Given my own experience working on compiler stuff with LLM, I'd say it's a very good decision.
LLMs jump at the first opportunity to use regex for EVERYTHING instead of doing proper lexing/parsing, for example. You need to repeatedly tell it not to use regex. In the end you might as well hand write your code, because you actually know how it works, unlike a clueless LLM.
LLMs jump at the first opportunity to use regex for EVERYTHING instead of doing proper lexing/parsing, for example. You need to repeatedly tell it not to use regex. In the end you might as well hand write your code, because you actually know how it works, unlike a clueless LLM.