I am using MATLAB a lot for prototyping, for communicating ideas and for documenting stuff for later use. Whenever you make a Matlab plot and include as attachement to a collegue or store it for future use, questions such as "what if you had increased variable x", "I prefer log x-axes", "what does that variable really mean?" arise.
My collegue uses IPython notebook, allowing him to integrate script, plots, etc in on apparent "document" with the flexibility of editing parameters and re-running the calculations right in the document. In one way, this is perfect scientific reporting as you can (in principle) have one self-documenting paper that only relies on a (e.g. recorded) dataset, showing every operation carried out on that data in order to reach some number/plot/... leading to some conclusion. Unlike so many papers I read where you might have to spend weeks figuring out how the formulas left out important subtleties that affected the conclusions (and quite often the relevance of their approach to me).
This seems like what Matlab has "almost" done for years. With the "%%" sections of .m files, Matlab Mobile, export to latex, etc.
Anyone using something similar for MATLAB or any ideas for getting this kind of setup without learning/porting to Scipy? Is this a good time to pester Mathworks staff about making such a feature?