Ideas From Paul Graham

10 Jul, 2020 | People Notes, Startups, Writing


Write as you speak. Write with simple words. Write things that surprise people.

School is full of bad tests; tests you can hack. Life is less like that, you have to actually do the work. So try not to hack tests, but actually see the valuable work to be done and do it.

Keep your identity small, as it will let you see both sides of an argument, making you smarter and more rational.

The way to generate startup ideas is to focus on noticing problems. Preferably problems that effect you, but also in general. Get good at rejecting the status quo. You can also generate startup ideas, but they will generally be weaker, so you need to be stricter at filtering them.

The way to create wealth is to start a startup. Why? Because you want to build two characteristics: measurement and leverage. Being small gives you measurement. Working on an important technical problem gives you leverage.

What to do before a startup? Understand they are counterintuitive. So you’ll have to calm your instincts. You’ll want to follow your curiosity. They may prove to be important, which is good for a startup.

Startup ideas are like a well. You want to find one that is narrow but deep and preferably with gold at the bottom.

Startups take pushing. You do this by doing things that don’t scale. Indeed, that is your advantage of being small - you can focus on giving your users exactly what they want and the best experience too.


Don’t ignore your dreams; don’t work too much; say what you think; cultivate friendships; be happy (Top of My Todo List).

A key part of success is determination and competence. A lesser known one if having a disinterested obsession with something. Caring about something so much you just do it. Importantly, that thing should be important. But it’s hard to know if an obsession is important. Some heuristics though: better to being something that creates, rather than just consumer. Better if its hard, especially hard to others relative to you. There are some difficult outcomes. If risk and reward are proportional, then perhaps we have to be willing to explore avenues that amount to nothing? Plus, perhaps there are people that took multiple bets, none of which panned out. It probably happens, they just don’t become famous. One thought might be to hedge your bets (but maybe you don’t go deep enough). If this interest is a big component, it has some interesting directions for which to explore or cultivate genius. We should relax a bit, be more willing to try stuff just for interest. Also for children, let them go deep on things. Encourage them to explore and go in on something, no matter how random (Bus Ticket Theory of Genius).

Keep you identity small. Politics and Religion have something in common - people attach them to their identity. This makes it hard for them to have an objective discussion about it. It turns into a war. So one outcome of this is to keep your identity small. Attaching more labels to your identity makes you dumb, as you’re less capable of seeing both sides objectively (Keep Your Identity Small).

Don’t hide your problems from VC’s when raising investment. If they don’t hear any problems, they’ll assume you’re lying. So better to be honest (Twitter, May 2019).

On Startups

Full Essay notes:

  • How To Get Startup Ideas
  • Do Things That Don’t Scale
  • How to Make Wealth
  • Before The Startup

If he started a company again, he’d try his best to keep it small. At every point he’d want people to be amazed at how much they did with so few people (Twitter, May 2019).

A startup that helps people make money is a great business. AirBnb does this. So does Shopify. Help people unlock resources they already have (Twitter, May 2019).

On Writing

What should essays be? Useful. They should be correct. Not generic. They should say something as strong (4) as possible without being false (1). Precision and correctness are like apposition forces in which you must balance. Additionally, it should say something important (2) that at least some didn’t know (3). Those are 4 metrics of usefulness. All 4 can be measured on a scale, like a spectrum to give a measure of usefulness. So, how to  ensure correctness? Well don’t publish stuff that is bad. It’s like only speaking when you have something valuable to say. Proofread everything and don’t publish the bad stuff, or rewrite it until it’s better. Occasionally might let through something clumsy, but not anything false. How to ensure importance? Use yourself as a proxy. If it’s important to you, that’s a good start. Importance has two sub metrics; how many people, and how much they care. How to measure novelty? Again, you can use the Morris check here. If it doesn’t feel new, don’t publish it. How to ensure strength of writing? As you write, you want to balance assertions with qualifications (like perhaps). If you are certain of an idea, say it baldly. Otherwise, qualifications are okay. Finally, aim to say things simply. Whilst it’s not needed to be useful, it is considerate of the reader. So useful essays = correct + novel + strength + important. Doesn’t make people happy. Because novel means you’ll likely contradict a belief. Simple means it’s rude, etc. So people will try to challenge you. One way is to predict their challenges and write these in the end notes. So how to get good? Practice. So which constraint of the four to relax? First, importance, specifically don’t care about how many people care. Start with a small topic and get going. Also, don’t feel the need to publish. Practice for yourself is also good.

Essays start with a question and beg exploration. They are not merely explanations of thoughts, but rather a search for truth. When you send it to a friend; two things to consider - where is it boring? Where is it unconvincing? It forces you to think better. Essays may start with a question, but they should finish with an answer. You should tell the reader something the didn’t know. You just don’t publish the ones where you didn’t find an answer (sort of like a failed experiment). In essays, it’s okay to meander as you are taking your reader on an intellectual exploration of the truth. Fundamentally, an essay is a cleaned-up trail of thought. You should write about what is surprising. You should only write about things you’ve thought a lot about. So a pre-requisite to good essay writing is thinking a lot about things and then fretting out the surprising. One way is to get better at observing. Ask questions beginning with Why - particularly when it seems like a mistake. Surprises get easier to see once you have more knowledge, as you have more things in your head that they can potentially contradict with (Essay). Like Mortimer & Adler, it’s sort of like a discussion between you and the writer.

Writing is very different to speaking. In writing, ideas matter. In speaking, delivery matters. So a good speaker may not actually share many good ideas. And vice versa. But a speaker can give emotion, or motivate (Writing & Speaking). (Note - this follows exactly his advice above, to share what you find surprising).

Being a good writer - Write the first draft quickly and have confidence to edit a lot. Cut all useless words. Ask your friends to help editing too. Use footnotes for digressions. Edit, write like you speak, use normal words, cut words (Writing).

On Education

What to do at college? Work hard, work on hard problems, work with smart people. That’s how you get good (He spoke about CS, so I have generalised). Math is also useful, as it provides models for broader life. You should try to learn a bit about everything though. In general, he recommends maths, hard sciences and history over social sciences and philosophy (which can be ideological and less time lasting). Liberal education isn’t really for jobs, although sometimes (like CS) you can learn a hard skill. Grad school is great, apart from dissertation, but you need good recommendations (College).

School and life are different. In school, you learn for tests. Tests are hackable, so its not the same as learning in general. You learn in specific ways to pass tests and get good grades.  Why are tests hackable? Because teachers haven’t spent time making them not so. It’s not their jobs, they are teachers. The real problem is grades. Grades are overloaded. They are no longer just feedback on how you are doing, they are measures of judgement in the workforce. Take college admissions as an example - there are whole industry around hacking the test (exam prep, admissions advice, etc). The problem is they develop a habit for hacking tests. So when it comes to starting a company, they ask how to raise investment? And look for tricks. No, build a good business. Well, what is a good business? They look for tricks. No, solve someones problem. And so on. Often founders take a long time to realise these basics, making things more complicated because they have been trained by education and life to hack bad tests. This is a bad habit. Part of why startups are attractive is because its less like hacking a bad test and more like just doing good work. So for the individual, learn to unlearn this. A decent proxy, bad tests have authorities. So as authorities are removed from a market (e.g. startups, publishers) people can focus on doing good work (The Lesson to Unlearn).

The more you feel like a noob, the more you are learning (so less likely to be a noob). So get comfortable with feeling like a noob (Noob).

© 2022,