ChatGPT has a brand new feature and it's amazing

How to use Custom Instructions to get better results from your prompts

The Prompt Warrior Cover

Read time: 5 minutes | Subscribe to this newsletter

Hey Prompt Warriors, Moritz here. Thanks for joining me on another edition!

The big news from yesterday: ChatGPT has shipped a brand new feature called Custom Instructions!

I’ve tweeted about this feature when it was leaked, but now it’s finally here.

And I am excited!

It’s a first step into making ChatGPT a truly PERSONAL AI assistant and it opens up so many new use cases.

In today’s post I want to explore what you can do with it.

Let’s dive in.

Before we begin, a big thank you to our sponsors who keep this newsletter free to the reader.

This week’s newsletter is sponsored by Drafthorse AI.

With Drafthorse you can drive traffic to your website without any effort. It uses AI to auto-generate SEO optimized blog posts based on your keywords.

Simply enter a keyword and several long-form articles of high quality will be created by AI.

How does Custom Instructions work?

Custom Instructions basically lets you insert two custom prompts (of 1500 characters each) in your settings.

This is allows ChatGPT to remember anything about you, which can be accessed in every chat.

So why is this important?

1. Context is everything

LLMs fundamentally are reasoning engines. Without context they aren’t very useful.

The key of prompting is feeding the model the right context, to get results that solve your problem.

This new feature gives you a tool to feed in context with less friction.

2. Personalisation

The future of AI is personalisation. You will be looking back at ChatGPT in a few years from now and be surprised how generic and robotic it was.

The problem is, ChatGPT does not know anything about you. Every prompt you enter gives the "same" result.

Enriching it with personal information will give you results tailored to YOU.

3. Future Integrations

This feature is only the first tiny step.

We are moving into a future where LLMs will be fed more and more of your personal data to make them more useful to you.

So where will all this data come from? Other apps.

Imagine a future where you can integrate the data from your:

  • Emails

  • Messaging apps

  • Note taking apps

  • Social media content

Scary? Maybe. But this is where we are heading.

What are some use cases you can try out now?

1. A ChatGPT that writes like you

This is probably the number 1 thing people want ChatGPT to do: Get it to write like you.

With the new feature we can get it to consistently do that.

Since we have a 1500 character limit, I would do it this way:

  • Step 1: Gather some representative pieces of your writing

  • Step 2: Open a new chat and use the prompt below

  • Step 3: Paste the output into the 2nd custom instructions box

Analyze my writing below in the following categories:
- Syntax and Grammar
- Vocabulary and Diction
- Tone and Voice
- Imagery and Figurative Language
- Rhythm and Pace

Provide an analysis for each category in one sentence. Also provide one example from my writing that is representative of this analysis.

My emails/blog posts/journals:
[insert your writing]

2. Product descriptions

If have a product or service and are using ChatGPT mainly for work around this product (ideation, brainstorming, content), then it could be very useful to add context for this product into the custom instructions box.

You can use this product canvas as a template:

Now a prompt like: “Help me brainstorm 10 new features for my product” should give you much better results.

3. Turn ChatGPT into an agent

This use case by Nick Dobos on Twitter is pretty crazy. He managed to turn ChatGPT into an agent by using a particular set of instructions.

Check it out here:

That’s it for today!

If you enjoyed this, please help spread the love by forwarding this newsletter to a friend or colleague.

How valuable was this email to you?

Login or Subscribe to participate in polls.

I hope to see you in the next one!


or to participate.