The LLMs use compilers, so I am not seeing the challenge here. Compilers do a lot of useful and important things like type checking and agents will typically check their output by invoking a compiler, then correcting errors reported by the compiler.

If anything what agents will mean is a move toward compilers for languages with more expressive type systems that can do more static checks for semantic correctness e.g. F*, which is an extension of F# that introduces dependent types.