Paul Graham — Leadership Playbook | Faster Than Normal
Paul Graham
Co-founder of Y Combinator, the most influential startup accelerator in history (Airbnb, Stripe, Dropbox, Reddit). Also wrote influential essays on startups.
In January 1995, a thirty-year-old programmer who had spent the previous several years trying and failing to become a painter walked into a gallery in New York and proposed to build it a website. The gallery was not interested. Neither was the next one, nor the one after that. Paul Graham and his friend Robert Morris — the same Robert Morris who, seven years earlier, had accidentally broken the entire internet with a self-replicating worm and become the first person prosecuted under the Computer Fraud and Abuse Act — had started a company called Artix with the plan of putting art galleries online. The idea was, by Graham's own later admission, thoroughly boneheaded. Art dealers, he discovered, were "the most technophobic people on earth. They didn't become art dealers after a difficult choice between that and a career in the hard sciences." Most had never seen the Web. Some didn't even own computers. Graham and Morris sank to building sites for free, and still couldn't convince galleries to participate.
The failure was clarifying. "Gradually it dawned on us that instead of trying to make Web sites for people who didn't want them, we could make sites for people who did." They ditched Artix and started Viaweb, the first software-as-a-service company — a web application that let anyone build an online store, hosted entirely in a browser, years before the term "SaaS" existed. Yahoo acquired it in 1998 for roughly $49 million. Graham suddenly had money, which meant he suddenly had to think about something he'd never considered: how not to lose it.
What happened next is stranger than the exit. Graham did not start another company. He did not join a venture fund. He did not move to Sand Hill Road. He went back to painting, then to writing essays — long, digressive, intellectually omnivorous pieces published on a personal website with the production values of a 1997 Geocities page. The essays were not about venture capital or startup strategy, at least not exclusively. They were about Lisp dialects, Renaissance art, the nature of taste, why nerds are unpopular in high school, how to disagree productively, and whether cities have personalities. They were, in aggregate, the closest thing the internet has produced to a philosophical system — one that happened to generate, as a side effect, the most influential startup accelerator in history.
Y Combinator, which Graham co-founded in 2005 with his wife Jessica Livingston, Robert Morris, and Trevor Blackwell, has funded over 4,000 startups. Among them: Airbnb, Stripe, Dropbox, Reddit, Coinbase, DoorDash, Twitch, and Instacart. More than 200 are now valued at over a billion dollars. The combined valuation of YC companies exceeds $600 billion. These numbers are staggering enough that they tend to obscure the deeper strangeness of what Graham actually built — not a financial vehicle but an epistemological one, a machine for identifying talent in its rawest form and subjecting it to a theory of knowledge derived, in roughly equal parts, from Lisp programming, Italian painting, and an Englishman's abiding suspicion of institutional authority.
Part IIThe Playbook
Paul Graham's influence operates through a distinctive mechanism: he writes things down, clearly, and the clarity itself becomes leverage. The principles below are not slogans extracted from motivational talks but patterns distilled from decades of making things, funding people who make things, and writing obsessively about what makes the difference between the things that survive and the things that don't.
Table of Contents
1.Start with the wrong idea.
2.Select for determination over intelligence.
3.Compress judgment into structure.
4.Do things that don't scale — then notice when they do.
5.Write to think, not to publish.
6.Make the deal terms a no-brainer.
7.Use taste as a competitive weapon.
Protect the maker's schedule.
In Their Own Words
You'll become like whoever you work with. Do you want to be like these people?
There are few sources of energy so powerful as a procrastinating college student.
— Hackers & Painters: Big Ideas from the Computer Age
Being strong-willed is not enough, however. You also have to be hard on yourself. Someone who was strong-willed but self-indulgent would not be called determined. Determination implies your willfulness is balanced by discipline.
At every period of history, people have believed things that were just ridiculous, and believed them so strongly that you risked ostracism or even violence by saying otherwise. If our own time were any different, that would be remarkable. As far as I can tell it isn't.
A programming language is for thinking about programs, not for expressing programs you've already thought of. It should be a pencil, not a pen.
If you leave a bunch of eleven-year-olds to their own devices, what you get is Lord of the Flies.
It's hard to say exactly what it is about face-to-face contact that makes deals happen, but whatever it is, it hasn't yet been duplicated by technology.
A restaurant can afford to serve the occasional burnt dinner. But in technology, you cook one thing and that's what everyone eats.
If you're in a job that feels safe, you are not going to get exceptional, because if there is no danger there is almost certainly no leverage.
The through-line of Graham's career is not entrepreneurship. It is the relentless compression of complex processes into their simplest possible expression — the search for the function that generates the function, the Y combinator of the company's name. In mathematics, the Y combinator is a higher-order function that enables recursion without requiring a function to have a name. Graham named his firm after it partly as "a secret signal to the kind of people we hoped would apply." The signal encoded the entire thesis: that the most important things are built by people who don't yet have names, working on ideas that don't yet have categories, in a process that refers only to itself.
By the Numbers
The Graham Orbit
4,000+Startups funded through Y Combinator since 2005
200+YC-backed companies now valued at $1B+
$600B+Estimated combined valuation of YC portfolio companies
~25MAnnual page views on paulgraham.com
$49MApproximate Yahoo acquisition price for Viaweb (1998)
227Applications to YC's first experimental summer program (2005)
6%Median equity stake YC took in early batches
The Philosopher Who Learned to Program
Graham was born on November 13, 1964, in England. He grew up partly in Pittsburgh — Monroeville, specifically, a suburb he would later describe with the affectionate contempt of someone who understood exactly what the place lacked. "When I was a kid, this was a place young people left." He has written that during the entire period he lived there, from 1968 to 1984, the city was in free fall, the steel and nuclear industries both dying, the flight to suburbs draining whatever energy remained. Pittsburgh had nerds — Carnegie Mellon was and is among the top computer science programs in the country — but no rich people willing to fund them. "The record skips at that point," Graham wrote, comparing Pittsburgh's startup output to Stanford's and MIT's.
