Wow! Quite an achievement over a wide range of CS fields. :-) The direction to go from here really depends on what you want to achieve for yourself. Do you wa

Wow! Quite an achievement over a wide range of CS fields. :-) The direction to go from here really depends on what you want to achieve for yourself. Do you want to maximize the breadth of your knowledge? Then how about an RTL implementation of a CPU that implements your spec? Going for static analysis or developing an IDE would also have that benefit. Do you want your tools to be useful beyond being a toy / learning experience? Then I would go for feature completeness in the C compiler. Also, a debugger would be needed. I quite like the minimalism of your approach. How about a single-page spec for a CPU with an MMU? :-) (Then you could actually tackle a real Linux port instead of being limited to uClinux. But I'd refrain from trying to port any Linux kernel right now because you'd have to extend your compiler and tool chain quite a bit; Linux is written in GCC, not C.) How about running your compiler as a JIT compiler?