Paul Graham Quotes

Recipe for low quality responses: Write a tweet that (a) threatens the identity of people who (b) aren't very smart.

One thing I didn't get about technical fields as a kid is that rigor is the standard ideas have to meet, not the source of them. The source is often the cheesiest sort of guessing.

The reason ideological enforcers want to make you recant is not that they care what you believe. It's to demonstrate their power, and to show which beliefs are forbidden.

That's why forced recantations are public. If they cared what you believed, a private one would suffice.

Prediction: Established companies will use cherry-picked examples violating broad definitions of AI safety as a way to suppress competition from startups. New technology is always somewhat unreliable, so startups will always be vulnerable to this type of attack.

Merely getting publicity for a cause doesn't automatically translate into generating support for it. If you get publicity for a cause in an obnoxious way, you'll generate opposition to it.

A math problem with a built-in dad joke: if a circular cake a foot in diameter is shared equally between 36 people, how big a slice does each one get?

It would be funny if as AI technology progressed, the legal situation tightened so that they were only trainable on images that were out of copyright, with the result that AI image generators produced very high quality traditional illustrations.

Competition can be an effective motivator, but don't let it choose the problem for you. Don't let yourself get drawn into chasing something merely because there's a lot of competition for it.

It's a good sign when you cautiously suggest some ridiculously ambitious future plan to a startup, and discover they've already thought a lot about it.

Some of the most successful founders start out tentative and lukewarm. Especially the young ones, who are often far from certain they want to run a startup. As they should be, because it's not for everyone, and they don't know yet what they like.

1. Don't design your site to force visitors to consume information in a particular order unless you're sure this is what all visitors want.

2. You can't be sure what visitors want.

As long as all newly founded startups prefer your product, don't worry about competitors who can convince larger customers. You'll own all future cohorts of companies, and that's more than enough. Plus in practice you'll eventually convert the old cohorts too, if you bother to.

"Make something people want" is fundamentally advice about being empirical, instead of assuming you already have all the answers. The equivalent for investors would be "Let founders teach you what people want."

One difficulty when trying to ban ideas on social networks is that it's the edge cases that matter most, but bans are enforced by low-level employees who don't understand them.

I meet a lot of founders who are worried by the rapid rate of technological change. They shouldn't be. It may feel uncomfortable, but techno-turbulence is net good for startups. They're much more likely to adapt successfully to some big change than incumbents are.