Before college, the two things Graham worked on outside school were writing and programming. The stories were awful, by his own account — "hardly any plot, just characters with strong feelings, which I imagined made them deep." The programming was only marginally more productive, constrained by the IBM 1401 that the school district housed in the basement of his junior high. "The only form of input to programs was data stored on punched cards, and I didn't have any data stored on punched cards." His clearest memory from that period: the moment he learned a program could fail to terminate, a social as well as a technical error on a machine without time-sharing, "as the data center manager's expression made clear."
At Cornell, Graham studied philosophy — not computer science, though he took enough CS courses that most CS majors assumed he was one of them. "In college I was going to study philosophy, which sounded much more powerful," he later wrote. "It seemed, to my naive high school self, to be the study of the ultimate truths, compared to which the things studied in other fields would be mere domain knowledge." What he discovered was that other fields had already occupied most of the interesting territory. All that remained for philosophy were edge cases.
He earned an AB from Cornell and a PhD in computer science from Harvard. Then, in a move that bewildered nearly everyone who knew him, he enrolled at the Rhode Island School of Design and later studied painting at the Accademia di Belle Arti in Florence. The decision seems eccentric only if you don't understand what Graham was actually looking for. He was not becoming an artist. He was studying the nature of making — the common substrate beneath programming, painting, architecture, and writing. "What hackers and painters have in common is that they're both makers," he would write in the essay that became the title piece of his 2004 book. "Along with composers, architects, and writers, what hackers and painters are trying to do is make good things."
The art school years gave Graham something that computer science alone could not: a theory of taste. The conviction that beauty matters in software as much as in painting. The understanding that "brand and craft became divorced" in art when mass media made it possible to build a reputation independent of quality — an insight he would later apply ruthlessly to startups, where the divorce between narrative and substance is a daily temptation.
The SaaS Before SaaS Had a Name
Viaweb was, in retrospect, one of the most structurally important startups of the 1990s, though it is almost never discussed in those terms. The company built what was essentially Shopify — fifteen years before Shopify existed. The entire application ran in a web browser. There was nothing to download, nothing to install. You could build and manage an online store from any computer with an internet connection. This was 1995. Netscape had gone public three months earlier. Most people had not yet heard the phrase "electronic commerce."
Graham and Morris wrote the software in Lisp, a language that most programmers regarded — and many still regard — as an academic curiosity. Graham had a different view. He believed Lisp's expressiveness gave a small team an asymmetric advantage, that the abstractions available in Lisp allowed two or three programmers to build what would require dozens in Java or C++. "In a big company, you can do what all the other big companies are doing," he later wrote. "But a startup can't do what all the other startups do. I don't think a lot of people realize this, even in startups."
The acquisition by Yahoo in 1998 — which turned Viaweb into Yahoo Store — gave Graham his financial independence. It also gave him something more subtle: a dataset. He had lived through the full lifecycle of a startup, from terrible first idea (Artix, the gallery venture) through pivot, growth, acquisition, and the disorienting aftermath of sudden wealth. He knew, from direct experience, that the path from founding to exit was not a straight line but a series of lurches, that the idea you started with was rarely the idea that succeeded, and that the founder's capacity to survive the lurches mattered more than the brilliance of any individual insight.
The Yahoo experience also taught him what happened when founders lost control. He watched the company that had acquired him decay from the inside — a process he would later anatomize in a devastating essay titled "What Happened to Yahoo." The lesson was seared in: professional management, applied to companies that still needed the founder's instinct, was not a neutral force. It was often actively destructive.
The Essay as Startup Weapon
After selling Viaweb, Graham spent several years painting, programming in Lisp, and writing. The essays began appearing on paulgraham.com in 2001. They accumulated readers the way a coral reef accumulates mass — slowly, then all at once. The site now receives around 25 million page views per year, a number that would be remarkable for any media property and is astonishing for a personal website with no images, no ads, no social sharing buttons, and a design that hasn't meaningfully changed since the Clinton administration.
The essays function as a body of work in the way that a philosopher's collected papers do: each one self-contained but gaining force from proximity to the others. "How to Make Wealth" argues that startups are a mechanism for compressing a lifetime of work into a few years. "Maker's Schedule, Manager's Schedule" identifies the fundamental incompatibility between the way programmers use time and the way executives use it — a distinction that has since become embedded in the vocabulary of the industry. "Do Things that Don't Scale" articulates the paradox at the heart of every early-stage company: that the habits that create growth are the opposite of the habits that sustain it.
The way to get startup ideas is not to try to think of startup ideas. It's to look for problems, preferably problems you have yourself.
— Paul Graham, 'How to Get Startup Ideas'
The rhetorical strategy of the essays is distinctive and worth examining. Graham writes with the syntax of a programmer — short declarative sentences, nested conditionals, assertions that build toward a proof. But the content is that of a humanist. He moves freely between discussions of Florentine painting and web application architecture, between the sociology of high school and the economics of wealth creation. The effect is of someone who has refused to accept the boundaries between disciplines, who treats all of intellectual life as a single codebase that can be refactored.
