I believe it can reduce tokens. It essentially obfuscates text as well if you want the security theatre of your prompts being hidden. Most importantly it makes

I believe it can reduce tokens. It essentially obfuscates text as well if you want the security theatre of your prompts being hidden. Most importantly it makes more complex chains where multiple questions are asked and uses the response from the llm to decide if you need a further question.