Optimize a Verbose Prompt

Verbose prompts are not just wasteful — they actively hurt performance. When a prompt contains filler phrases like "Please kindly", "As an AI language model", "I want you to", and "In order to provide you with the best possible answer", the model allocates attention to these meaningless tokens instead of the actual task instructions. This example shows a prompt with common filler patterns and demonstrates how removing them reduces token count while improving clarity and consistency of outputs. The optimization process removes five categories of filler: politeness padding ("please", "kindly", "I would appreciate"), meta-commentary about the AI ("As an AI", "As a language model"), verbose preambles ("In order to answer your question thoroughly"), redundant role assignments that restate the system prompt, and unnecessary hedging that weakens instructions ("try to", "if possible", "when you can"). Each removed phrase saves tokens and sharpens the instruction. A good rule for prompt quality: if a phrase would be equally true (or equally false) for any prompt, it is filler. "Be helpful" is true for every prompt ever written and can always be removed. "Respond in exactly three bullet points using present tense" is specific and constrains behavior — keep it.

As a helpful AI assistant, I would like you to please kindly summarize the following text for me. I want you to provide a thorough and comprehensive summary that covers all the main points. Please make sure to be as helpful as possible and provide me with the best possible summary you can. In order to do this effectively, try to identify the key themes and important details. If possible, please also mention any relevant conclusions.