"Reading Paul's essays is like having a conversation with a genius who doesn't need to score any points by proving it to you," Eric Raymond, author of The Cathedral and the Bazaar, wrote. The description captures something essential: Graham's writing is persuasive not because it is forceful but because it is generous. He shares his reasoning process, including the dead ends and uncertainties. He admits when he's wrong. He treats the reader as a collaborator rather than an audience.
The essays also served a nakedly practical function. They were, in effect, a filter — a self-selecting mechanism that attracted exactly the kind of people Graham wanted to work with. "We named the company after [the Y combinator] partly because we thought it was such a cool concept, and partly as a secret signal to the kind of people we hoped would apply." The essays were the same kind of signal, broadcast at a wider frequency. If you read "How to Start a Startup" and felt the hairs on the back of your neck stand up, you were probably the kind of person Graham wanted to fund.
The First Batch
Y Combinator was born from a talk. In the spring of 2005, Graham was invited to speak to the Undergraduate Computer Society at Harvard. He needed a topic. "I thought, 'All right. I'll tell them how to start a startup.'" The talk became an essay. The essay became, inadvertently, a recruiting document.
Graham had been thinking about angel investing — the practice of putting small amounts of money into very early companies. He realized, while giving advice to these undergrads in real time, that the standard model was broken. "The best thing was to raise money from angels who had themselves made their money from doing a startup because then they could give you advice and connections as well as money." But there weren't enough such angels, and the ones who existed operated idiosyncratically, one deal at a time.
What if you could batch the process? Fund many companies simultaneously, give them a shared infrastructure of advice and community, and compress the critical early months into a structured program? The idea was so simple it seemed almost trivial — and yet nobody had done it. In the summer of 2005, Graham, Livingston, Morris, and Blackwell launched an experiment: they would fund a cohort of startups over the summer, primarily young founders, and see what happened.
Jessica Livingston — a former vice president of marketing at a Boston investment bank who would go on to write Founders at Work, the definitive oral history of startup origins — was the co-founder whose contribution is most systematically underestimated. Graham himself has been explicit about this. Livingston possessed what the team came to call "social radar" — an uncanny ability to read people in high-pressure situations, to detect the micro-signals of dishonesty, defensiveness, or interpersonal dysfunction that predicted failure. "I'll meet somebody I really like, and she'll say, 'You know, there's something off about him.' And it always turns out she's right. Always," Graham told Mixergy. "So I've learned, when Jessica says these people are good or these people are bad, I should really listen."
Robert Morris — the erstwhile internet worm author, now a tenured professor at MIT — served as the technical conscience, the person whose instinct for what was computationally interesting or fraudulent was refined by decades of operating at the edge of what systems could do. Trevor Blackwell, a roboticist and hardware hacker, rounded out the founding team with a background in physical engineering that grounded the group's otherwise software-centric worldview.
The first batch received 227 applications. The deal terms were stark: approximately $20,000 in funding — later supplemented with capital from outside investors — in exchange for roughly 6% to 7% equity. The financial arithmetic was deliberately designed to be a no-brainer. "If we take 6%, we have to improve a startup's outcome by 6.4% for them to end up net ahead," Graham explained. "That's a ridiculously low bar." Founders who didn't grasp this, he said, had "failed an IQ test."
Among the first batch's graduates: Reddit. Alexis Ohanian and Steve Huffman, barely out of the University of Virginia, had applied with a mobile food-ordering app called My Mobile Menu. Graham rejected the idea but liked the founders. He suggested they build something inspired by Delicious, the social bookmarking service — "Paul was enamored with this idea of like what is going to replace The New York Times when the front page needs to be more than just what one editorial board can decide," Ohanian later recalled. What emerged was Reddit, with a $12,000 grant and a name that Graham thought was "terrible" and a mascot — the alien Snoo — that he wanted killed. "He was like 'The name is poison to potential investors, and that mascot, that bug thing, looks so dumb,'" Ohanian told Fortune. Graham even suggested an alternative name: Octopop. "He liked that one because he said I could turn the Reddit alien into an octopus if I insisted on keeping it." Reddit, of course, kept its terrible name and its dumb mascot all the way to a market capitalization exceeding $38 billion.
The Determination Thesis
As batches accumulated — two per year, growing from a handful of companies to dozens, eventually to hundreds — Graham and his partners assembled an empirical dataset about founder success that was, by his own account, unprecedented. "Between the volume of people we judge and the rapid, unequivocal test that's applied to our choices, Y Combinator has been an unprecedented opportunity for learning how to pick winners."
The most important discovery was also the least glamorous. "We thought when we started Y Combinator that the most important quality would be intelligence. That's the myth in the Valley. And certainly you don't want founders to be stupid. But as long as you're over a certain threshold of intelligence, what matters most is determination."
This was a genuinely heretical claim in 2005. Silicon Valley's self-image was built on the cult of the genius — the dropout visionary, the ten-thousand-IQ technologist who saw the future before anyone else. Graham's empirical finding was that genius was necessary but not sufficient, and that the variable that actually predicted survival was a dogged, occasionally unreasonable refusal to die. "Most startups have at least one low point where any reasonable person would give up. That bottleneck is the reason there are so few successful startups. The only people who get through it are the ones who have an unreasonable aversion to failing."
Graham also observed that the qualities he valued most in founders existed in tension with each other. Determination and flexibility. Imagination and naughtiness. The metaphor he reached for was a running back: "He's determined to get downfield, but at any given moment he may need to go sideways or even backwards to get there." The ideal founder was not a person with a fixed vision but a person with a fixed destination and infinite tactical adaptability.
