startups 27
- Starting a startup and returning to college are incompatible paths.
- If existing model companies slow down, starting a new model company becomes slightly more feasible.
Paul Graham
Co-founder of Y Combinator and Viaweb; essayist at paulgraham.com.
Paul Graham did not write this page.
We collected these quotes from things they published elsewhere, and every quote links to where it was said. They have no account here and have not endorsed this site. Quotes are word for word; the short line under each one is our own restatement, not their wording. Their own site. Is this you? Claim it or ask us to remove it. Or tell us what is wrong here.
What Paul Graham says most about
design 19
- Prewar watch movements were often designed to fit the rectangular cases that were fashionable then.
- Great software requires a fanatical devotion to beauty: the parts no one is supposed to see should be beautiful too, because unseen details become visible in the aggregate.
ambition 15
- The founders who look unambitious are not: they have been trained by obedient upbringings to hide their ambition rather than lacking it.
- A founder needs ambition rather than a sense of duty, because a startup is grueling enough that dutifulness will not carry anyone through it.
taste 11
- Even the smartest people seem to have to learn taste in problems from experience.
- A company name can have different appeal to corporate buyers versus their customers.
writing 8
- A style guide is only useful if it reflects current educated usage.
- Good transitions make an essay feel well-structured, and complaints about lack of structure often come from clumsy or abrupt transitions.
management 6
- In practice "hire good people and give them room to do their jobs" often means hiring professional fakers and letting them drive the company into the ground.
- There are two ways to run a company — founder mode and manager mode — and Silicon Valley has wrongly assumed that scaling a startup means switching to the second.
AI and human skill 5
- Many things only worked because there is a limit to the rate at which humans can operate.
- Watch repair seems a promising career that is fairly resistant to AI.
LLMs 5
- Skeptics who say AI can't conjecture, theorize, or explain are describing last spring's models rather than current ones.
- The place to be in AI is probably at either extreme: close to the technology making LLMs, or close to the customer using AI to serve them.
discipline 4
- The persistent and the obstinate are both hard to stop: the persistent like a boat whose engine cannot be throttled back, the obstinate like one whose rudder cannot be turned.
- Persistence is five qualities combined (energy, imagination, resilience, good judgement and focus on a goal) and only resembles obstinacy in that it keeps you from giving up.
investing 4
- Being formidable means getting what you want in any situation, and that is exactly why investors want it: a formidable founder's interests and theirs are the same.
- Unlike later-stage investors, YC has to be prepared to give advice to founders about practically everything.
venture capital 4
- Being formidable means getting what you want in any situation, and that is exactly why investors want it: a formidable founder's interests and theirs are the same.
- Unlike later-stage investors, YC has to be prepared to give advice to founders about practically everything.
America 3
- American labs leading is good because Chinese lab employees would be unable to warn the public about threats.
- Chinese labs will encounter the same threats as American labs within a year, in complete silence.
AGI 2
- AGI is not a finish line to be crossed but a smear with width, and we are already standing on it.
- Skeptics who say AI can't conjecture, theorize, or explain are describing last spring's models rather than current ones.
AI and jobs 2
- Watch repair seems a promising career that is fairly resistant to AI.
- Many things only worked because there is a limit to the rate at which humans can operate.
capitalism 2
- Superlinear returns are a feature of the world rather than an artifact of capitalism, and the same pattern runs through fame, power, knowledge and benefit to humanity.
- The way to win at capitalism is to make customers' lives better rather than to squeeze them, because creating new things returns 10x where squeezing returns 2x at best.
data centers 2
- Blocking data center construction in America would not slow the rate of AI progress, only the rate of AI progress in America.
- Stoke's reusable launch vehicles will be shipping Starcloud's data centers into orbit within a few years
humility 2
- Founders' reputation for overconfidence is sample bias; most good startups actually underrate themselves.
- Soon it will become clear which things only worked because of limits on human operating speed, as they break.
optimism 2
- As soon as founder mode is an established idea people will misuse it: founders excusing an inability to delegate, and managers deciding to act like founders.
- Skeptics who say AI can't conjecture, theorize, or explain are describing last spring's models rather than current ones.
simplicity 2
- Not streaming a talk can be surprisingly powerful
- Good design is simple, because being forced to be simple forces a maker to face the real problem and deliver substance instead of ornament.
space 2
- Launch costs are on a declining trajectory comparable to Moore's Law-driven chip cost declines in the 1990s.
- Stoke's reusable launch vehicles will be shipping Starcloud's data centers into orbit within a few years
war 2
- A war made round watches fashionable because all military watches were round.
- Superlinear returns are a feature of the world rather than an artifact of capitalism, and the same pattern runs through fame, power, knowledge and benefit to humanity.