Paul Graham and Y Combinator

Abstract

Paul Graham (born 1964) is a Lisp programmer, painter and essayist who sold his web-store startup Viaweb to Yahoo in 1998 and in 2005 founded Y Combinator, the startup accelerator that funded Reddit, Airbnb, Dropbox and Stripe and turned the seed investment into a mass-produced, twice-yearly event. His essays, published as plain HTML on his own website since 2001, became the founders’ handbook of the following two decades: ideas such as “do things that don’t scale,” the “maker’s schedule” and “founder mode” started there. He made Sam Altman YC’s president in 2014.

Lisp and Viaweb

Graham was born on 13 November 1964 in Weymouth, England, and grew up in the United States. He studied philosophy at Cornell (BA 1986), then computer science at Harvard (PhD 1990), and afterwards painting at the Rhode Island School of Design and the Accademia di Belle Arti in Florence. His books On Lisp (1993) and ANSI Common Lisp (1995) made him one of the best-known advocates of the language John McCarthy had designed in 1958.

In 1995 he founded Viaweb with his Harvard friend Robert Tappan Morris (who was listed under a pseudonym, “John McArtyem”) and Trevor Blackwell. Viaweb let small merchants build online stores in a web browser, at a time when most software still had to be installed. Yahoo bought it in 1998 for 455,000 shares, worth about $49 million, and renamed it Yahoo! Store.

Graham later explained Viaweb’s speed in “Beating the Averages” (April 2001): the company was written in Lisp, which let three programmers add features faster than competitors working in conventional languages. The essay introduced the Blub paradox: a programmer who knows only a middle-ranked language (“Blub”) sees what weaker languages lack, but looks at more powerful ones and sees only “weird” features he does not need.

The Essays

From 2001 Graham published essays on paulgraham.com, on startups, programming, wealth, taste and what may and may not be said in public. They are written in short sentences without images, and several became standard reading for startup founders. “A Plan for Spam” (August 2002) popularized Bayesian spam filtering (see The History of Spam). The collection Hackers & Painters (2004) argued that programmers are makers in the way painters are, not engineers in the way bridge-builders are.

Some of the ideas became Silicon Valley vocabulary:

“Maker’s Schedule, Manager’s Schedule” (2009): a programmer needs half-days without meetings; a manager’s day is cut into hours, and a single meeting can wreck a maker’s afternoon.
“Do Things That Don’t Scale” (2013): early startups should recruit users by hand, one at a time, rather than waiting for growth to happen.
“How to Do Great Work” (2023): a long essay on choosing a field, getting to its frontier and noticing gaps.
“Founder Mode” (2024): written after a talk by Airbnb’s Brian Chesky, it argued that the standard advice to “hire good people and give them room” fails founders, who should stay involved in details deep inside their companies. The term spread within days and was criticized as a license for micromanagement.

Y Combinator

In March 2005 Graham, his partner (later wife) Jessica Livingston, Morris and Blackwell launched Y Combinator, named after a construct in the lambda calculus. Its first program, the “Summer Founders Program,” gave students a few thousand dollars each, three months in Cambridge, Massachusetts, and weekly dinners with speakers, in exchange for a small share of the company. Eight teams were chosen from 237 applications. They included Reddit, founded by Steve Huffman and Alexis Ohanian; Aaron Swartz’s Infogami, which later merged with Reddit; Justin Kan’s calendar startup Kiko, whose founders went on to Justin.tv and Twitch; and Sam Altman’s Loopt.

YC changed how startups were funded. Before it, a seed round was a negotiated, one-off deal with an angel investor. YC funded companies in batches, twice a year, on standard terms, and ended each batch with a Demo Day at which the founders pitched to investors all at once. The batch model let one small firm fund hundreds of companies a year. By 2014 YC companies included Airbnb, Dropbox and Stripe, each valued at over $1 billion; in 2019 the combined valuation of YC’s top companies was about $155 billion. Graham launched Hacker News in February 2007 as YC’s discussion site; it became one of the most-read forums in software.

Handing Over

In February 2014 Graham stepped back from running YC and named Sam Altman, a 28-year-old founder from the first batch, as president. “YC needs to grow, and I’m not the best person to grow it,” he wrote. In 2016 Graham moved his family to England.

Altman left YC’s presidency in 2019 to run OpenAI. After the OpenAI board crisis of November 2023 several outlets reported that YC had pushed him out. Graham replied on 30 May 2024 that YC had not fired Altman; Livingston had told him that if he was going to work full-time on OpenAI, YC needed someone else to run it, and Altman had agreed. “We didn’t want him to leave, just to choose one or the other.”

Graham’s influence on the ideas of the industry is mapped in Silicon Valley Philosophies, where his essays are part of the “founder creed”: the belief that small teams of founders, given money and left alone, can solve problems that institutions cannot.

Dead End: Arc

Graham spent years designing a new dialect of Lisp, Arc, with Morris. He described the goal in “The Hundred-Year Language” (2003): a language whose core would still be worth using a century later. Arc’s first public version came out on 29 January 2008, after years of anticipation among Graham’s readers. It was criticized immediately: it supported only ASCII characters, and its web library built page layouts from HTML tables. Graham called the release a snapshot of unfinished work. Few programmers outside YC adopted it. Its one large application was Hacker News, which ran on Arc for years. The essays arguing that Lisp makes programmers more productive found a far larger audience than the language written to prove it.