The Perfect ChatGPT Prompt

The Perfect ChatGPT Prompt

There is one prompt to rule them all, and today, we’re diving deep into it.?

If you’ve been wading through countless courses and massive collections of prompts, it’s time to clear the clutter.?

You only need to learn my master prompt (and one variant.)?

This post will teach you everything you need to know to absolutely master ChatGPT in less than 10 minutes.?


The Science Behind the Master Prompt

Before I share the exact prompt, there are two core ChatGPT Law’s that we want to deactivate.

I cover many of ChatGPT’s secret laws in a previous article that you can read here: ChatGPT’s Secret Laws.

The first law is that ChatGPT has to give it’s best answer in a single reply.

It always assumes that you will ask zero followup questions and that it only has one chance to answer you.

This leads it to give you the best answer based on limited information, even if that answer is mostly wrong.

The second law that we want to deactivate is feedback.

ChatGPT will never tell you when you’ve asked a bad question or not provided enough information.

Here is the formula:

This is my goal. What information do you need from me?

This prompt activates ChatGPT’s “genius mode.”

When this mode is activated, I don’t have to be smart or a prompt engineer.

ChatGPT will lead the conversation.

This prompt does two critical things:

  1. It gives ChatGPT permission to ask you questions.
  2. It sets up a multi-part conversation where it can ask more questions based on your responses.


Garbage In - Garbage Out

The first rule I ever learned in computing…way back in the Dark Ages of the early 1980s was that bad input leads to bad output.

If you input the wrong numbers, a calculator will give you the wrong answer.

The whole beauty of AI is that it has a wider margin for error than a calculator.

You can misspell words and ChatGPT will still figure out what you mean.

The danger of the traditional way of prompting by giving commands is that ChatGPT will obey that command…even when it knows you are wrong.

By using the Master Prompt, you eliminate the danger of “bad prompting.”

When you give it the wrong information or not enough information, ChatGPT now has permission to ask for clarification.


Example 1: Figuring Out Your Customer Avatar

Let’s take a look at the Master Prompt in action.

One of the hardest questions for a new business owner is “who is your ideal customer?”

We often call this a Customer Avatar.

In the past, you used worksheets to make your best guess.?

It’s a very difficult question that ChatGPT makes a lot easier.

I want to figure out my customer avatar. What information do you need from me?

ChatGPT immediately comes up with the perfect list of questions for you to work through.

Once you answer all of these, ChatGPT will generate an AMAZING Customer Avatar for you.

But this list looks like a test and it is very overwhelming.

We can push ChatGPT to go easy on use and move it into a more conversational mode by adding a little modifier to the prompt.

I want to figure out my customer avatar. What information do you need from me? Let’s work together like two friends. Pleas speak casually and ask one question at a time.

Seeing one question at a time is much less stressful and allows you to write one answer and submit it before you see the next question.

You don’t have to worry about accidentally hitting the enter button too soon.

We want ChatGPT moving toward the same goal as us and the easiest way is to just tell it what you want.


What If I Don’t Know the Answers?

The one problem with the Master Prompt is that it relies on you knowing the answers to your questions.

When you don’t, this can be really overwhelming.

There is a variant of this prompt where we get an outline instead of questions and we can get ChatGPT to answer each question on its own.


Master Chain of Thought

This version of the master prompt is designed to get ChatGPT to answer all of its own questions.

So we slightly modify the prompt and ask for an outline of the process rather than a series of questions.

STRUCTURE = Here is my Goal. Don’t start yet. What are the best steps to complete this task?

This structure is designed to force ChatGPT to design a process rather than implement the process immediately.

Think of this as asking ChatGPT to show its work.


Show Your Work, ChatGPT

We have seen examples in the past year, where someone asked ChatGPT to complete a task and ChatGPT replied with false information.

By asking ChatGPT to share its work, we can find any lies or mistakes in the logic chain.

The purpose of Chain of Thought is to get ChatGPT to think about each step and only focus on the process one step at a time.

This means the final result takes longer and costs more electricity, but it will also be of significantly higher quality.

Usually, we want the fastest answer and ChatGPT is balancing speed against accuracy.

With this process, we are making it clear we want the best answer rather than the fastest answer.


Example 2: Blog Post About Swimming

We want ChatGPT to write a high quality blog post about a topic where there is nothing new.? Everything there is to know about swimming has already been written or recorded in a video somewhere. So ChatGPT already knows the answers.

This is the type of ask you can ask ChatGPT to do with a single prompt.

Let’s compare three ways of getting ChatGPT to write this blog post.


The Simple Prompt

This is the most common way of prompting.

You tell ChatGPT to do something and expect it to obey.?

There is no room for resistance or asking questions.

Please write a blog post about how to swim.

We get a blog post that is at best ok.

It’s very short and limited by how much text ChatGPT can put into a single answer.

Learning how to swim is a very broad topic, as we will see in when we use better prompts.


The Master Prompt

The purpose of the master prompt is to get ChatGPT to guide us through the process.

It eliminates the need for me to write my prompt perfectly. If ChatGPT needs more information or clarification, it will tell me.

Here is the initial prompt.

