Paul Graham on Start-ups, Innovation, and Creativity
Aug 3 2009

Paul Graham, essayist, programmer and partner in the y-combinator talks with EconTalk host Russ Roberts about start-ups, innovation, and creativity. Graham draws on his experience as entrepreneur and investor to discuss the current state of the start-up world and how that world has changed due to improved technology that makes it easier to start a software company. Graham talks about his unusual venture firm, the y-combinator, and how he and his partners work with start-ups to get them ready for more advanced funding. Along the way, Graham discusses why hackers are like painters and how to survive high school.

Michael Munger on Permissionless Innovation
Michael Munger of Duke University talks with EconTalk host Russ Roberts about permissionless innovation. Munger argues that the ability to innovate without permission is the most important concept of political economy. Munger defends this claim and explores the metaphor of...
Jonah Lehrer on Creativity and Imagine
Jonah Lehrer, staff writer for The New Yorker and author of Imagine: How Creativity Works, talks with EconTalk host Russ Roberts about the science of creativity. They discuss focusing vs. ignoring as a way to solve problems, the potential for...
Explore audio transcript, further reading that will help you delve deeper into this week’s episode, and vigorous conversations in the form of our comments section below.


Bo Zimmerman
Aug 3 2009 at 2:43pm

Mr. Graham nailed it when he compared developers to painters, though he didn’t do justice to Dr Roberts request for elaboration. In my opinion, the similarity is not just in their essentialist approach to what they absolutely need to do their stuff, but more importantly, that for both the (good) developer and the (good) painter, the quality of the thing produced trumps any consideration of how one gets from here to there. For the (good) scientist and mathematician, the opposite is true — they are supposed to be neutral to the results, and focus on making sure the process is correct. Most of the worst developers I’ve met focus their energy on good process, and produce nothing, expensively.

My own University (Texas State) understood this: they offered a Bachelor of Arts in Computer Science.

Dustin Klang
Aug 3 2009 at 4:03pm

This was a very interesting talk; one critique for Mr. Graham. Sir, your eloquence would be dramatically improved if you were able to remove the rhetorical affirmation for comprehension and/or agreement ‘right?’ from your active vocabulary.

Mr. Graham, it also seemed as if you believed that there was a certain benefit to living in a certain manner (pizza boxes and impromptu office space). Am I correct? If so, do you believe that the benefit is primarily derived from the knowledge of current self sacrifice of station for the possibilities of later return? If not, why do you believe that so many start-ups seem to share this, shall we say, “less organized” quality? I’m very interested in your response.

Aug 3 2009 at 4:13pm

You’re right about hackers & painters, wrong about mathematicians & scientists.

The latter two are also striving for beautiful results (check out QED, or executable biology, if you don’t think beautiful results exist in math & science) – the difference is that they have to get there using a process anybody else could follow. Their “code” has to be perfectly reproducible by *everybody* else – or, they have to document *every single* stroke of the brush, and the precise chemical composition of the colors used for each one, to achieve the painting. (When I say everybody, I mean everybody with sufficient technical knowledge, but not necessarily any domain-specific knowledge).

In other words – even achieving something un-beautiful in math or science is difficult. It is also often worthwhile, because other people can build beautiful things upon that basis.

It is generally less worthwhile to create ugly things in hacking or painting – but it is not always the case. Check out the linux TTY code if you want to see some horrifically ugly hacking upon which other, more beautiful things have been built.

Mads Lindstrøm
Aug 3 2009 at 5:24pm

Bo Zimmerman I am curious to know, in the context of your comment:

What do you see as the process of a mathematician? (the proof maybe)

What do you see as the product of a mathematician? (the proofed statement perhaps)

What do you see as the product of a software developer? (the code, a compiled program, …)

Bo Zimmerman
Aug 3 2009 at 9:04pm


I think of a mathematician’s process as being one of rigor and discipline; applying rules from axioms to the symbols they manipulate carefully and consistently. One wrong move, and the answer is pointless. I have great respect for such minds, not being one myself. 🙂

The product of the mathematician is rigor and correctness of the process they followed on the inputs they’ve been given. The answer may be all their consumers care about, but those answers are only as good as the process followed, every time.