Though the most successful founders are usually good people, they tend to have a piratical gleam in their eye. They're not Goody Two-Shoes type good. Morally, they care about getting the big questions right, but not about observing proprieties.
— Paul Graham, 'What We Look for in Founders'
The college pedigree finding was equally surprising — at least to Graham. "One of the most surprising things we've learned is how little it matters where people went to college," he wrote. Y Combinator's data showed that Graham and his partners had been systematically overvaluing graduates of elite universities. "We'd interview people from MIT or Harvard or Stanford and sometimes find ourselves thinking: they must be smarter than they seem. It took us a few iterations to learn to trust our senses."
The Airbnb Test
No YC company better illustrates Graham's philosophy — or the gap between his intuition and his conscious analysis — than Airbnb. Brian Chesky and Joe Gebbia, two Rhode Island School of Design graduates with no technical co-founder, had been sleeping on air mattresses in their San Francisco apartment and renting them out to strangers to cover rent. The idea — that large numbers of people would want to stay in other people's homes — seemed, by any conventional analysis, insane.
Graham funded them anyway. Not because he believed in the idea. "When we funded Airbnb, we thought it was too crazy. We couldn't believe large numbers of people would want to stay in other people's places." He funded them because of a single detail in their application: they had been supporting themselves by selling Obama- and McCain-branded breakfast cereal during the 2008 election — custom boxes called "Obama O's" and "Cap'n McCain's" — at $40 a box. "As soon as we heard they'd been supporting themselves by selling Obama and McCain branded breakfast cereal, they were in."
The cereal boxes were a signal — a compressed proof of the founders' resourcefulness, creativity, and willingness to do absurd things to survive. Graham's essay "Do Things that Don't Scale" would later codify this principle, but the Airbnbs were living it before it had a name. During the YC batch, Chesky and Gebbia were flying to New York every week, going door to door, recruiting hosts and personally improving their listings. "When I remember the Airbnbs during YC, I picture them with rolly bags, because when they showed up for Tuesday dinners they'd always just flown back from somewhere."
The Airbnb story also illustrates what Graham calls "fragility" — the precariousness of companies that later seem inevitable. "Airbnb now seems like an unstoppable juggernaut, but early on it was so fragile that about 30 days of going out and engaging in-person with users made the difference between success and failure." The thirty days of fragility became one of Graham's most-repeated teaching examples: evidence that the gap between a billion-dollar company and oblivion can be measured in weeks of manual effort.
The Scaling Paradox
As YC grew — from eight companies per batch to forty, then a hundred, eventually to several hundred — the institution faced a version of the same problem it taught its startups to solve: how to scale something whose value depended on intimate attention. Each batch had to retain the intensity of the first one. Each ten-minute interview had to compress the same quality of judgment that Graham and Livingston had applied when they knew every applicant personally.
The solution was characteristically Grahamian: standardize everything except the judgment. Investment terms were simplified and made uniform. Office hours were structured on the maker's schedule — concentrated into specific days, leaving the rest free for the founders to work. Hacker News, the social news site Graham built and administered, served simultaneously as a community platform, a recruiting tool, and a real-time intelligence feed about the problems and interests of technically minded people.
But the core of YC remained irreducibly personal. Every batch still included the Tuesday dinners. Every company still got ten minutes of interview, during which Graham and his partners made decisions that would shape billions of dollars in value. Graham once told an interviewer that by about a minute into a YC interview, the interviewers' minds were unlikely to change. The rest of the time was for confirming or — more rarely — overriding the initial read.
The model was, in a sense, anti-institutional by design. YC ran on the maker's schedule even though it was an investment firm. "Nearly all investors, including all VCs I know, operate on the manager's schedule," Graham wrote. "But Y Combinator runs on the maker's schedule." The team was kept deliberately small — "we still only have four people, so we try to standardize everything" — not out of austerity but because smallness preserved the texture of judgment. Hiring employees would have meant managing employees, which would have meant meetings, which would have meant the death of the unbroken blocks of time that Graham believed were essential to doing anything difficult.
The City That Whispers
Graham has written with unusual precision about the relationship between geography and ambition. "Great cities attract ambitious people. You can sense it when you walk around one. In a hundred subtle ways, the city sends you a message: you could do more; you should try harder." The essay "Cities and Ambition" — published in May 2008, when this observation was still slightly unfashionable — mapped the invisible messages of different cities with the care of an anthropologist cataloguing kinship structures.
New York, he argued, tells you to make more money. Cambridge tells you to be smarter. Silicon Valley tells you to be more powerful — but a specific kind of power, the power of effect. "What matters in Silicon Valley is how much effect you have on the world. The reason people there care about Larry and Sergey is not their wealth but the fact that they control Google, which affects practically everyone."
The essay was not merely descriptive. It was prescriptive — and personal. Graham had tried living in Berkeley, reasoning that it would be Cambridge with good weather. It wasn't. "The message Berkeley sends is: you should live better." Cambridge with good weather, it turned out, was not Cambridge, because the people who chose Cambridge chose it despite the bad weather, the expense, the grubbiness. The self-selection was the mechanism.
Years later, Graham moved — not to another American city but to the English countryside. "Originally we only meant to stay for a year," he told Antonio García Martínez in 2020. "We wanted the boys to see what it was like to live in another country. Plus I was born here and had always wanted to try living here again. But we liked it so much we've stayed longer." The reason, distilled to a single word: calmer. "Part of the calmness comes from things being old. A lot of the houses where we live are four or even five hundred years old. Some fields are thousands of years old."
