Would you be interested in sharing your findings? I'm currently experimenting with LLM-generated rust and honestly think it works quite well, however I'm looking for ways to improve the "taste" of the agent.