The Larry approach sounds more vision-led and its interesting how it looks when it fails. Its difficult to see beyond the survivorship bias and only see it as egotistical and ignorant failure.
The real mistake with Wave was that a group of incredibly smart engineers were basically allowed to go into a bubble and build whatever they wanted for an extended period of time. Successful innovation needs to be anchored by a real problem to solve. E.g. the best programming frameworks always come from someone who is building a real app on top, not from someone who set out to build a programming framework. The feedback loop is necessary to keep developers focused on the actual needs of users, rather than on whatever problem they find most interesting to solve.
I think Eric is referencing the poor usage numbers as a proxy for the fact that the Wave team was not building something that users really wanted.
(FWIW my own startup, Sandstorm.io, had a pretty poor record on this point as well.)
OTOH, when Larry took over, things became much more top-down, with Larry trying to be Steve Jobs and tell the company what they should be building next -- most notably when it came to Google+.
I left before Sundar took over so I don't know how he runs things, but it seems like Google these days is just sort of chugging along, building solid products but not doing anything revolutionary.