The man who had spent two decades telling founders to move to San Francisco — who had argued, with considerable evidence, that geography was destiny for startups — chose, when it came time to optimize his own life, a place where the central message was not power or money or intellect but durability. Fields that had been fields for millennia. Houses built before the discovery of the New World. The signal was legible: the work of making things that last requires a different kind of environment than the work of making things that grow.
Founder Mode and the Manager Fallacy
In September 2024, Graham published an essay that, within hours, became the most discussed piece of writing in Silicon Valley. "Founder Mode" did not introduce a new concept so much as name something that founders had felt but could not articulate: the systematic failure of conventional management advice when applied to companies still led by their creators.
The catalyst was a talk by Brian Chesky at a YC event. "Most founders I talked to afterward said it was the best they'd ever heard. Ron Conway, for the first time in his life, forgot to take notes." Chesky described how, as Airbnb scaled, well-meaning advisors told him to hire good people and give them room to do their jobs. He followed this advice. "The results were disastrous." He had to figure out a better way on his own, partly by studying how Steve Jobs ran Apple.
Graham's argument was structural, not merely anecdotal. Two modes of running a company exist: founder mode and manager mode. The conventional wisdom assumed that scaling required switching from the former to the latter. But the advice to "hire good people and give them room" often meant, in practice, "hire professional fakers and let them drive the company into the ground." Founders who resisted this advice were told they were micromanaging. Founders who accepted it watched their companies deteriorate.
The essay resonated so explosively because it gave language to a common experience. "I know multiple founders who felt like they were being gaslit from both sides," Graham wrote, "by the people telling them they have to run their companies like managers, and by the people working for them when they do." Jeremy Fiance, founder of The House Fund, described the essay as "a rallying cry" for founders who had been "beat up and looking for permission to unleash their instincts."
What made the essay characteristically Graham was its admission of ignorance. He did not claim to know what founder mode actually looked like in practice. "There are as far as I know no books specifically about founder mode. Business schools don't know it exists. All we have so far are the experiments of individual founders who've been figuring it out for themselves." The essay was not an answer but a question — a naming of the gap in knowledge, which is the first step toward filling it.
Writing as Thinking, Thinking as Survival
In October 2024, a month after "Founder Mode," Graham published "Writes and Write-Nots," a brief, severe essay about the consequences of artificial intelligence for the act of writing. His argument was compressed to a cutting edge: writing is thinking. AI would eliminate the pressure to write. Therefore AI would eliminate, for most people, the pressure to think.
"If you're thinking without writing, you only think you're thinking," he wrote, quoting the computer scientist Leslie Lamport. The essay predicted a world divided into "writes and write-nots" — which was really a world of "thinks and think-nots." The analogy was to physical strength in a post-industrial economy: once everyone's job made them strong; now strength is a choice. Writing, and the thinking it enables, would follow the same trajectory. "There will still be smart people, but only those who choose to be."
The essay was, among other things, a defense of Graham's own medium. He had spent more than two decades publishing essays on a website with the aesthetic of a Usenet post, in an era that rewarded video, podcasts, and social media brevity. The essays survived and thrived because they did what Graham said writing does: they forced clear thinking into existence. The form was not incidental to the content. It was the content. Each essay was a proof-of-work — evidence that the ideas had been subjected to the discipline of being written down, rearranged, compressed, and refined until they were as clear as the author could make them.
Graham has described his writing process as closer to programming than to conventional prose composition. Each essay begins as a sketch, a hypothesis that develops through iteration. "As in an essay, most of the ideas appear in the implementing." The parallel to startups is exact and deliberate: you don't know what you're building until you build it. The essay is the minimum viable product of an idea.
Develop a habit of working on your own projects. Don't let "work" mean something other people tell you to do. If you do manage to do great work one day, it will probably be on a project of your own.
— Paul Graham, 'How to Do Great Work'
The Field of Thousand-Year-Old Fields
There is a passage in "How to Do Great Work" — a 2023 essay that represents perhaps Graham's most ambitious attempt at a unified theory — where the advice resolves into something simpler than strategy: "What are you excessively curious about — curious to a degree that would bore most other people? That's what you're looking for."
The sentence reads differently depending on where you are in your life. At twenty, it sounds like permission. At forty, it sounds like a diagnostic. At Graham's age — approaching sixty, living in the English countryside, his children growing, his formal role at YC long since handed over (first to Sam Altman in 2014, now to Garry Tan) — it reads like an autobiography compressed to a single question. Graham was excessively curious about Lisp, about painting, about the nature of making, about how young people with no credentials could build things that changed the world. He followed that curiosity through a philosophy degree, a computer science PhD, art school in Florence, a failed gallery startup, a pioneering web company, a quarter-century of essays, and an institution that reshaped venture capital from the bottom up.
In 2019, he published Bel, a new Lisp dialect written in itself — a project of pure intellectual passion, undertaken for no commercial reason whatsoever, years after he could have spent the rest of his life doing nothing at all. The choice was perfectly legible to anyone who had read the essays. A project of one's own.
He lives now among the old houses and the old fields. The essays still appear, at irregular intervals, on the same austere website. They are read by millions of people, many of whom will never start a startup but who have absorbed, through Graham's prose, a way of thinking about work, ambition, and the craft of making things — a way that values clarity over complexity, curiosity over credentials, and the willingness to be embarrassed by early work over the paralyzing fear of imperfection.
