... and make the stuff we've already got work right.
Err, maybe that would be innovation. I'm semi-serious here, but the previous posters have already made the serious points better than I could... the web browser's a good example. It's a fairly new sort of program, too new to be, like, old hat already or anything. But, Netscape and Microsoft started "innovating" it all to hell and gone, without ever making it do its basic job reliably, and every one they bring out now, they've put their time into adding all sorts of crap that either programmers or marketers (or, I dunno, maybe even users) thought sounded nifty-- so they're feature-rich bug-ridden crash-a-holic pieces of crap.
But, so, on the other hand, p2p stuff for example is HOT HOT HOT right now, brand new paradigm, blah blah blah- it may be overhyped, but there's a lot of smart people trying to make something interesting out of it. Isn't that innovating? I mean, I don't know, I'm asking.
And, on the third hand, what's an example of a brand-new genre of software that would be useful (or even just cool)?