Personally, my ideal language would be a spaghetti-code type language that would treat each gene as say a TI-83 would treat a program. I guess what I mean is that I want simplicity and ease of use, coupled with a powerful yet easy to use math library. TI-BASIC was very easy for me to learn (nesting condtitions was very easy, GoTo's made branching easy, and the math functions of the calculator were simply indespensible) and allowed me to code out complicated ideas with a very simple and easy syntax. Once I created a system which would pit two A.I. players against eachother in dozens of rounds of rock paper scissors. The first one I made (based on statistically finding the most probable response to a given move) became pretty dang advanced, turning out to be a very crude neural net once I started adding some of it's secondary functions (a tiebreaking routine, an 'evasive' routine in case an opponent 'locked on' to it's output, and some other stuff related to data maintinence) but I sadly lost all of it to a RAM reset. I also find it to be easier to use than Python. If you look, Python's function system works almost just like calling a program as a subroutine in TI-BASIC, even allowing recursion. What I want is brain dead simple. Although I can live without goto's, which are impossible in DNA code.