I remember having a discussion yesterday about the feasibility of porting gCAS2 to the TI-OS in an odd fashion. Since it's all in C, it won't be as efficient and will be much greater in size, likely exceeding the 8k program execution limit. However, I had an idea to convert the 8Xp (or just the binary output) to 8Xk (an app). (This deserves its own topic, but I want to be semi-free before trying to code it
DrDnar is right in many aspects - I also know of the hacker (who I should mention is a pro at Z80 assembly and very knowledgeable of the TI-OS) who played around with ZoomMath. You'd be better off learning the math yourself! Besides, real STEM people in the field don't use CASes** all day...
If you really want to try making a CAS, you can study existing CASes and understand how they work. Not only will you be able to write a CAS, but you'll also develop the mathematical knowledge that the CAS attempts to use!
** OK, professionals do
use CASes - most notably, Mathematica. But they use it for much, MUCH more complex mathematical operations then solving equations, probably operations that are understood in theory but impossible for humans to perform by hand. That is a legitimate use of a CAS.