Comment by zufallsheld

This only works if all you care about is the code, like the author mentioned. But if you want to understand what and why the llm did something, or you want it to explain it to you, you're again stuck with the output.