Rahul Verma: My Petty Bits

A very quick and easy trick for prompt refinement is letting an LLM ask clarifying questions, thereby flipping the interaction mode, for the subsequent request-response. This is very helpful for beginners in prompting as well as to experienced ones dealing with a new problems that they don’t fully understand. It’s also helpful when you are just feeling lazy and don’t feel like typing much :-).

Give the following as a root prompt (the first prompt in a chat) or you can even experiment by providing a variant of this in custom instructions that impact all chats thereafter.

“Hello there! From now on act as my brainstorming partner. To begin with solving a problem, I’ll give you a starting prompt, which may not be well defined as I have not worked out the dimensions of the problem yet. Ask me 5-7 clarifying questions. Based on my answers, suggest a better and effective prompt that gives me a better response that effectively covers what I am looking for. Let’s start?”

Remember – don’t cram prompts – locate the patterns and building blocks. Understand what is being done here and then tweak it as per your needs.