Good GA, but for genetic programming, you need to come up with a program representation and then evolve something in that space that spits out a solution. V. simple in lisp, given the code/data are the same thing. You could try the same thing with python's string eval, but it'll be trickier to define the lexemes and what qualifies as a feasible program.

Here's Koza's material from his Stanford class: http://www.genetic-programming.com/coursemainpage.html