your innovation as a startup will not be reinventing wheels -- building and debugging frameworks that, fortunately for you, others have put [tens of? hundreds of?] thousands of man hours into, solving seriously unsexy problems like browser incompatibilities, security nits and writing init scripts. i can't see how it's faster to tackle all of these things _and_ your application code than just, um, building your application code leveraging one of the very solid foundations that already exist.
while your competitors are solving problems that matter to your users, you'll be running around feverishly putting out fires when someone types an é into your application, or starts getting funny with strings involving apostrophes, semicolons and "drop table", or script tags in comments.