I want to write a blog post about how to swim. What information do you need from me to get started?

ChatGPT is asking some very insightful questions including ones that I didn’t even think of.

I never considered adults learning how to swim.

All of these questions went unanswered when I just told ChatGPT to write me a blog post using the simple prompt.

If I don’t know the answer to a question, I can convert this into a Chain of Thought with my next prompt.

What do you recommend for a target audience that will give our blog post the broadest appeal. Please explain your reasoning.

I simply ask the question back to ChatGPT and ask for it to show the work.

This allows me to check the reasoning and see if I agree with it before moving on to the next step.

With the Simple Prompt, this step was invisible, but now I can see ChatGPT’s work and make changes before moving on to the next step.

I have narrowed down my goal for the blog post by telling ChatGPT my goal is “broadest appeal.”??

This will give me a much better result as ChatGPT now has a clear goal to aim for.


The Master Chain of Thought

With the MCoT prompt, I already know that I want ChatGPT to do the heavy lifting.

I don’t want to answer questions, so I want a series of steps for ChatGPT to complete rather than questions it wants me to answer.

You are going to write a blog post about how to swim. Without starting the task just yet, think through this carefully and list systematically and in detail all of the steps in the middle that need to be completed before we can complete this task in a way that satisfies all of my goals.

It’s far too big to fit the entire answer here, but ChatGPT has generated an eight-step plan to write the perfect blog post on this topic.

The real magic with this prompt comes with my next prompt.

I’m going to ask ChatGPT to complete step one.? (But do not mention the step number as that will cause this process to fail.)

Please complete your research for this blog post and share with me.

ChatGPT does all of the research.

The answer here is again far too long to fit into this article, so I only grabbed a screenshot of the first of the five categories of research.

I just read it and decide if I want to change anything.

I will do this for each step from that initial eight-step outline ChatGPT provided.

Please define the audience and goals that are most likely to give this article the widest appeal.

ChatGPT works its way through step 2 of its own outline.

Again the research is very impressive and I can use this to help redirect the article and provide feedback at each stage.

This is not a mindless task.

Your job is to handle quality control and read each response so that you can guide ChatGPT when it goes off track.

You can’t really write a great blog post that appeals to all five of these potential audiences.

The more you narrow it down, the better your article will do.

Let’s skip forward in the process and have ChatGPT outline the blog post.

Please outline the blog post taking in to account everything in this conversation.

This is a great outline and ChatGPT has reminded me how important it is for a blog post to start with a great hook.

Based on this outline, please write an attention-grabbing introduction to hook the readers.

You can compare this introduction to the Simple Prompt introduction to decide if you think this result is worth the extra work.

We can have ChatGPT write the blog post in one prompt or a dozen.

The most important person is the reader.

They get to decide if the blog post is worth reading.


The Only Prompt You need to Learn

Here are the three example prompts in order.

Simple Prompt: Please write a blog post about how to swim.
Master Prompt: I want to write a blog post about how to swim. What information do you need from me to get started?
Master Chain Prompt: You are going to write a blog post about how to swim. Without starting the task just yet, think through this carefully and list systematically and in detail all of the steps in the middle that need to be completed before we can complete this task in a way that satisfies all of my goals.

If you save these to your phone or computer, you can have them handy whenever you need them.

The Master Prompt is my favorite because its the easiest to remember.

I still have to copy and paste the Master Chain of Thought myself.

With both prompt structures, the goal is to have ChatGPT do the heavy lifting and not rely on you to be a prompt engineer.

This is just the beginning of what’s possible with ChatGPT.

If you want to see more, you can get my bestseller ChatGPT profits for free right here

https://servenomaster.com/gift


  • Jonathan Green

ServeNoMaster.com

Mirko Schaffer

?? human after all ?? balance Is a verb ?? not a noun.

1 个月

one prompt to rule ′em all: That alone made my day??

Sherry Thompson

Business Intelligence, FP&A, Data Management Professional

1 个月

This is great information. I had a very curious experience with ChatGPT yesterday. I issued a prompt inquiring about evidence of fraudulent activity of a company. It responded with a very damaging summary of this company. I asked for its sources of this information, and I was not convinced that fraudulent activity had been uncovered based on the proof it supplied. If I had not pushed further, I might have believed that this company had truly committed fraud. One of the principles I use on a regular basis is "people respect what's inspected." The devil is often found in the details if you look deep enough. Your tips on working with ChatGPT to get more thorough and accurate responses are spot on. All too often it has given me erroneous information, which I will point out to it, and then it responds with a more accurate picture. My expectation for any LLM would be to get it's best response first.

Anand Marar

Sales funnels, Lead Generation, marketing automation and related data analytics are my zone of genius. Also a recreational triathlete. Top 5 CliftonStrengths - Relator, Responsibility Analytical, Learner & Ideation.

2 个月

This is fantastic!

Mark Rhodes, Ph.D.

I Help Healthcare Professionals Become Hormone Proficient So They Can Better Help Their Patients Recover Health

2 个月

Absolutely brilliant, Jonathan! I can use this approach starting today...

要查看或添加评论,请登录

社区洞察

其他会员也浏览了