I had that same though pop into my head. Right after claiming that Digg hired a lot of "B" and "C" coders, he says that he's handpicking everyone who is working on his new project. I don't see how that will prevent the same issue from occurring.
However, consider the reverse, is there anybody who is consistently good at picking good developers? Is there any interview process that consistently works? If so, I am not aware of it. Hence I don't consider the hiring process to be 'solved'.
I can see the reasons behind Joel Spolsky's "make sure they write code during the interview" mantra, but you look at some of the other stuff he wants and it's all pointless pointer twiddling. Then he went and broke all the rules anyway by working with Jeff Atwood who by his own admission wouldn't know a C pointer if it bit him in the face.
It makes me wonder, how do they do it in other professions? E.g. is there a process by which you can weed out the good Civil Engineers from the bad ones? Or good Lawyers (leaving aside the obvious and droll joke), or good Doctors? What about a people oriented profession like Priest? Surely they would be able to weed out the good from the bad.
But no, it turns out that for Doctors and Priests we're always getting bad ones in the news (Doctor's lying about their qualifications, practicising without a licence etc).
Okay then, the 'fluffy' professions haven't got this sussed, what about the technical ones?
Yes, this the process of licensing. In those professions you mentioned, all people who (legally) work in the field must undergo a process of training and subsequent certification. This includes both work experience and exam requirements over the course of several years prior to certification. Yes, there are people that cheat the system, but the licensing exists to weed them out and protect the consumer.
Software "engineers" undergo no such process or requirements. I can learn ruby in my basement in a year, call myself a software engineer, and get hired building any number of apps. Is my work critical enough that someone may die if I am not certified? Usually not, so the need for certification isn't as strong. Nor have we had the time to develop such continuously relevant certifications in a field that changes several times a year.
I recall a story about a flight instructor who said to a freshly licensed pilot something like, "Now you've got your license to go out and really learn how to fly".
The point is that true competence and most certainly brilliance requires individually driven effort and expertise that is unlikely ever to be measured by government or professional organization administered exams.
Put it another way: Do you feel that licensing assures us that all drivers on the road are "good"?
(Sorry if my argument sounds abrasive; the subject's a pet peeve of mine.)
And getting a pilot's license is a little bit different from being licensed to practice medicine. It also depends on whether you're referring to a private or commercial pilot. Private pilots only need 40 hours[1] of time in the air. And even once a commercial pilot is licensed, there are other regulations in place. And a driver's license is an even lower bar, of course.
So no, licensing (in any form) does not ensure that the licensee is "good" or "brilliant," but it does allow us to ensure a minimum bar that does not currently exist in the software industry today.
(And no worries, it's a pet peeve of mine as well. At least at it relates to the software industry.)
1. http://en.wikipedia.org/wiki/Pilot_certification_in_the_Unit...
The historical pattern in the US seems to be that professional licensing, introduced as protection from incompetency, inevitably evolves into a way of limiting competition in a field by restricting the number of practitioners. At least on architectural exams, to the best of my knowledge, the maximum number of those who "pass" each year is set in advance.
It's a slippery slope once you start regulating entry into a field. The regulating body comes under political pressure from anyone with a financial stake, especially practitioners, educators, and major consumers. I'd wager that soon after you need a license to call yourself a software engineer, you'll need an accredited degree and years of internships before you can sit for the licensing exam. It's played out in similar fashion in the other professions.
There are other ways of ensuring competency without the use of government or quasi-government institutions. In software development, incompetency generally wastes time and money, not lives. With stakes this low, perhaps something as simple as a standardized testing service that could be used by HR departments to screen applicants would be viable... (Please pardon my designer's ignorance if something like this already exists in the development world :)