And it’s amazing how much more rewarding I find writing Ruby, Rails, and JS than I did Java and Android. You couldn’t pay me enough to write Java at this point. Life is just too short not to use high-level languages. (Except – I do want to learn more C and related languages for Arduino-type devices.)
I’m now beginning to look for work. If you have a project or company in need of a bright, motivated full-service web dev – please get in touch.
Paul Graham’s recent essay “Startup = Growth” has evoked a strong reaction, on Hacker News and elsewhere. The comments are full of phrases like “an epic essay with tremendous depth” and “one of my favorite PG essays of all time.”
Even much of the criticism is combined with praise. Many commenters repelled by the essay’s vision of startups still expressed gratitude for its insight. For example:
Thank you Paul. You’ve actually freed me from a dream that I now realize will never make me happy…. You gave me back my life, my real one.
But the more I think about it, the more I find that the main line of criticism misses the mark — and meanwhile, there’s a more important criticism, of particular relevance to founders, which isn’t getting enough attention.
To summarize the heart of the essay:
PG offers two criteria with which we can evaluate newly-founded companies: (a) the size of their market, and (b) their ability to reach it. Most companies are constrained in at least one of these departments. A company that makes Tibetan language software for Hungarians, for example, can probably dominate their market – but it’s a tiny one. Conversely, a brick-and-mortar English school for Chinese speakers has a huge market, in an abstract sense – but it won’t reach much of it beyond its local area, no matter how excellent or innovative it is.
But an online English school for Chinese speakers has both a huge market and the potential to reach a large percentage of it – and thus, the greatest upside. Companies with this kind of massive growth potential due to their market size and potential reach, PG calls startups.
And this is where the critics get it wrong.
The top comment on HN argues the trend in the past decade has been towards greater founder independence, and with it the ability for founders to grow more slowly, if they wish, accepting VC funding on their own terms (if at all). Now is therefore a bad time to narrow the definition of startup to strictly “super-rapid-growth” startups.
But that’s not what the essay does. The essay privileges rapid-growth startups as best exemplifying the essence of startups, but goes no further. In PG’s own words:
“A profitable startup could if it wanted just grow on its own revenues. Growing slower might be slightly dangerous, but chances are it wouldn’t kill them.”
In other words, a startup doesn’t need to grow rapidly – it’s just that a successful startup can grow rapidly if it chooses, and arguably should (due to the rewards of so doing and the costs of not). But there’s nothing in the essay that says all startups must grow rapidly, or that they must take VC funding. Several passages, in fact, explicitly say the opposite.
Other highly-ratedcomments make the same mistake. This is unfortunate, because they’ve distracted attention from the more trenchantcriticism. Which involves not the upside case (growth) so much as the downside (failure).
The downside of starting a startup is that it’s difficult to succeed. All of the obvious ideas meeting these two criteria have been taken. So you have to do some combination of the following: find a golden opportunity everyone else has overlooked, pull off something no one else has figured out how to do, or enter a crowded market and outcompete everyone in it. All of which are hard to do. For that reason, the vast majority of startups fail.
Given that, why would any rational person want to start a startup, or invest in one? Because, PG argues, the expected value is still quite high. If we assume the reward for startup success is $100 million, and the chance of success a mere 1 percent, your expected value is still $1 million. If you’re an especially gifted hacker, your chance of success might be as high as 20 to 50 percent, and your expected value $20 to 50 million. 
And here we arrive at a real problem. With numbers like these, startups resemble gambling. (Which, as a modestly successful former professional gambler, is a topic I know a thing or two about. )
Before I flesh out this comparison, let me get the biggest contrast out of the way. Occasional bubbles aside, startups are generally positive-expectation, whereas gambling is generally negative-expectation. The set of all startups has created untold trillions of dollars of wealth, whereas the set of all gamblers has lost untold trillions.
So we can restate our claim more flatteringly as startups resemble advantage gambling. “Advantage gambling” means placing bets with a positive expected value, via card counting, for example, or a host of more advanced techniques.
Great – but we’re still miles from establishing that starting a startup is a good idea.
Why? Because of variance. In the long run, your actual return from an activity should closely resemble your expected value, but in the short term, anything can happen.
Consider the following bet. You may bet $10,000 on the roll of a pair of standard dice. If you roll a 12, you win $100 million. If you roll anything else, you lose. You may play as many times as you wish.
