I definitely have the mindset of your schoolfellow - I like to get input & output from early on, even if it bears little resemblance to the final product and I work very much by progressive revision/elaboration

I think it's Simon who's the odd one out on this point. It's been accepted wisdom for decades that incremental development generally works much better than 'big bang' style programming. Brooks has a quote in The Mythical Man-Month somewhere about people being able to 'grow' much bigger programs than they could 'build'.