The product of a commercial developer is a bug-free, feature-complete program that is profitable. Such programs can and have been produced by processes ranging from self-taught loners on their home computers following reference manuals, on up to College educated engineers as part of a massive corporate software engine. Some write compact spaghetti code in Assembly, some follow rules regimented to eliminate style. Some are thrown into Waterfall factories, some mixed up in Scrums, others huddle around a bedroom desktop alone or in small groups. I’m sure you get the picture.

Brian Gordyko
Aug 4 2009 at 12:59am

A reality show for start ups already exists in Canada. It’s called Dragons (I think…) and it airs on the CBC. As an aside I think the show exists primarily as entertainment, so that we may all bear witness to some of the most idiotic ideas on the face of the planet get shot down by rich people. As a by-product they do end up funding some of the ideas but I do not believe they are attempting to fill some greater function to the economy at large.

I am an Engineer by training and have recently started some object orientated programming at work to help me in a project. Mr. Graham is right that programming at many levels is much more accessible to somebody who is not formally trained in it. I have learned there is convention and structure to most programming languages, it makes it easier for others to review and for you to review later (which is important when you forget what you did at the time and have to sort through it all again). It all sounds like a very rigid formula but when you set about to doing it there are differences in the approach. As an example whenever my co-worker would help me fix my broken code he would say that it was “messy.” At first I found this very insulting; however, as I have spent more time the particular code I am using I am starting to understand and develop a style of my own. This leads into why I think you are wrong that other disciplines are also like the painter.

I beleive that the same analogy to the painter also applies in Engineering which shares elements of math and science. There are dimensions in solving a problem that goes beyond a simple question and answer. If you give 10 engineers the same simple static’s problem. You will (hopefully) see 10 similar results within some reasonable rounding error. You might also expect to see all draw a free body diagram in all 10 answers (which is convention). Outside of that each may try to work through the problem a different way using accepted methods and the laws of physics. So with 10 engineers there may be 1 answer, and 10 distinct methods of coming to the 1 answer. I also guarantee that some are better then others even though they come to the same result. The best way I would describe a properly executed engineering calculation is that it reads like a story, but this may not be obvious to somebody outside the profession.

Aug 4 2009 at 12:42pm

I enjoyed your talk.

The BBC in the UK has a program called Dragon’s Den which is exactly the “reality show” you postulated. Participants vie for funding by a panel of four venture capitalists : the dragons. (Must be the same CBC program Brian refers in the previous post.)

Sometimes interesting, sometimes funny, often very painful to watch.

The BBC link is here :

I do not know if they do follow-up programs on the success and failure stories.

Aug 4 2009 at 1:16pm

Hi, Hans, Brian.

I have never seen or known about “Dragon’s Den”. How might it compare to PBS’s “Everyday Edisons” (, which I have watched and enjoyed in sporadic reruns? What are these programs’ success rates for the startups they advise? Is there a conflict of interest between the success such a program experiences versus the success its entrepreneurs experience?

Is there a substantive difference between physically manufactured goods, services that require sales space (e.g., hair-cutting), and internet services? I was intruigued by Graham’s drawing the analogy between simplifying the processes of manufacturing a car via a physical assembly line versus simplifying the production of funding startups. But I think Graham has found a way to translate that analogy mostly to internet-type startups, which don’t require more than garage space.

Specifically: is there a difference in how to pitch these different kinds of startups to prospective investors? Is there something that other investor-oriented firms might learn from Y Compositor that might translate to specializing in these other kinds of businesses?

Mads Lindstrøm
Aug 4 2009 at 1:20pm


I would think there is a lot more to a mathematicians process than “applying rules from axioms…”. To prove a mathematical statement, you must use the rules of math rigorously, but figuring out which rules to apply is not a rigorous process. Few people care how a mathematician came up with a prove, only if the prove is correct and that the proved statement is interesting. My point is that a lot of what a mathematician do, is not a well-defined (mechanical) process, but requires intelligence, experience and intuition.

In some way the mathematicians job is not so different from a software developer. Part of the software developers job is done rigorously – after all, a compiler/interpreter is very unforgiving about syntax errors. But coming up with a good algorithm or structuring your program in a good manner, requires intelligence, experience and intuition – just like for the mathematician.