How good a bet you think that is depends entirely on how much money you can afford to lose. Sure, it’s massively positive-expectation, to the tune of $2.7 million per roll. But if your net worth is $100,000… well, there’s a good chance (75 percent, in fact) that you lose all of your first 10 rolls, and now you’re toast. A positive expectation does you no good if you go broke before reaping its benefits.
On the other hand, if you’re already loaded – suppose you’re a Republican presidential candidate with a net worth of $200 million – well, now this is a fantastic bet. Even though your expected value is exactly the same, $2.7 million per roll, you have a much better chance of realizing that value, because you can afford the possibility of losing $10,000 dozens of times en route to rolling a 12 for the big $100 mil payout.
And if your net worth is somewhere in the middle, other factors come into play, like your level of risk aversion, or more generally, your utility function for money.
Professional gamblers refer to this as bankroll management. It’s not enough to find a positive-expectation game; you must also scale your bets according to how much you can afford to lose. Otherwise, you’re just gambling. 
And in PG’s examples, the logic of startups is almost exactly parallel.
The startup bet
Consider the following proposition. You may move to Silicon Valley and start a startup. If your startup succeeds, you make $100 million. If it fails, you’ve gained nothing financially (though perhaps a lot in more qualitative terms). You may play as many times as you wish.
How good an opportunity you think that is depends entirely on how much of your time and energy you’re willing to expend while possibly seeing no financial reward beyond a founder’s salary. Sure, it’s massively positive-expectation, perhaps as high as $50 million if you’re extremely talented and hardworking. But if you’re 40… well, there’s a good chance that your first half-dozen startups all fail, and now you’re 60. A positive expectation does you no good if you retire before reaping its benefits.
On the other hand, if you’re young – suppose you’re Bill Gates, who was 19 when he started Microsoft – now this is a much better opportunity. Even though your expected value is exactly the same, you have a much better chance of eventual success, because you can better handle the possibility of being a serial failed entrepreneur for several years until you finally hit a home run.
And if you’re somewhere in the middle, the role of other factors increases. PG highlights several, like how smart and determined you are, and your utility function for money.
Hackers might describe this as… life management. It’s not enough for a startup to be a positive-expectation opportunity; you must also consider your age, as well as your goals, risk preferences, and an honest assessment of your own intelligence and determination.
So now we’re ready to make a qualified endorsement: starting a startup can indeed be a rational choice, especially if you’re young, as long as you consider these factors first.
Or in PG’s words, “Most people should not try to start startups.” Most people shouldn’t gamble either, even with an advantage. But if you’re the right kind of person, work smart, and work hard, you just might pull it off.
Conclusion: what’s the distribution?
All that is true, so far as it goes. But there are a couple issues with the picture we’ve drawn.
The first is that, even if you’re a talented 19-year-old hacker, the odds still have a harsh edge to them. Assuming a 20 percent chance of success per startup, there’s still an 11 percent chance that all of your first 10 startups fail. Think about that.
Which brings me to my final point: the numbers PG and I are using are simplified for the purpose of making a point, and perhaps make startups seem like a worse deal than they actually are. At least, so says Andrew Chen:
The whole 1% of $100M versus 10% of $10M calculation vastly oversimplifies the outcome of these companies as binary. This is totally wrong.
In my experience in silicon valley, people start with building something small/simple (but in a big market), get little drips of funding from investors as they show progress. If they fail at any point along the way, there’s value in what they’ve created, and they exit for whatever they get. The later you exit, typically the further along you get, and the bigger the exit. That’s why the diversity of outcomes in the valley are everything from zero to billions, and companies raise anywhere from zero to a dozen rounds of funding.
At any inflection point in the business, you have lots of options: you can sell, raise more money, raise more and cash out some shares, you can quit, you can make yourself chairman and have your cofoudner run it, you can do nothing and grow it organically, etc., etc.
Each one of the choices above are part of your arsenal of options at almost any point. The people who choose to raise tons of money, not cash out at all, and then who fail- well, they made a series of active decisions to do all of that. They’re big boys.
My point is, when you’re building a company you can make a lot of choices along the way, and it’s not just setting out for a suicide run of either 1% of $100M or 10% of $10M.
This point might not matter much to investors, who smooth out their variance by investing in a large number of companies. But it’s highly relevant to founders, particularly those who aren’t 19 and aim to be rational in making major life decisions.
So this is where I’d like to see more discussion: Is Andrew correct? What does the distribution curve of recent, real-world startup outcomes actually look like? And what are the best strategies for founders who want to keep a range of exit options open, while maintaining rapid growth as their primary goal? 