The fields, he wrote, are thousands of years old. Someone cleared them once, long ago, with no certainty about what would grow.
8.
9.Treat cities as operating systems.
10.Build the filter, not the funnel.
11.Be suspicious of conventional wisdom — especially your own.
12.Work on your own projects.
Principle 1
Start with the wrong idea.
Graham's first startup, Artix, was a disaster — a company that tried to put art galleries online when art dealers didn't have computers. Microsoft's first company was called Traf-o-data, and it didn't do well either. Graham draws a direct parallel: "If we, who were 29 and 30 at the time, could get excited about such a thoroughly boneheaded idea, we should not be surprised that hackers aged 21 or 22 are pitching us ideas with little hope of making money."
The failure was productive precisely because it was fast and cheap. The lesson — that you should make things for people who want them, not people you wish wanted them — seems obvious in retrospect. But Graham's deeper point is that the wrong idea is often a necessary precursor to the right one. The process of pursuing a bad idea teaches you things about the market, about yourself, and about the nature of problems that abstract reasoning cannot. Viaweb emerged from the wreckage of Artix not despite the failure but because of it.
At YC, Graham institutionalized this insight. The interview process evaluated founders more than ideas, because ideas were expected to change. "The idea doesn't matter much; it will change anyway." Reddit's founders were accepted despite their original concept (a mobile food-ordering app) being rejected. Airbnb's founders were accepted despite their idea seeming insane. The pattern was consistent: bet on people who would iterate faster than others, and the right idea would emerge.
Tactic: Evaluate yourself and others not on the quality of the current idea but on the speed and honesty with which you've abandoned previous bad ones.
Principle 2
Select for determination over intelligence.
"We thought when we started Y Combinator that the most important quality would be intelligence," Graham wrote. "That's the myth in the Valley." The data said otherwise. Determination — specifically, an unreasonable refusal to accept failure — was the single best predictor of startup success, provided a baseline threshold of intelligence was met.
The distinction matters because it changes who you fund, who you hire, and who you bet on. Intelligence is relatively easy to assess: test scores, school pedigrees, the ability to perform well in a structured interview. Determination is harder. It reveals itself only under stress, only over time, only when the situation becomes genuinely dire.
Graham's example is vivid: Bill Clerico of WePay, a finance startup that required endless negotiations with large bureaucratic companies. "When Bill Clerico starts calling you, you may as well do what he asks, because he is not going away." The metaphor of the running back — determined to get downfield but willing to go sideways or backward at any moment — captures the specific kind of determination Graham values: not stubbornness, which refuses to change course, but persistence, which refuses to stop moving.
⚖
The Determination–Flexibility Matrix
Graham's most-valued founder quality is the combination of two traits that seem contradictory.
Tactic: When evaluating founders or team members, weight the evidence of what they did when things went wrong more heavily than what they did when things went right.
Principle 3
Compress judgment into structure.
YC's core innovation was not the money — $20,000 was trivial even in 2005 — but the batch model. By funding dozens of companies simultaneously and running them through a structured three-month program culminating in Demo Day, Graham solved several problems at once. The batch created peer pressure (founders worked harder when they had to present to each other at Prototype Day). The structure created deadlines (Demo Day forced a level of effort that abstract ambition could not). The cohort created a network (founders in the same batch became each other's first customers, advisors, and emotional support systems).
The genius of the structure was that it encoded Graham's judgment into a reproducible process without eliminating the judgment itself. The ten-minute interview remained. The Tuesday dinners remained. The personal office hours remained. But these irreducibly human elements were embedded within a system that could scale — not infinitely, but far beyond what any single investor could manage operating deal by deal.
The economic design was equally deliberate. Standardized terms eliminated negotiation, which eliminated time waste, which eliminated the transaction costs that made angel investing impractical for small checks. "Till now investment terms have been individually negotiated. This is a problem for founders, because it makes raising money take too long."
Tactic: Identify the single most important judgment call in your process and protect it; automate or standardize everything else.
Principle 4
Do things that don't scale — then notice when they do.
The Airbnb story — founders flying to New York weekly, going door to door, photographing apartments — is the canonical example. But Graham's essay on the subject identifies a deeper pattern: nearly every successful startup had a phase where the founders did things that no rational person would do at scale. Stripe's "Collison installation" — where the Collison brothers, instead of sending beta invites, would say "Right then, give me your laptop" and set up the software on the spot. Wufoo releasing their form-builder before the underlying database was even functional, because the feedback from 83,000 people who came to "sit in the driver's seat and hold the steering wheel" was more valuable than a finished product.
The principle operates on two levels. The tactical level: manual effort generates the initial data and relationships that automated systems cannot. The strategic level: founders who are willing to do unscalable work reveal themselves to be the kind of people who will do whatever it takes — which is, again, the determination thesis in disguise.
Graham's nuance, often missed by people who quote the essay's title as a mantra, is the second half: you will eventually have to stop doing those things. "You'll be doing different things when you're acquiring users a thousand at a time, and growth has to slow down eventually." The skill is not in doing unscalable things forever but in knowing when and how to transition.
Tactic: For any new initiative, identify the one thing you'd have to do manually that would be embarrassing at scale — then do it, and track when the manual version starts generating enough signal to automate.
Principle 5
Write to think, not to publish.
Graham's essays are not content marketing. They are a thinking process conducted in public. "As in an essay, most of the ideas appear in the implementing" — a sentence that describes both the essay form and the startup form, and means the same thing in both contexts. You do not know what you think until you have tried to write it down.
