Startups are on balance a good thing. Their founders and early employees can be much more productive than they'd be working for an established company. Y Combinator's goal is to cause there to be more startups, by helping founders to start them.
Y Combinator represents the union of two ideas that had not previously been combined: the application of mass production techniques to startup funding. Funding startups in batches is not only more efficient, but also better for founders.
YC's value is the number of startups we help times how much we help them. Make both factors surprisingly big, and the product will be surprising squared.
From the point YC funds a startup we should put the founders' interests first, before even our own. That may seem counterintuitive in a for-profit business, but in this business it works; it's more scalable (in much the same way telling the truth is), and empirically the benefits of benevolence are greater than the costs. And since the only way to be consistently benevolent is to actually be a good person, YC's employees must be.
We must remember that we're investors, not bosses. We can advise and persuade, but not command. Good founders don't need more than advice anyway. And since you can't know what it's like to start a startup without having done it, those who advise the founders should be mostly people who have.
YC is also a startup itself, and (what's more difficult) must remain one. Many of the startups we fund get their first injection of startup culture from YC, so it's critical that YC practice what it preaches. YC has to be fast, cheap, informal, and focused on essentials. If something seems like the sort of bullshit a big company would engage in, it's probably a mistake. As with benevolence, the first line of defense is hiring.
The most successful founders are motivated less by money than by a consuming interest in what they're building. YC showed that this principle extends to investing too. What drove us in starting YC was that it seemed a cool hack: that if we helped founders in the earliest stages, there could be a lot more successful startups. That hypothesis turned out to be correct, and it has a long way to run. Focus on helping founders, and everything else will follow.