Have you ever felt frustrated when you just can’t get the right prompt? Maybe you think that the output (or in technical terms, the completion) you get from ChatGPT isn’t living up to your expectations?

The truth is, getting a quality output from AI large language models like ChatGPT depends heavily on the quality of your prompts and can simply be augmented by thinking about what you’re writing before you write and crafting the best input. In other words, a little bit of prompt engineering.

Commonly, receiving a sub-standard output doesn’t stem from the AI system itself, but rather from the inadequacy and imprecision of the input (prompt) it receives.

Stop open-ended prompting now.

It's important to understand that as an AI language model, ChatGPT works best when it's given clear and well-structured prompts.

So, to avoid prompts that are too broad or open-ended, such as "generate a LinkedIn post about AI for my account" or "suggest some ideas to me to write about programming".

Try to provide specific and focused prompts that clearly outline what you're looking for.

Always remember, poor instructions lead to poor results.

So how exactly do you do that?

There are four main steps you can take to get the job done. Let's break them down together and make it easy!

#1. Understand your needs and requirements

To order something, you first need to know what you want.

Right?

So you need to know exactly what you want from the AI, why you want it, and how you want it delivered.

So, first of all, start standardising all types of outputs you require from ChatGPT.

Here are some examples. Think about what you need in the prompt:

• I am starting to be active on my Twitter account — so I would like both tweet ideas and Twitter thread structures.

• I am really active on Medium — so I would like to generate skeletons for articles.

So from this exercise, I need 4 different types of outputs.

• Tweet Ideas for Twitter

• Threads structures for Twitter

• Writing Ideas for Medium

• Article skeletons for Medium

#2. Treat AI (ChatGPT) like a digital intern or assistant

Just like you wouldn't expect an intern to excel at their job with only one brief explanation, it's important to understand that AI language models like ChatGPT require clear guidance in order to perform well.

Say for example you want to post a Twitter thread about using the Google Cloud Platform. Simply asking for a thread without providing any additional information is unlikely to produce the results you're looking for.

Instead, it's important to provide specific instructions, such as the main points you want to cover, the tone or voice you'd like to use, and any other relevant details that will help ChatGPT generate a relevant and engaging Twitter thread.

Always provide your AI model with a detailed checklist. Explain the purpose behind the task, and be open to clarify any questions ChatGPT might have.

So, steer clear of the below:

The prompt is way too vague. Be clearer. Think about the below questions.

• How many tweets do you want?

• What writing style?

• What subtopics should ChatGPT emphasize?

• What is my language tone — friendly, professional, etc?

When the prompts are too vague or open-ended, ChatGPT is forced to make numerous decisions on its own, which can result in outputs that are messy or irrelevant.

What’s super important is to always use AI tools to leverage your work — not to substitute your work.

#3. Create constraints and avoid assumptions

This is critical to the process. It's super important to provide clear and specific instructions within your prompts. This can include:

• Accurate contextualisation: Clearly explain the context and purpose of the output you want from ChatGPT.

• A specific topic with subtopics: Define the specific topic you want ChatGPT to focus on, and provide any relevant subtopics or points of emphasis that you'd like it to cover.

• A specific structure: Determine the structure of the output you want, such as the number of tweets or words required.

• A specific format: Determine the writing style and tone that you'd like ChatGPT to use, and any other specific formatting requirements.

• A specific list of things to avoid: Clearly state any topics or information that you do not want ChatGPT to include in its output.

Let’s create our own prompt to generate a Twitter thread.

1. Add accurate contextualisation

I want ChatGPT to generate a Twitter thread for me. However, I need to ask the question - what is a Twitter thread?

Firstly, I need to make sure ChatGPT understands what I mean by "Twitter thread”.

Then I explain to ChatGPT what this specific output consists of.

“Twitter threads are a series of tweets that outline and highlight the most important ideas of a longer text or a specific topic.”

2. Add a specific topic

I want ChatGPT to write a twitter thread about a specific topic. Now is the time to explain more about the topic.

In this case, I want this thread to write about Google Cloud Platform free tiers.

Explain the main topic to ChatGPT.

“The twitter thread will talk about Google Cloud Platform free tier services."

Outline exactly what I want it to mention and what to emphasize.

“Write a 500 word summary of the Google cloud platform environment, include all its services and its utility for data science. Emphasize the advantages of using the platform over other services. Mention and reference Google big query and cloud functions.”