This insight has practical implications far beyond essay writing. Graham's argument in "Writes and Write-Nots" — that the decline of writing will produce a decline in thinking — is an argument about the nature of cognition itself. Writing forces linearization. It forces you to decide what comes first, what depends on what, which assertions are supported and which are merely asserted. These are the same skills required to build a company, to evaluate an investment, or to make any complex decision under uncertainty.
The practice is also a compounding asset. Graham's 200+ essays, accumulated over more than two decades, form a publicly accessible body of thought that functions as both a recruiting tool and a reputational moat. No one can replicate the advantages of having written clearly about the same set of problems for twenty-three years.
Tactic: Before making any important decision, write down your reasoning in complete paragraphs — not bullet points — and see if the argument survives contact with full sentences.
Principle 6
Make the deal terms a no-brainer.
Graham's design of YC's economic terms was an exercise in removing friction. The median equity stake was 6%. The math was transparent: YC had to improve a startup's outcome by 6.4% for the founder to be net ahead. "That's a ridiculously low bar," Graham said. "So the IQ test is whether they grasp that."
The implication was deliberate and slightly insulting: anyone who turned down the deal was not exercising good judgment but failing to do arithmetic. By making the terms obviously favorable, Graham eliminated the negotiation that consumed most of the time and energy in traditional angel investing. Founders didn't have to hire lawyers. They didn't have to compare term sheets. They could say yes immediately and get back to building.
The broader principle is that every transaction in a system should be designed so that the rational choice is immediate and obvious. When you force people to deliberate, you lose the ones whose time is most valuable — which is to say, the ones you most want.
Tactic: In any deal, partnership, or offer, ask: would a smart person accept this in five minutes? If not, simplify until they would.
Principle 7
Use taste as a competitive weapon.
Graham's training as a painter — RISD, the Accademia in Florence — is not an eccentricity on his resume but a core competitive advantage. His understanding of visual design, of what makes software beautiful versus merely functional, gave him an evaluative lens that most investors lacked entirely.
"Somehow, I seem to be able to look at a web app and think, 'No, this is wrong. This is right,'" he told Mixergy. "I have a weird ability to do this. I don't know why. I often worry that it will stop working." The modesty is probably sincere — taste is, by definition, difficult to explain — but the ability is real and consequential. In a world where thousands of startups compete for attention, the ones that feel right — that have the quality Graham would recognize from Florentine painting — have an asymmetric advantage.
The deeper argument, articulated across multiple essays, is that taste and quality are not luxuries but survival mechanisms. "It used to be that the best artists were the best craftspeople. Once art started to be reproduced in newspapers and magazines, you could create a brand that wasn't based on quality." In software, the same divorce is possible. Graham's role, implicitly, was to fund companies where brand and craft were still unified — where the thing worked beautifully because the people who built it cared about beauty.
Tactic: Develop your aesthetic judgment in a domain outside your primary work. The cross-domain transfer of taste is one of the least-recognized sources of competitive advantage.
Principle 8
Protect the maker's schedule.
The distinction between the maker's schedule and the manager's schedule is Graham's single most widely applied idea. "There are two types of schedule," he wrote in 2009. "The manager's schedule is for bosses. It's embodied in the traditional appointment book, with each day cut into one hour intervals." The maker's schedule works in half-day blocks minimum. A single meeting can "blow a whole afternoon, by breaking it into two pieces each too small to do anything hard in."
The insight is not merely about time management. It is about the cognitive cost of context-switching, which is itself an argument about the nature of difficult intellectual work. "Ambitious projects are by definition close to the limits of your capacity. A small decrease in morale is enough to kill them off." The meeting doesn't just consume an hour. It consumes the adjacent hours, the hours spent anticipating it and recovering from it.
Graham applied this principle to YC itself. Office hours were concentrated. The team stayed small. The schedule was designed around the assumption that the most important work — evaluating founders, reading applications, writing essays — required unbroken blocks of deep attention.
Tactic: Audit your calendar for meetings that fragment your most productive time blocks. Move them all to a single day per week, and defend the remaining days absolutely.
Principle 9
Treat cities as operating systems.
Graham's "Cities and Ambition" essay is, at one level, an entertaining classification exercise. But at a deeper level, it's an argument about the mechanism by which environments shape outcomes. Cities don't merely attract certain kinds of people; they program them, continuously, through subtle ambient signals. The message is not explicit. No one stands on a street corner in Silicon Valley saying "you should be more powerful." But the cumulative effect of thousands of interactions — the conversations overheard in cafés, the news that circulates, the things that are admired and envied — produces a specific vector of ambition.
The practical implication is that choosing where to live is not a lifestyle decision but a strategic one. Graham's move from Silicon Valley to the English countryside was not a retirement but a reclassification: a shift from a city that optimized for effect to an environment that optimized for depth and longevity.
Tactic: Identify the ambient message of the place where you live and work. If it doesn't align with what you're trying to build, the friction will be invisible but constant.
Principle 10
Build the filter, not the funnel.
The essays, Hacker News, the YC application, the ten-minute interview, the Tuesday dinners, the name itself (a mathematical concept from lambda calculus) — every element of Graham's ecosystem functions as a filter. The goal is not to attract the maximum number of applicants but to attract the right ones and repel the wrong ones.
This is the opposite of how most institutions think about growth. Universities want more applicants so they can be more selective. Companies want more candidates so they can fill more positions. Graham wanted fewer, better applications — people who had read the essays, who understood the reference in the name, who were the kind of people for whom $20,000 and three months in a room with other obsessives sounded not like a sacrifice but like a gift.
