E.g., he wrote the first version of G-mail in a day. Okay. We can't expect more Googles from such efforts.
Paul, YC, and Silicon Valley have been taken in from their successes, their Sand Hill Road echo chamber, and their own Kool-Aid.
Instead of such influences, we have some much better examples of doing big things from the all-time, unique, world-class, grand champion of really big successes from really big technology, the US DoD, with high irony, what got Silicon Valley going, along with Dean Terman. For runner up, we have Bell Labs with the transistor, solid state lasers, and optical fibers. These successes were not done in a day. And with high irony, they were not risky, that is, the batting average was really high, much higher than VCs in Silicon Valley.
Paul is making another big, fundamental mistake, really implicit in the subject and title of the interview, what can be learned from 200 investments: He does have some good lessons, but broadly what he explains is very much the wrong approach.
Why? Because for his $20 billion criterion, necessarily he is looking for really exceptional, rare events. There's no hope at good lessons for such successes, such rare events, except just to count on luck, when looking at 200 projects where all or nearly all were just mediocre.
Yes, these events are so rare and difficult to do that maybe most of them, that is, the Googles, Facebooks, etc., were more likely from luck, Moore's law, and the Internet than sufficiently good work with little or no good luck. He's just not looking at approaches that are good at detecting the desired rare events.
Instead, it really is possible to conceive of, plan, and execute really big, earth changing work, with quite high reliability. E.g., did I mention the US DoD and Bell Labs? Uh, there's no way to see from Paul's lessons from the 200 startups just how the US DoD and Bell Labs were so successful. E.g., those organizations didn't need 200 rolls of the dice to get the Manhattan Project, GPS, stealth and the F-117, the SR-71, Keyhole (the DoD's Hubble, aimed at the earth instead of space), the transistor, solid state lasers, or optical fibers. And now we can include RSA, WMAP, LIGO, LHC, etc. -- no 199 failures to get one big success. Indeed, in all that work, nearly no failures. Also include the Human Genome project and the work of Craig Venter -- worked great, right away.
But the conception, planning, and execution of such low risk, high payoff projects are not illustrated by Paul's, YC's, or Silicon Valley's investment or project evaluation approaches.
Uh, hint: What Paul is describing is available to middle school students. So, we can set aside all education and research past middle school?
SMTP, HotMail, Gmail -- those were simple projects, flashes in the pan from luck and the rapid growth of Moore's law and the Internet, and do not illustrate the approach to $20 billion for the future.
Uh, early in e-mail, the software I had put one icon on my desktop for each e-mail message sent or received. I could wait for minutes for a single e-mail message to appear on my desktop. Total brain dead bummer. So, I got out the SMTP RFCs, read a little, use the interpretive language Rexx that had the TCP/IP calls, to write my own e-mail. For reading, writing e-mail, I used just my still favorite text editor -- KEdit. For converting to/from base 64 for the MIME attachments, I wrote some code in Rexx. My e-mail had a file that permitted nicknames and groups. It worked great. It put all the e-mail I sent/received, just as it was sent/received, exactly, in a file -- appended to the file with a delimiter line after each e-mail message. When the file got over 10 MB, I started another one. I still have all the files, and finding messages in the files is easy. I wrote all of it in just one afternoon except did the base 64 stuff in another afternoon. My e-mail set up was good enough that I had a cute feature: In files of other work, I'd put just the date line from an e-mail message as a link to that message, and one keystroke in KEdit would cause KEdit to get and display that e-mail message. Sure, the date lines do not have to be distinct, but I never found a duplicate! I was not lucky enough to get rich from it, but I used it for about 10 years, until I got a copy of Outlook 2003. I still wish I knew the syntax of the Outlook PST files or whatever they are called so that I could implement links to those. So, really in some ways, my little one afternoon project remains better than Outlook 2003. Writing basic client side e-mail is dirt simple and no big accomplishment in any sense.
We don't want to be duped into believing that the last $100 million lottery winner has some really great ideas!
In comparison with the DoD, etc. work, the work of Paul, YC, SV, etc. is, really, brain-dead. For more, Paul's approaches are not what got us to 14 nm and won't get us to 10 or 7 nm.
Uh, NASA and, IIRC, JPL, launched some equipment into space and eventually sent back some astounding pictures of -- sit down for this one -- Pluto. Up close. First time. No 200 failures for the one success.
But, wait, there's more: With the current power and prices of computing and the Internet, there is a very special case of information technology. There we take in bits, manipulate them, and put out bits. The goal is to have the output bits be darned valuable, well over the $20 billion criterion.
Then, what is the key to evaluating such work? The team? Personal need? The attitude? The determination? Their desire to get the main work done in a day? Their grit? Their novelty? Their interview impression? Other imponderables and tea-leaf reading? Hand shake style? Eye contact? "Taking the measure" of the person? And may I have the envelope please? [Drum roll}. And the winner is ....
I'll keep that a secret (hint: None of the above.).