Paul Graham Quotes

Interesting insight from Andy Lapsa of Stoke Space: If you're running a hard tech startup, you should be able to fabricate everything you need yourself. You may choose to farm something out to a supplier, but you shouldn't need to.

A startup's ascent often consists of alternating rises and flat spots. This puzzled me when I first saw it at Viaweb. But it turns out just to be what happens when you add a function representing "random ups and downs" to one that's consistently increasing.

"The best lack all conviction, while the worst
Are full of passionate intensity."

This is always true, but it's truer now than usual.

"Keep discouraging venture debt."

— email from a founder

It would be very hard to make yourself read a lot if you didn't like reading. Even if you made yourself "read" lots of books, you wouldn't do it with sufficient attention. So only people who enjoy reading will be able to write well.

There are now so many "accelerators" that I get spam from people making products for them. Very weird. I wrote all YC's original software, so every spam is either offering something I wrote, or something I didn't think was worth writing.

Addiction is where you go from something taking you from normal to good to it taking you from bad to normal.

If you want to ensure that something you're writing has a sufficiently neutral point of view, imagine that you're an alien from another planet writing a report to your supervisors. But an alien with a sense of humor.

Err on the side of verbs. If you have to choose between expressing something via the verb or in some other way, choose the verb.

when customers tell you why they won't buy your product, that you should listen to. They can occasionally be mistaken, but not as often, and there is an iterative process for converging on the truth.

Asking the right questions has always been a critical ingredient in great work, but AI is making it even more critical.

If you want to start a startup, don't learn "entrepreneurship." Learn how to build things. The hard part of startups is not "entrepreneurship" but product: to know what to build, and to be able to build it.

The sweet spot for present-day AI seems to be projects that were constrained by the rate at which humans could produce text. That's why it works so well for programming. Basically programmers produced valuable text. But there are lots of other projects with this quality.

A lot of having taste about something is just caring about it enough to be honest with yourself, so that you can get past "I like what I like" to "Is this actually good?"

Twitter mobs are conformism storms.