I saw something like this a while ago... Codalab (http://codalab.org/). It's run by Percy Liang (https://cs.stanford.edu/~pliang/). It lets you create executable worksheets to go with your papers; and people can validate both the code and data used in your experiments.
Looks pretty cool indeed, bookmarked. However I believe it is focused on Computer Science. Nothing wrong with that but I think they key is convincing scientists that are not exposed to ideas like version control, open source (and I'd argue data sharing due to the rise of data mining) by default (or are less computer savvy in general).
