Maybe to some extent, but with the way LLMs work, code generation is about as ideal of a use case as you get. Especially given the available training data.