I'm fairly well-versed in interactive use of MATLAB and ipython/numpy/scipy. I also have an HP-50g next to my monitor, with the USB cable plugged in so I don't need to keep replacing the batteries every few months.
I use ipython if :
- the problem has a significant non-mathematical component (string processing or some other general-purpose programming language task), or
- I want good looking 3d graphs
I use MATLAB if:
- I need to plot a bunch of things (2d), or
- the problem is almost entirely linear algebra
I use the 50g if:
- the problem is short enough to quickly write on paper, or
- the programming only requires a handful or short routines, or
- neither of the above programs is running, and the problem won't take more than a minute or two to type in and run on the 50g
There's a strong inverse relationship between the power of a calculating tool and it's usability. For really short stuff with only one or two variables, the ergonomically-ideal 15c is much much quicker than the clumsy notation of numpy. The 50g, with it's math-optimized keyboard and GUI, is often much faster than either desktop option because the bottleneck is the typing process, not the computation process.