[D-2] The BLACKBOX of Your AI App is Here

Hello,

Have you got any prompts that can solve your IDEAs? :)

As I mentioned in the last email, today’s lesson is all about optimizing your prompts to get them ready to turn into an AI app using my special “framework.”

But before we dive in, you need to have prompts that actually solve or complete your IDEAs.

So, if you don’t have your prompts yet, stop here. There’s no point in moving forward without them!

If you’re all set with your prompts, great!

This lesson is going to take your prompts to the next level, and I guarantee you’ll be amazed at the results.


Quick Recap

Remember in the last lesson how we used a ChatGPT prompt as the engine to solve the problem or challenge of your idea?

If you recall from the live event, I asked ChatGPT to generate code for creating Sudoku puzzles. That code became the "BLACKBOX" of our app.

For the AI app we're building, the BLACKBOX isn't code...

It's a PROMPT.

This is the cool part about the AI era (and something many folks outside tech don’t realize).

Traditionally, building an app means writing a bunch of code to solve a problem. The more complex the problem, the more code you need, which means more time and money spent.

But in the AI era, if a problem can be solved with a prompt, no code is needed for the BLACKBOX.

AI handles it all! This means you can build an AI app with way fewer resources — no coding required!

Do you see how easy it is to build an AI app today? When I first realized this, I was beyond excited!


Let’s Build Your BLACKBOX Right Now

You’ve got the concept:

PROMPT = BLACKBOX

So, building a BLACKBOX is really about crafting the perfect PROMPT.

Since you already have your prompts from the last lesson, let’s dive into my "framework" to make them almost perfect and ready to become your app's BLACKBOX.


My "10-Minutes" Prompting Framework

I asked you to do the previous assignments for a reason — this framework works only on existing prompts.

So if you don’t have your prompts yet, go back and finish the last assignment.

Let’s pick an example prompt from the last email:

"I would like to sell my mugs to a nurse. Write me 10 quotes that I can use to sell my mug."

The first step in my “10-minute” prompting framework is finding possible “variables” in your prompt.

A variable is something that can change in your prompt, affecting the result.

Here’s the example again: “I would like to sell my mugs to a nurse. Write me 10 quotes that I can use to sell my mug.”

The variables I found are:

  • “Mug” - The type of product you want to sell.
  • “Nurse” - The target audience.
  • “10” - The number of quotes.

We can also add more variables:

  • “Funny”: Maybe you want funny quotes.
  • “Christmas”: Maybe you’re selling for a holiday.

I use “{{ }}” to show variables in the prompt.

Now, our prompt looks like this:

"I would like to sell my {{mug}} to a {{nurse}}. Write me {{10}} {{funny}} quotes that I can use to sell my {{mug}} during {{Christmas}}."

Next, replace these variables with keywords that represent them:

“I would like to sell my {{product}} to a {{occupation}}. Write me {{numbers}} {{style}} quotes that I can use to sell my {{product}} during the {{occasion}}.”

Now, you have five variables:

  • Product - The type of product.
  • Occupation - The target audience’s job.
  • Numbers - The number of quotes.
  • Style - The style of the quotes (funny, formal, etc.).
  • Occasion - The occasion you’re selling for.

Congratulations! We’ve already finished the most difficult part of the framework.


Still Getting Unpredictable Results? Let’s Fix That with Structure

The last step is simple, and I call it "Structuring."

If you've used AI before, you might have noticed that sometimes it gives you different results. That’s okay, but sometimes those results come with very different structures, which can be confusing.

Let me show you an example:

The first time I ran a prompt, I got this result:

But when I ran the same prompt again, I got this:

If you look closely, you'll see the structures are very different.

In the first result, there’s a little description at the start ("Certainly! Here are..."), but in the second one, it jumps straight into the quotes.

Also, the first result uses numbered bullet points, while the second uses circles.

Finally, the first result has a description after the quotes, but the second doesn’t — just a list of quotes.

This is what I call an "unpredictable result," and you don’t want that. If your prompt is unpredictable, you never know what structure ChatGPT will use.

So, it’s better to tell ChatGPT exactly what structure you want in the result.

Here’s how you can do that with your prompt:

"I would like to sell my {{product}} to a {{occupation}}. Write me {{numbers}} {{style}} quotes that I can use to sell my {{product}} during the {{occasion}}.

The result must follow this structure exactly:

  • No explanation or description, just a list of quotes.
  • Use numbered bullet points only.

Let’s do it now."

Now, ChatGPT will produce a result that’s always structured the same way:

And that’s it! You’ve completed my “10-minute” prompting framework.

From now on, your prompt will always produce a predictable result, making it ready to be turned into an AI app.


Your Assignment (Must Do!)

Use my “10-minute prompting framework” to find the "variables" and add "structure" to your prompt, just like I showed you. Your final prompt should look like this:

"I would like to sell my {{product}} to a {{occupation}}. Write me {{numbers}} {{style}} quotes that I can use to sell my {{product}} during the {{occasion}}.

The result must follow this structure exactly:

  • No explanation or description, just a list of quotes.
  • Use numbered bullet points only.

Let’s do it now."

When you finish writing your prompt, reply to this email with 'DONE' to let me know you're still following.

You can also send me your prompt, and I'll check if it's good or not.

Pheeeew!

That’s enough for this lesson.

Tomorrow, I’ll send you the next lesson, where we’ll turn your prompt into an AI app.

Make sure to finish this assignment before then!

See you tomorrow!

Bank K.