FORTUNE — Paul Graham has touched a lot of startups.
Not in the sappy, sentimental way — though some founders may largely credit him for their success — but where influence is concerned. Because few in the Valley can claim they have interacted with, advised, and funded as many early-stage businesses as Graham has. Indeed, Y Combinator, the startup incubator he co-founded with wife Jessica Livingston, Trevor Blackwell, and Robert Tappan Morris, has seen 632 startups come and go since 2005, including Dropbox, Airbnb, and online payments platform Stripe.
Last week, Graham announced he was stepping down as Y Combinator President, handing the reins off to partner Sam Altman. In doing so, Graham will still offer advice part-time to YC startups but cease day-to-day administrative tasks like poring over applications and interviewing applicants. “I spend all the time on that crap,” Graham half-joked onstage at this year’s LAUNCH Festival in San Francisco.
Graham also offered some parting general startup advice, culled from nine years of running YC.
Get a co-founder …
Businesses are more likely to succeed with more than one founder, because each co-founder may bring different strengths. Graham used Apple (AAPL) as an easy example: Steve Jobs was a pro with sales while the tech-focused Steve Wozniak couldn’t have cared less. That’s why Graham usually tells YC applicants to get a co-founder. In fact, that’s exactly the advice he offered Dropbox CEO Drew Houston when he applied for YC in 2007. Houston supposedly did just that, recruiting MIT classmate Arash Ferdowsi in two weeks. Now, seven years later, Dropbox has raised $350 million in its latest round with a valuation of $10 billion, far less than the amount Graham recalls Houston would have been willing to sell Dropbox for in the early YC days. “The answer was definitely under $10 million — it might have been under a $1 million or $2 million,” he recalled.
But get a co-founder you actually like …
Basic? Sure. But you’d be surprised how many startup founders actually don’t gel. With YC’s three most successful startups — Airbnb, Dropbox, and Stripe — much of Graham and crew’s acceptance decision was based on the co-founders themselves and how well they got along. (Of course, a promising idea didn’t hurt.) And if they don’t? Well, that’s easy to tell quickly enough, even during a YC interview, which spans a pithy eight to 10 minutes. “One guy literally rolled his eyes when his co-founder was talking and pitching us — a surefire tell,” Graham said.
Focus on a specific subset of users …
Longtime startup investors usually have one startup that had a nifty idea but never quite took off. For Graham, it’s Etherpad. Released in November 2008 by Appjet, Graham accepted the startup into YCombinator because he “liked the idea so much.” Essentially, Etherpad was a Google Docs (GOOG) competitor: a feature-rich, web-based text editor that let multiple users collaborate, write, and edit documents in real-time. (Google acquired the software in late 2009.) Problem was, the product lacked focus. “They should have launched it as one thing and let people realize, ‘Hey this is much more than real-time typing,'” Graham offered.
Instead of a catch-all approach, Graham said the secret to growth is starting a “small, intense fire” — first focus on a subset of users and meet their needs. Again, Graham brought up Apple as an example: The company produced just 200 units of its first personal computer, the Apple I, but sold 175 of them within the first 10 months on sale, largely to computer hobbyists. Explained Graham: “You’ve got to know who those first users are, sit with them, have a party, and focus on them.”