One of the things that I tell young people considering startups is that I actually think there's probably something a little different or off—you might describe [it] as wrong—with people who are very motivated to start companies. I don't think they're normal ... Would you sign up to get punched in the face every day? Does that sound like a perfectly logical [and] rational thing to do? Most people wouldn't.

A founder who truly understands what they're working on is able to explain it to a layman, so the ability of the founder to clearly communicate their ideas in verbal form is really important.

The beautiful thing about a founder is their ability to lie to themselves, and the number one thing a founder lies to themselves about is whether they have product-market fit, and the great founders limit the amount they lie to themselves.

The base thing that I want a founder to have is some special or different insight about the problem that they're trying to solve. I want to be clear: not necessarily a special or different insight about the solution. Solutions come and go; solutions change and iterate; solutions are something that really get molded by the user. But some special or different insight about the problem, I think, is really important.

People learn by doing. You don't do at school. It's that simple. You can learn on best practices are to fire someone or you can have someone in your company who's not performing and it hasn't been performing for six months and every time you look at them your stomach hurts because you know, you need to fire them but your stomach hurts more because you're too scared to. That's when you learn. That's learning.

When I describe a startup, I basically say, 'You signed up to get punched in the face every day—forever.' In the early days, who's going to be the partner [whose shoulder you're going to cry on], and they're going to cry on your shoulder, when you're both getting punched in the face every day.

You have to be a couple standard deviations away from average in order to have an extraordinary outcome.

The first motto of a YC application reader is 'don't be too smart.' I think there's all of this false conventional wisdom around being able to see the future ... These founders are shooting for a point ten years from now, and many, many things that look like a bad idea now will look like a good idea 10 years from now.

If I have two companies to fund, they're starting at the same position, they're going after the same market, and one believes software is commoditized [while] the other believes that they can build best-in-class software, I am always funding the second one—always, always, always. To me, when I think about the kind of founders I'm excited about, I think of founders who are excited to build best-in-class software.

Older and more experienced founders tend to be more comfortable with execution risk, meaning 'I know there's a need in the market, but I don't know whether I can fill it.' Younger and less experienced founders tend to be a little bit more interested in product risk—'I'm confident I can build it, but I'm not sure people want it.'

You are going to start learning skills because there's a gun to your head and if you don't have the skill, someone's going to pull the trigger, and then you learn. You do that enough times, you dodge enough bullets, [and] maybe you make a good company.

These are pretty impressive people. Their companies might not work, [but] you have to be be pretty impressive to even want to start a technology company, and then you have to be pretty impressive to be stack-ranked in the top 10% of a YC applicant pool. This is not just a straight cross section of people on the street ... Generally, everyone that comes in the interview has something to be confident and proud of.

I would say Silicon Valley by and large is built to fund software companies, and the further away you get from that software core, the less reliable it is as a funding mechanism.