and nothing at all is where formation actually happens.

There is a competition worth running, and it isn’t the one he’s describing. His own footnote explains why, which I find delightful.

And those quotation marks he puts around the word entrepreneurship? They belong somewhere else entirely. They belong around startup.

One more thing, and then I’ll stop.

Paul — if you ever want to see what a purpose-built density machine looks like at a public flagship in the Mountain West, four hundred students living where they build, at an institution that inherited exactly none of this and had to go buy it: the door is open. Genuinely. Come look. I’ll drive.

I suspect the argument is more interesting from inside the building than from outside it.