 Strictly speaking, an expected value calculation should include not just the rewards for possible successes, but the costs of possible failures; for a binary outcome, EV = p(success) * reward – p(failure) * cost of failure. But for startup founders working with VC money, evaluating that expression is difficult because the first term is denominated in dollars and the second mostly in time.
 For five months in 2010, I traveled across North America as a professional slot machine hustler, beating select games from the new generation of digital slot machines. Yes, I’m serious. I summarize the crazy story in this blog post, and recount it in several Facebooknotesfromthattimeperiod. For the curious and/or skeptical, this post explains the simple mathematical principle behind the vast majority of beatable slots.
 Not every professional gambler is mathematically savvy enough to fully appreciate this, unfortunately. One of my pro gambling partners in 2010 really wanted me to leave beatable slots to count cards with him, and I declined, because we would’ve lacked a sufficient bankroll to safely apply a betting scale high enough to support a good average hourly rate.
 Shameless plug if there are YC decisionmakers reading: this is a conversation I would love to have at Startup School next month.
An MIT graduate student I know wrote the following about the Iranian post-election protests, after I asked what he thought of the use of technology there. Money quote:
The crux, I think, is this: twitter et. al. provide more interesting and useful communication tools. But communication isn’t enough, you have to wield power, and power doesn’t happen on the Internet… Communication is still really important to enable action. But that communication doesn’t have to be new or fancy, and it may work better if it isn’t.
So what’s next in digital activism technology? There’s a great quote in this Time article: “The sky is falling, but here we are — millions of us — sitting around trying to invent new ways to talk to one another.” I think there’s something to that, and I think there’s something of a distraction and time sink that the Internet brings to efforts to enact meaningful social change. I think we might be better served learning about what to say to one another than what incremental improvements we can make to the medium. Learning how to influence people and change their minds, get them to be more aware of the plight of everyone else. My personal research goals are around finding out how to get technologists to listen more deeply to communities in need about what their problems are, rather than what seems cool or exciting or technically challenging to the technologist.
Gonna brush up my French for the next few weeks, and go to the Association of Québecois in New England pre-party for the Fête Nationale on Georges Island – I think the ferry from Boston is $14 round-trip, or free if you RSVP.
Our Transsexual friend Ebru has been stabbed to death at her own apartment on March 10.
We; who have been saying that Transsexual and Transvestite murders are political murders, are going to protest the gay, lesbian, bisexual, transsexual phobic system which does not define “Hate Crimes”, awards the Killers by reducing their prison sentences, does not provide any constitutional rights to Gay, Lesbian, Transsexual, bisexuals, and transvestite population, and makes these type of murders much easier.
To commemorate Ebru and many others like her, We will be having a press conference infront of Ebru‘s house on Purtelas street. We are asking all to join us and support us on our cause for a just world for all.
I’m taking a course in genetics and evolution this semester — a good month to start, no? — to fill in my biology background as I work towards a future involving neuroscience. So I am likely to post more about science here, and also less frequently.
Especially since I had an epiphany about “focus” last month. For most of my life I’ve sympathized with Sylvia Plath’s fig tree predicament, too engrossed in possibilities to fully realize any of them. But perhaps some process of maturation has culminated, because it’s become easier to prune away activities and possessions (tangible and electronic) that aren’t contributing to my goals.
(If you’re looking for inspiration along those lines, try the chorus of Saul Williams’ “Break” — NSFW)
No, Obama isn’t progressive enough. And yes, people need to keephimaccountable. And yes, people need to work for progressive change in ways other than elections.
But from primary season through today, I have been on his side – not as a Kool-Aid drinker (been there done that), but as a realist. People who think elections don’t matter need a reality check. And may get one this week, if children’s healthcare is passed and Guantánamo is ordered closed.
So, I’m not embarrassed to be happy today. Tomorrow, we’ll see.
A commenter on the latest Stanley Fish blog post “The Last Professor” tries to make lemons out of lemonade out of lemons:
I also think that those who love the [humanities] enough to engage in arcane arguments in journals will continue to do so, whether paid or not by universities. We’re therefore looking at a future full of independent, hobbyist scholars — not the worst of all possible worlds. A return to the world of the gentleman-scientist, a reversal of the halcyon days when all the funding went to theology and fledgling geologists roamed the hills on their breaks from work as masters of divinity and grammar-school teachers. The world moves on, and so can we.