That's what each person in the article did. They found a problem, learned what they needed to solve it, and solved it.

It's probably way less about specialization versus willing to depart from your existing specializations. If you think, "Oh, I'm a software engineer, I can't do electrical", then you can't solve some problems. Or you can say, "I'm a software engineer, but I need to solve an electrical problem. Guess I'll go learn some electrical engineering."

Over the last week, I've given myself a crash-course in Arch Linux, driver code, and how feature points (as in computer vision) work, because there's a problem I need to solve that needs me to know these things. Until I changed my attitude from "I'm a web dev! I can't do that!" to "It's a problem I'm going to solve", doing those things was hard. When I changed my outlook, doing those things was easy.

Do the work, get the output - work's still work, but it's not... "hard". Just involved and time-consuming.

That's what we have flight simulators for (the good kind, not the arcade ones).

> run a marathon

Lots of people do that :). Hell, I might even do it myself one day.

> design a nuclear reactor

Well, why not? Of course your design isn't going to cut it because of tons of engineering trade-offs involved, but I think one should be able to make a "complex enough" design that makes sense.

> prove Fermat's last theorem, climb the K2, dance the Swan Lake, discover new drugs, speak 7 languages

Oh come on. :).

I get your point about perspective, but I still think we're specializing a bit too much. Given that the boundaries of our occupations are pretty arbitrary, and the reality doesn't care about them, we're losing out on knowledge at the boundaries. T-shaped skills, or whatever is the buzzword du jour for it these days, seems like a good idea.

That, and I think Heinlein is arguing for people to have a little thicker horizontal bar in their "T". :).