I would suggest that development processes often fail, as they try to turn development into a mechanical process. Like working in a factory. It is not that it is undesirable to invent mechanical processes, but that if a process is mechanical it should be automated, not turned into a procedure carried out by humans. Also, software developers are good at automating stuff, so if some desirable process is mechanical in nature, it is likely that it has already been automated (by software developers).

Bo, you write:

“The product of a commercial developer is a bug-free, feature-complete program that is profitable.”

I am a software developer myself, and I have yet to see a profitable, bug-free and feature-complete program. I think this has a lot to do with the difference between math and software development. Mathematicians prove their “programs”, and thus the “programs” are bug-free. The software developers programs are so big, that it is not practical to prove them correct.

However, I would suggest that software developers can learn from mathematicians. We might not be able to prove our applications 100% correct, but we could prove part of them (e.g. type systems or other types of statical analysis). Mathematicians have been developing structures for literally 1000 of years – it would seem likely that they came up with something of interest to software developers.

Aug 7 2009 at 10:03am

Really, a very enjoyable podcast. Graham’s got both a refreshingly humble perspective, and he comes off as a fun, laid-back kind of guy.

I found his insight into the impact of the recession on startups to be quite interesting. Anyone know where one might find some public data on what he described?

Bob Calder
Aug 8 2009 at 8:34am

I thoroughly enjoy EconTalk and thought Russ’ choice of Graham was typically enjoyable.

Since everyone is interested in the entertainment/education value of a reality show, I would like to comment on it.

The “rich people shooting down” ideas would be essential to misunderstanding the process which is what, I beleive, Graham immplied.

There wasn’t enough time to talk about it but I think he tried to clarify it by talking about early deployment with adjustment to users’ expectations.

The point of this is that “rich people” meaning market experts I suppose, would be the arbiters of reality in the television model. But the fact is that when faced with a novel interactive process, users often pick it up and use it for things the creators never really intended.

Expecting people with market knowledge to somehow divine the outcome of what a group of people will do in the situaiton above is worthless on its face.

The essential message Graham has is that getting an idea to market should not be about guiding/filtering it and that the process of filtering can be deadly to creativity and ultimately, profit. I am excluding Graham’s own filtering process which I believe has to do with capability and character. Y Combinator provides the capital, and voila you’ve got the 3Cs necessary for performance. An obvious application of a basic economic principle.

Bob Calder
Aug 8 2009 at 8:53am

Graham’s next step should be moving Y Combinator to a location that is highly wired, has a low cost of living, and generous visa rules.

How about South Korea? I would do it in a heartbeat. The culture supports cooperative living and working and there’s plenty of ramen.

Aug 9 2009 at 12:15pm

I enjoyed the podcast. It is making me evaluate my use of “…,right?” in my vocabulary. However, I noticed a difference in the way it was annunciated by Professor Roberts and Graham. I wondered which one I sounded closer to, right?

I agree with Graham on many things, especially his thoughts on making stuff and how there’s much less science and more art than many people think. In my town there’s two pizza shops, separated by two blocks with generally the same level of investment, largely the same type of equipment and the same type of ingredients. One regularly has lines out the door, the other stays barely busy enough to stay open. The difference in these operations is primarily an art of how all of the “same stuff” is combined to produce a different experience for the customer.

I also see this in the business that currently pays my bills. We’re loaded with the “get it done”/scientific group, but we don’t have many of the “artistic” business people. In fact, those people are strongly repelled from our business and I believe that’s the primary reason why the business is struggling to maintain relevance. Until the leaders of our business realize that applying B-school 101 strategy via prestigious, high paid consultants and executing through our strong “get it done” team is the problem, we will continue to struggle.

Tim Bugge
Aug 15 2009 at 2:41pm

Great interview as usual. I was particularly struck by the final conversation about high school education, and the sentiments expressed. They were very similar to the critisisms expressed in a book I just finished by John Taylor Gatto titled “Weapons of Mass Instruction”. In it the author, a retired public school teacher, claims that the “corraling” and “prison”-like character of modern schooling (terms used in the interview) is in fact by design, and sites the public rhetoric of many progressive era public school advocates (both political and industrial [bootlegers and baptists?]) to demonstrate his thesis that “schooling” should be used primarily to cultivate in the larger share of the population a habit of uncritical deference to authority in order to make them more easily managed as both a workforce and a political class.

