Created by Ideogram 1.0 using the prompt, "Blueprint on table"

Prompt Engineering Guide

You might want me to give you some fancy prompt template or try to sell you a book of amazing prompts, but that’s not really how large language models work. Prompt templates that I used to use with fancy keywords like this:
“[Tone Guide: (Goal), (Tone), (Recipiant), (Content), (Length). Let’s think step-by-step.]
Generate a (blog post, email, list, ect.) about (subject).”
are not the answer. You need to treat robots like they’re humans.
Language models are trained on all the data on the internet, and they have figured out by now that writing doesn’t work by giving one long list of instructions and getting a polished article in return.
The way you write with ChatGPT is like having an employee. Give instructions, have a clarifying conversation, and go through a few iterations, and then you get an output you can really use.
Does this mean you can’t speed the process up? No. One way to speed your process up is to give the GPT more information upfront. I have created a GPT that already knows how I want it to operate, and already has examples of my writing. I don’t have to spend 5 minutes priming it to write. I can have a good output in a minute or two, but if you are trying to skip conversation all together that defeats the purpose of a conversational assistant, and your results will suffer.
So, in conclusion, don’t buy prompt books full of one-shot instructions. Instead build streamlined workflows customized to you, and I can do that for your business today! Any questions? Contact me.