[aristus]

That cuts at the root of the disagreement. Categorically stating that diversity is "completely wrong" for a startup ignores plenty of counterexamples, so, yes, he's wrong. Diversity creates communications latency. It creates other things like tolerance and broader understanding. It also (ahem) widens your hiring pool.

And no, I wouldn't want a company that reflects the demographics of my city. That would mean no black people and mexicans do all the cooking. :)

Levchin is free to run his companies any way he wants, inside the law, and he's done amazingly well. But rejecting people who simply use an odd phrase once is, to me, not a example to celebrate, especially in a course for students.

"I did this and succeeded" is not equal to "everyone else is wrong". Creating a mirrortocracy is equal to claiming "I am right, all the time, and I do not require outside opinions. All I require are clones." That's a pretty heavy bet to make, and its arrogance is not excused by success.

[byrneseyeview]

It's hard to come up with a counterexample when you're looking at outliers. A successful startup is already statistically improbable, so I don't think you can easily say that if Paypal had hired more jocks instead of nerds, it would have done even better.

So we have to look at this from first principles. When you have a less diverse team, you don't just have short-term communications latency. You also have the advantage that really similar people can push each other more, because it's easier to tell if you're doing 1% better than someone else if you're both doing the same thing.

I don't know how valuable "tolerance" is. That depends on what you're tolerating. If diversity trains your ability to "tolerate," it sounds like a bad thing, the way having a broken air conditioner teaches you to tolerate heat. The belief that it creates broader understanding is also up for debate--in my experience, people express a wider range of thoughts in a more homogeneous environment, because they're less worried about offending people.

One last thing: "its arrogance is not excused by success." That's not very Bayesian. If someone has an unusual opinion, puts it into practice, and beats the competition, shouldn't you lower your estimate of their arrogance, and raise your estimate of your own arrogance?

[aristus]

A good counterexample would be Lotus, and possibly Flickr. I'm not claiming PayPal would have been more successful otherwise. I don't know. Many strategies can be successful. Levchin is the one teaching kids that everyone else is "completely wrong", in the face of proof to the contrary.

And there are problems with Levchin's strategy. In the next breath he lamented that it narrowed the company's hiring pool by 50%. That should give anyone pause. His strategy had a direct, measurable cost that could turn into a strategic disadvantage.

There are also serious effects down the road, as today's engineers become tomorrow's investors. If only Brand X people get funded, and hire Brand X people, who turn around and become investors themselves, isn't that a serious problem? You can't adjust your "priors" about other kinds of people if they never get a chance in the first place. Maybe Brand X is optimal, but you don't know, and claiming to know is arrogance.

Kapor has a good essay on this: http://mkapor.posterous.com/beyond-arrington-and-cnn-lets-lo...