Towards a Theory of Prompts

 It does not take long when working with an AI to discover that the major limitation in ordinary interaction is the prompt. Here is my take on the proper composition of a prompt.

A prompt should contain exactly the information necessary to elicit exactly the response desired, no more and no less. Insufficient information leads to ambiguities, confusion. Too much information leads to confusion. Irrelevant information leads to confusion.

A successful prompt will be 'clean', of a piece philosophically and bereft of extraneous information. An unsuccessful prompt will be 'noisy', cluttered with nonessential information and scattered in its organization. 

This analysis has achieved excellent results in working with claude.ai and applies as well to feedback which should relate directly to the prompt and be 'clean' and not cluttered with extraneous information.


Do Well and Be Well.

Comments