Is it, though? Grammar is more like a set of tools than a rigid mechanized system. Formal proofs have recursion and logical rules which generate vast structures, structures which, in our era of computation, can move and act without our full understanding of their function. But whenever we seek to describe how something functions, even in our most clear and rigid proofs we use natural language, and I'm sure many of those proofs contain "grammatical errors."