3. Add a specific structure

Now let ChatGPT know what the structure of the output is. This part can be more generic or more detailed, depending on your needs. I usually include as much detail as possible.

Specify the whole structure I want to receive from ChatGPT

For example, the prompt could read like:

“Write a tweet with a concise message, letting people know what the thread is about. Write more than 30 words, use key hashtags and use engaging language to convince people to read the whole thread. Emphasize the utility of the thread for them.

Write a second tweet that makes a short intro and lets the reader understand why they are still reading the thread. It is important to keep the reader reading.

Write 4 or 5 tweets that outline and describe the most important parts of the article. These should summarise the main ideas of the topic I explained to you before.

Write a last tweet with some conclusions, letting people know why the thread is worth reading.

Write a final tweet inviting them to retweet your thread and follow you.”

5. A specific list of things to avoid

In this case, if there’s something that you don’t want ChatGPT to mention, let it know. In my case, I don’t want it to use complicated vocabulary.

Therefore, I’ll add;

“Avoid using complex vocabulary.”

#4. Iterate and refine your prompt until successful

Again, if the AI generates a poor output, it’s probably due to an issue with your input. Don’t be afraid to rework your prompts multiple times.

Remember, even though you’re using natural language to talk with a machine, you should think of it as writing code for the AI.

Prompt writing is an iterative game — you won’t get it right the first try. But like training an employee, the upfront investment is worth it. Because once you have a working, reliable prompt, you can use it forever.

So, clocking off. What’s our point?

Overall, we at Promptstacks think it's important to remember that the limitations of ChatGPT and other AI tools are often a result of the way we interact with them. Rather than expecting the AI to do all the work for us, we need to refine our approach and take a more proactive role in guiding the AI to execute the tasks we need.

With the right mindset and approach, ChatGPT and similar tools can be incredibly powerful resources for improving our productivity, creativity, and problem-solving abilities.

Author

Josep Ferrer - Edited by Johnny

Twitter

LinkedIn 

Upskill Tutorials

1. Build an AI Chat Bot using ChatGPT API - Click here

2. ChatGPT for resumes - click here

3. ChatGPT for researchers - click here

From our Partners

Fast and Easy Compliance.

Promptstacks recommends Secureframe as it lets your team focus on the good stuff.

Let Secureframe unblock opportunities and accelerate your sales cycle without the need to invest in new resources or the risk of overburdening your team.

From comprehensive compliance policy templates to over 150 integrations with your core technology services, the Secureframe platform significantly increases the speed with which organisations can confidently begin a SOC 2 audit, without increasing overhead costs or slowing your team down.

Schedule a quick demo with SecureFrame and let them show you how to keep compliance in check.

Bitesize News 🍔

PALANTIR

Palantir, a data analytics company, has recently announced the release of their AI platform for military use. The capabilities of this platform are extensive and frankly, a bit unnerving. The AI is able to analyze complex situations and provide actionable plans by accessing data from a variety of sources, including satellites and drones.

The potential implications of this technology are significant, particularly when it comes to matters of national security. Some have even drawn comparisons to the Metal Gear Solid video game series, which features advanced military technology capable of analyzing and responding to complex situations in real-time.

GRIMES

The Canadian musician, Grimes, has made history by being the first artist to offer splitting royalties with AI-generated songs. In addition to this, she has been working on developing a program that is capable of simulating her own voice.

This move by Grimes highlights a growing trend in the music industry, where musicians are taking greater control over their creative output and exploring new ways to monetise their work. With the rapid advancement of AI technology, it seems inevitable that musicians will begin to experiment with using AI-generated music as a means of "owning" their voices and maintaining some level of creative control over how their music is used.

It will be interesting to see how this trend develops in the coming years, and whether other artists follow in Grimes' footsteps by embracing AI technology in new and innovative ways.

Editors’ Product Pick 📺

LunaAI

Create better LinkedIn content 10x faster with AI.

Magical.so

Automate your meetings, scheduling, and notes with AI

SecureFrame (sponsored)

From comprehensive compliance policy templates to over 150 integrations with your core technology services, the SecureFrame platform significantly increases the speed with which organisations can confidently begin a compliance audit without increasing overhead costs or slowing your team down.

Make Me A Midjourney 🤛 

Leaked pictures from the ISS

Teach me how how to make these — Midjourney

Reddit Spotlight 🔦

The Book of Genesis (in the style of Donald Trump)