Aug 26 2009 at 10:08am

The whole idea that only teenagers are using facebook and similar networking sites could be challenged. See the NY Times article below.

Comments are closed.


About this week's guest:

About ideas and people mentioned in this podcast:Articles:

Web Pages:

Podcasts and Blogs:



Podcast Episode Highlights
0:36Intro. [Recording date: July 17, 2009.] What is the Y Combinator? A venture capital (VC) firm, but how did you get involved in it? Wouldn't use the term venture capital firm, which means you've raised money from university endowments, venture partners on boards. Do seed funding, no name for it yet. To copy us they have to mention us. What does it do? Trying to use the name incubator, but that already meant something different. A venture capital fund makes a small number of large bets; invest one or two million at a time, one or two deals per partner per year. Y Combinator invests twenty thousand in fifty or sixty ideas; four partners. Likely success rate--don't know yet. Will not be as high as a good VC fund--ought to be a lot of failures or being too careful. Don't know success rate yet, but happy if it's a third; means most will fail. How are they screened? Patterns, but not in ideas. Batch mode. Applying mass production techniques to venture funding. Didn't set out to do this. Usually one thing made with a lot of time; someone comes along and rationalizes it, making it possible to produce a lot. Usually with a lot of technology. Fund a whole bunch of startups at once, bring all to Mountain View; they all start up at once--literally. Like a boot camp. Horse race. Graduate program. Teach things: common materials, mass production part. Software used internally: get many hundreds of applications; two batches a year, might get a thousand for next batch. Can't look at printed forms or Power Point presentation. Application form; software reads through the applications. Those who do best are invited to meet in person in November. Talk to them for ten minutes each. Do not like or have time to be pitched. Ask questions, like court martial, four interviewers. How long? Started summer of 2005. First batch did anomalously well; included Loopt, application that shows you where your friends are on your cell phone, used by high school students. Like having an ant farm. Reddit;, more viewers than was doing a different startup that failed. TextPayMe got bought instantly.
9:19What happens afterwards? Tell everybody yes or no; tell them to show up in January or June. Thought of doing a reality show? Would affect measuring what happened. Don't know if Y Combinator will work; not making lots of money yet; don't want to mess with something that might keep it from not working. Accept 25-30; they come from all over the world. Last winter batch, 7 out of 16 start-ups were from overseas. Not regional; can't start the Y Combinator of St. Louis; competing for same pool of applicants. Founders at this stage are completely mobile, two guys with laptops eating pizzas. If a city was a little better than others for startups, would vacuum them up. Easier to start a startup now--it's cheaper. Four reasons. One is hardware. Moore's law has made computer hardware effectively free. Internet has made promotion free: used to have to buy advertising or PR firm; now word can spread more quickly. Programming languages have gotten more advanced, more abstract; don't have to do as much work to get a given program done. Used to be that to build a startup you had to have a team of developers, five or six people writing C++; doesn't work to have that many partners, so had to hire them. Now, you can do it with just the founders, if 2-3 or them are programmers. Don't need to raise money to hire people. That's the reason there are so many more startups, and the reason they are more mobile. Anyone can do it. Don't have to raise money to build a factory; don't have to be plausible, old, well-dressed to raise money. Doesn't cost any more than hanging out. Larger pool of people and more variation. More risky ideas. Always going to lose the bottom people anyway, so larger pool is a good thing. See it already: Facebook and Google. Both too risky when they first started up. VCs and angels: VCs invest larger amounts of other people's money; angels invest smaller amounts of their own money. Ron Conway, famous angel; invested in both Google and Facebook.
17:03Common guess: when you are investing other people's money you can be riskier. Other way around. VCs have to worry. If they invest in a startup that looks plausible but tanks, they can point to how plausible it looks; if they invest in a 19-year-old whose idea tanks, the investors will not be convinced. An angel can go with his gut. Often angels founded startups themselves. VCs rarely were startup founders; most are money managers. Google and Facebook were both angel investments; and ideas seemed wacko. Social networking didn't exist. Just like fantasy as a genre didn't exist before Tolkien's Lord of the Rings. Social networking has taken over thousands of people's lives but didn't exist three years ago. Tempting to say there won't be anything like that again; but three years from now, who knows? Probably being built now. Asked earlier if any patterns in ideas. Line in