A Conversation with Paul Graham - Moderated by Geoff Ralston

Find someone you trust, and then find someone they trust. In other words, take recommendations from people. When you hire via the web, don't hire too fast.

Paul Graham picked Robert Morris as a cofounder because he was his co-conspirator. Robert was an exceptional programmer. He could program as fast as he could type. When they needed more programmers, they recruited the smartest person Robert knew.

Advise for startups are often counintuitive and not obvious. For example:

• Don't do things that scale at first. Don't try to do too much in the beginning. You should do things that don't scale, meaning in a sort of handmade artisanal way so you can learn a lot from it.
• Do things very manually for your early customers. If you think you know the solution and build lots of stuff for it, you might be wrong because you don't know yet. Once you know what you should be building, then focus on scaling.
• As long as you have a constant growth rate, it means exponential growth.

The hardest part of being a sole founder is morale. 

When you start out, there are a million reasons why what you're doing isn't going to work. When you're a sole founder, there's no one to keep you going when things are going badly and no one to cheer you up.

• Most of the time, startup companies fail because of poor execution by the founders. 
• Startups seldom fail because of competition. 
• Startups are more likely to succeed if they start with a good founding team that knows one another well, can work together and can stand stress together.
• Founders are not just smart people. More importantly, they have to be determined to succeed.
• You don't have to be that creative if you care enough about users. You can just follow what makes users happy.

The risk of launching early is not so great as the risk of launching late. 

• Launch as soon as you have a quantum of utility, which means as soon as there are ten people who are glad that you launched because now they can do something that they couldn't do.
• If you're not embarrassed with what you launched, you launched too late.

• Figure out not just what users think is wrong but what's actually wrong. What's missing in their life?
• Ask them "What would you like to do that you can't?" They'll tend to give you an answer that's like a subset. 
• Try to find out what they really mean. Ask them, "What if you could do such-and-such?"

If you don't pay attention to users, you will make up some idea in your head that you will call your vision, and then you will spend time thinking about your vision by yourself and build some elaborate thing without talking to users.

You are better off finding someone with a problem that they will pay you to fix and then see if you can find more people like that.