Paul Graham Quotes

The way I can tell I didn't say a quote attributed to me is not by remembering everything I've ever said, but by knowing what I would say. When I read something and think "I'd never write that," I know I didn't.

In office hours with startups I often try to figure out a secret weapon they can use to protect themselves from competitors. Today to my delight I talked to one whose most natural secret weapon was to be genuinely benevolent.

One big difference between talking to AIs and talking to people on Twitter is that AIs usually admit when they're mistaken and correct the error, whereas people on Twitter will do anything to avoid this.

Idea: Adversarial rendering. Builders proposing a new project should have to release enough of their design files for third parties to create renderings of what the project will actually look like.

Sometimes people who discover a new kind of work know that they have. Other times they just do something so well that imitating them becomes a genre.

Some of the patterns in the numbers are intrinsic (e.g. the alternation of even and odd) and others are artifacts of our use of base 10 (e.g. multiplying by 10 is trivial), but both enable shortcuts in calculation.

When something is fundamentally bogus, it ends up being surrounded by a cloud of subsidiary bogus things. It's like the way you can prove any proposition if you start with contradictory premises.

I noticed an interesting equivalence that obtains between two apparently unrelated things. The percentage of live music in restaurants that's not loud enough = the percentage of awards ceremonies that are too short.

You don't know how big a fish you are till you try a big pond.

If terriers had long enough legs to reach more things, the breed would not have survived.

One heuristic for making startups AI-proof is to make things human intelligence can't compete with either. For example, marketplaces are proof against AI competitors for the same reason they're proof against human ones.

AI will be very good at attributing paintings. The clues you use in attribution are exactly the kind it's good at. And this is going to generate some interesting news, because misattributions in both directions are common.

Prediction: Because the quality of abstract art is (ahem) hard to judge, it will be a lot easier to revise the history of that period to adjust the representation of great artists to suit present and future fashions.

I've never met a more consistently Darwinian group of people than nature guides. Supposedly rapacious business types are pollyannas by comparison.

No investor has any right to push you to get acquired if you don't want to. The company is your project, not theirs.