TV show, "The Saint"; Roger Moore, not just the bad Bond: "I'm never disappointed by anything." Never surprised by anything any more. Now funded 144 startups; seen so much stuff happen; surprised in both directions. Sit back and watch things happen. By definition, can't predict the next big thing. Cool about capitalism. French and German governments cooperating, spending billions to make their own Google. Google was a good idea in 1997. What it was, was a stupid idea proposed by outsiders. What are the odds that a government is going to end up working on a stupid idea proposed by outsiders? Worse than average. Hayekian idea of trial and error; progress is not centrally planned; many fail, market weeds them out. Standard procedure with web apps: Release something as soon as you possibly can, because the point of releasing is to start learning from your users. Then you watch what people do with it. Amar Bhide podcast. Nothing better than to have sophisticated users; dangerous to build a product for giant corporations. Want to build a product for clever, impoverished outsiders, who are leaders, harbingers, willing to switch, not driven by inertia.
23:52Bring folks in, work intensively for three months. Several gears. This is the incubatorish phase. Word used to mean that all the startups work on your premises, makes the startups feel like employees. Don't provide space. Successful startups all tell stories about working out of the garage, turn on washing machine when cold. Space is not the hard part. But they do have to be nearby. They come for dinner weekly, give talks about how things really work; off the record, founder-to-founder. Can't publish them; speakers would clam up. After 10 weeks, demo day--actually two days. Invite all the investors from Silicon Valley and the country; speakers present for five minutes each. That's the conversation that starts the next round of money. After three months, the dinners stop, but nothing else. Office hours, meetings still with startups from two or three years ago. Counseling and mentoring. Help with deal-making. What they're building and talking to investors. Hard part is to get an investor to want to even listen. Get equity in exchange, from 2-10%, average 6-7%. Reputation good; when people are applying can tell them to just pick any of their startups and ask.
29:21The times we live in. What is the impact of the macroeconomic recession; and California in particular, which has had problems? California's fiscal troubles have not had an effect so far. Taxes already hideously high; could have some effects. Economy at large: worried that investors would slap wallets shut; but people kept investing. For startups, the bad days are over, unless a second wave of disaster. When it was bad, valuations were lower, half what they would have been. If people invest a million dollars at a pre-money-valuation of two million, what that means is the post-money-valuation is two million plus the million they put in, so they own a third of the company. Investors were asking for more of the company. Scared? Just sensed that everyone was more desperate; supply and demand; fewer people investing, pushing things on their terms. Not everybody ran out of money. Some VC funds were structurally messed up, their limited partners (LPs) couldn't come up with funds; but there were firms, big angels, that kept investing. You'd think some people out of work who would turn to a startup; haven't seen a lot of that. What age? 25-26, a couple of years out of college.
33:05How cities might be able to capture startup market: urban economic development in the 1990s went through a bizarre phase where cities wanted to be more like Seattle. Good thing, but you can't get there from here. St. Louis thought it was a lot like Seattle: world class universities; Boeing vs. McDonald-Douglas; Microsoft vs. Monsanto. But more people want to live in Seattle than St. Louis, which is geographically flat and visually boring. As a place for 22 year-olds it's not as attractive. City funded a lot of incubators--public/private partnerships. Politicians like this kind of thing: T1 lines, labs. Bio-tech completely different world. Commercializing research; lead guy, patents, less risk, more money. You know who to back: can narrow it down to 20 people. Interesting geographical role that regions play in innovation. Now, a handful of places: money; people like them who they can hire; the weather is nice; fun to be around; often because of a university, which gets them there in the first place. Stanford for Google founders. It is possible to create a magnet but hard to do. Cities that try to do it are not critical about their own plans: revitalize the downtown, never stop and ask themselves if it's going to work; result of some political compromise; or send delegation to Seattle and try to imitate it, like the Google of France. Can't see Mount Ranier from St. Louis even on a very clear day.
