An LLM is an impressive, yet still imperfect and unpredictable translation machine. The code it outputs can only be as good as your prompt is precise, minus the often blatant mistakes it makes.