The filter operates through specificity. When you are specific about what you value and what you don't, you lose the people who would have been bad fits and gain the ones who recognize themselves in your description. "We named the company after [the Y combinator] partly as a secret signal to the kind of people we hoped would apply." The signal doesn't need to reach everyone. It needs to reach the right people with unusual clarity.
Tactic: Replace one piece of broad marketing with one piece of deeply specific communication aimed at exactly the person you're trying to reach. Measure by quality of response, not quantity.
Principle 11
Be suspicious of conventional wisdom — especially your own.
The "Founder Mode" essay is not, at bottom, about the superiority of founders over managers. It is about the danger of accepting advice uncritically — especially when the advice sounds reasonable, especially when it comes from smart people, especially when everyone agrees.
Graham's own history of errors is instructive. He was wrong about Reddit's name. He was wrong about Airbnb's idea. He was wrong about the relative importance of intelligence versus determination. He was wrong about college pedigrees. In each case, the data corrected him, and he adjusted. The ability to be corrected by evidence — rather than defending one's prior beliefs — is, in Graham's framework, a form of determination applied to one's own mind.
"Usually when everyone around you disagrees with you, your default assumption should be that you're mistaken," he wrote in "Founder Mode." "But this is one of the rare exceptions." Identifying those rare exceptions — the moments when the consensus is wrong and you are right — is the most consequential skill in investing and in life. It cannot be taught in the abstract. It can only be developed through the accumulation of specific instances where you were right or wrong, honestly catalogued.
Tactic: Maintain a written record of your predictions and their outcomes. Review it quarterly. The pattern of your errors will teach you more than the pattern of your successes.
Principle 12
Work on your own projects.
This is perhaps Graham's deepest conviction, the one that underlies everything else. "Develop a habit of working on your own projects. Don't let 'work' mean something other people tell you to do. If you do manage to do great work one day, it will probably be on a project of your own."
The advice sounds banal. It is, in practice, extraordinarily difficult. School trains you to do assignments. Jobs train you to complete tasks. The entire structure of most people's lives is organized around responding to other people's requirements. To work on your own project — truly your own, driven by your own curiosity, accountable to your own standards — requires an act of will that most people never perform.
Graham performed it repeatedly. The essays were his own project. Bel, the Lisp dialect he published in 2019, was his own project. YC itself began as a project of his own — an experiment, conducted in the summer, to see if a new model of startup investing was possible. Each of these projects shared the same structure: Graham became excessively curious about something, pursued it without institutional authorization, and produced something that existed because he wanted it to exist.
The connection to startups is direct. The best startups, in Graham's analysis, are themselves the founders' own projects — things built to solve problems the founders personally have, in ways that interest the founders personally. "The very best startup ideas tend to have three things in common: they're something the founders themselves want, that they themselves can build, and that few others realize are worth doing."
Tactic: Identify one thing you would work on even if no one paid you, no one acknowledged it, and it might fail completely. That's the project.
Part IIIQuotes / Maxims
In their words
It's better to make a few people really happy than to make a lot of people semi-happy.
— Paul Graham, 'Startups in 13 Sentences'
In college I was going to study philosophy, which sounded much more powerful. It seemed, to my naive high school self, to be the study of the ultimate truths, compared to which the things studied in other fields would be mere domain knowledge.
— Paul Graham, 'What I Worked On'
If you're thinking without writing, you only think you're thinking.
— Paul Graham, 'Writes and Write-Nots'
One of the biggest things holding people back from doing great work is the fear of making something lame. And this fear is not an irrational one. Many great projects go through a stage early on where they don't seem very impressive, even to their creators. You have to push through this stage to reach the great work that lies beyond.
— Paul Graham, 'Early Work'
VCs who haven't been founders themselves don't know how founders should run companies, and C-level execs, as a class, include some of the most skillful liars in the world.
— Paul Graham, 'Founder Mode'
Maxims
The wrong idea is a prerequisite. Every great company passed through a stage where its idea was terrible. The founders who succeed are the ones who iterate fast enough to find the right idea before they run out of time or money.
Determination is the bottleneck. Intelligence gets you to the starting line. Determination is what carries you through the low points where any reasonable person would quit — and those low points are where most startups die.
The essay is the MVP of an idea. Writing forces you to think clearly. If you can't explain your reasoning in complete sentences, you don't actually have reasoning — you have an intuition that hasn't been tested.
Scalable systems are built from unscalable beginnings. Go door to door. Install the software yourself. Photograph every listing by hand. The insights that make automation possible come only from doing the work manually first.
Make the terms so good that refusing is irrational. In any transaction where you want speed and trust, eliminate the need for deliberation by making the decision obvious.
Taste is an underpriced asset. The ability to distinguish beautiful work from merely competent work — in software, in design, in writing — is rare, difficult to teach, and disproportionately valuable.
Protect long blocks of unbroken time. Ambitious projects die from interruption, not from neglect. The maker's schedule is not a preference; it is a survival requirement for difficult intellectual work.
Geography is an ambient operating system. The city you live in whispers a message about what matters. If that message doesn't match your ambitions, you will drift toward what the city values, regardless of what you intend.
Every filter is also a signal. The specificity of what you value — expressed in language, in structure, in the design of your processes — attracts people who share those values and repels people who don't. This is a feature, not a bug.
Work on your own projects. The things that matter most are the things no one asked you to do. Curiosity that would bore most people is the only reliable compass for work that will turn out to be important.