39:21Essay: "Hackers and Painters." Similarities: studying programming in college, idea was that it was supposed to be like math. But the interesting parts, you can't make scientific, let alone mathematical. Cornell has department of poultry science, more science than computer science. What makes someone good at programming is not what makes a scientist good at science. More like what makes a painter or architect good. Taste, aesthetics, a certain knack. Had studied painting; a lot in common with hacking. You make stuff. People who make stuff have more in common with one another regardless of what vertical silo they're in. Horizontal brotherhood of people who make stuff. Programmers and painters next door neighbors. Did studying painting and painting itself make you a better hacker? Can look at a hex color and know what it's going to look like: html red, green, blue hex code for colors. Web startups succeed in part because of good design. Hacker model for cheapness is the artist's model for cheapness; artists live cheaply in a way that is not squalid. Poor but uneducated tend to live a caricature of rich people's live, plastic tawdry versions of the same material things. Artist won't try to live like that; live in industrial space, rent a loft, paint everything white, put cool bright-colored things on the wall. Don't try to do a tawdry, cheap imitation of Microsoft. La Boheme: tuberculosis at end. Society richer today. Baseline high now, many more opportunities to live super cheaply and yet still nice. Left-brain, right-brain podcasts. People think of hackers as left-brained and painters as right-brained; but obviously there is a right-brained, visionary--aha!--aspect of hacking. Creativity. In the world of painting, there are some people who are fabulously talented at drawing; at age 15 they can draw. Use this skill to produce anything, though, then they lose. Similar distinction in programming. Some people, you give them a spec to write a program and they will just produce it, no matter how hard, as long as you tell them precisely what to do. But they can't make up a product that someone would want. This is a mistake companies make. Programmers translate their ideas into code; no loop-back. Best programmers combine in one head both the ability to translate ideas into code and having the ideas. Cezanne could not draw, makes same drawing mistakes that everyone makes in introductory drawing classes. But what he was good at was deciding what to produce. Frustrated. In a room of paintings, it looks like there is a spotlight shining on his and the others have been coated with a light coating of mud.
48:23Related stereotype: technical people like bells and whistles, and so create products that are cool or fun but are not what the user wants. Hackers like gadgets; it's a tendency you have to fight. Hack: gratuitously cool by nature. Not too much of a problem with the startups because they have very little time. In an essay, wrote "Facebook killed TV." Literally meant that's what people are doing instead of watching TV. Not literally true, but as true as you can get in three words. People like to talk to one another. TV, like the Brady Bunch, a fake family; might have preferred to be talking to group of friends, but no mechanism. Telephone was one-on-one till parents got mad. Now you can talk to your friends all the time. Also something about gadgetry: TV versus computer, cell phone versus camera. What got people to use computers even though they are hard to use. Even 14-year-old girls; that's how they exchange pictures of their friends. People over the age of 50 don't get this at all--the Facebook and Twitter phenomena. Why do I care that you are feeding your dog right now? Maybe when you are 18, your life is minutiae. Teenagers waste their time hanging out with their friends. Used to be on cars. Tweeting, great distraction. Time sink. Feature, not a bug, for a certain demographic. Used to be TV; now snuck back in to my computer. Many people over 70 don't have a cell phone, many don't use email. But even many over 50 don't do Facebook or Twitter. Will there be 70-year-old Facebook users? Most people like to waste their time, rich and poor. The reason 70-year-olds don't do it is their peers don't do it. For today's kids, there will be a new thing they haven't grown up with. People will be younger and younger when they are left behind.
55:47Why does the United States have such a vibrant startup culture; and how does immigration affect the startup market? Both the same question. Immigrants start startups, disproportionately so. More than half started by immigrants. Single biggest problem that kills startups is visa problems--people who get kicked out of the country because of political problems back home. Commencement address--just a talk--written for a high school; but no permission granted to student-inviter. Surviving high school. Should treat high school as a day job. Functionally high school was a holding pen. Welcome to the real world: you have to spend a lot of time doing what you don't want to do. Best to use that time effectively. Compared to being a waiter, high school is a pretty good gig; sweet deal. Lots of free time. As long as you are in prison, you might as well learn to be a good chess player. High schools with good reputations academically and for football. Remembering little from high school. Remember college more; still unremembered stuff. Possibly nature of the brain, learning other things that were important. The Scarlet Letter in English class--such a boring book. Deep insights into the challenge of that experience.