I read the post S12 (post noidea) essay (http://www.paulgraham.com/startupideas.html) as a (if unacknowledged) reaction of someone trying to understand and elucidate why ideas are actually very important (going against the common refrain that 'execution' is all that matters) after having recently seen poor ideas get by a selective process.
Ever run into an n^2 algorithm?
But it was certainly an interesting experiment to see if there are people who accept that they have the skillset to build a company but not have the idea part figured out.
Although requiring them to come up with an idea could be a good ability test, even if it isn't what they end up building. Because a large number of companies end up doing something different that turns out to be successful.
ala "getting to plan b" (or the abused "pivoting") http://www.amazon.com/Getting-Plan-Breaking-Through-Business...