Excellent idea. I have done this myself several times.
My dilemma: At what point do you realize that your code is so cool that you don't want to develop it on someone else's machine (and dime)? I often find myself saying, "This is WAY more than they're paying for and will be worthless to me since they will own it."
To edw519: This where I'm going (in addition to looking for another line of work I might enjoy more). I've been making my own tools for about three years. I guess that's my "breaking point". I'm using them personally, but I don't want to be at shops without some kickass tools and practices already in place. I like my tools, but working at a place without any useful abstractions is no fun. In the meantime, I'm trying to make the best use of my tools and talents on my own dime, for my company (http://greaterscope.com). There I can solve the problems I want to solve, without restrictions (until I get customers). Hopefully it'll be a much better ride and I won't post another similar thing in another five years.
I think you can work on really cool stuff in any platform.
Similarly, you can also work on mind-numbingly boring stuff in any platform.
Perhaps he should just try to find something cool to work on (something that really turns him on), without regard to technology, and then just learn that technology in order to work on in.
It's not about which hammer or screwdriver you use.
It's about satisfying those who will live in the thing you build.