I'm currently working on an old cad sort of program, its a mess, 20 years of kludges. Working on it is hard - you have to keep all this stuff in your head and I keep saying why the f did he do that, as expected really. Recently I had to add a new bit and use some computational geometry algorithms, it was like a step into a clear dark pool, everything was ordered and nice, which is probably the world of the phd.

I don't know how you test for the ability to do real world stuff - I always think of doctors, they have a system of proctoring they've developed over the years, where you're judged by your peers and rated accordingly and even then its not 100%. I think this is the only way to do it in real life, but I don't know if programming will ever get to that point, it probably will be necessary sometime - when everything is driven by computers.