This is particularly timely given the recent success of the Mars rover landing, which is a good reminder that having a ton of meetings and spec-ing something out robustly in advance is still a valid approach in some contexts.
Generally software just doesn't work that way. (But maybe that was that you implied with "some contexts". In that case, I agree with you)
* I'm not saying they weren't trying out things until they decided how to land the Rover, but when they wrote the software they had very clear requirements.