The Curious Companion: Ep. 8 – Getting ChatGPT to Do What You Want

Curious Reader!

Welcome to this week’s ChatGPT Curious companion newsletter.

What you came for is below, and you can CLICK HERE to listen to the episode if you decide you prefer earbuds to eyeballs.

Happy reading!

In this episode, we’re talking about what it really takes to get ChatGPT to do what you want. From clarifying your own thoughts to understanding the basics of prompting, this is a how-to guide without the hoopla. Really though, it’s not that serious. Whether you’re asking it to write an email or act as your assistant, specificity and clarity are the name of the game.

Let’s Start with Some Props

I received some feedback from a good friend that the last episode of ChatGPT Curious (Ep. 7: What is AI?) felt more digestible and was easier to understand than previous episodes. If you, too, shared that sentiment, I want you to take a moment and clap for yourself.

While I will admit that generalizations made about AI are easier to conceptualize than deep dives into LLM parameters, I fully believe that her (and your) increased understanding of these topics is equally attributable to y’all getting smarter. You’re sticking with things, you’re pushing through, you’re exposing yourself to these terms…and you’re getting smarter.

Clap for yourself.

This isn’t basic bitch material. Some of it is actually pretty complex and may require a few passes and some re-reads/re-listens, and I’m grateful for the effort you’re putting forth.

Prompting = Communication

The technical term for giving ChatGPT commands is “prompting”. It’s the words you type in the box. That’s it.

People, especially in Reddit tech bro circles, have been doing the absolute most trying to crack the code on writing the “perfect prompt”. But here’s the thing: It’s not that serious.

Like…not even a little bit.

In my humble opinion, the entire prompt craze exploded because tech dudes suddenly realized that being specific, clear, and not assuming things leads to better outcomes. The tech bros discovered (a la Christopher Columbus), communication.

Welcome to what the rest of us already knew.

Prompting is communication. The clearer you are, the better things usually go.

You Don’t Need a Template

Models have improved significantly (one thing I guess we can thank the AI arms race for), which means the days of needing to be obsessed with perfecting your prompts are largely behind us.

That’s also why I’ve never made a “Top 10 Prompts for X” guide. I also just don’t teach that way.

Call me old school, but I’m not here to hand you a fish. I want to teach you how to think, and how to organize your thinking, so you can use ChatGPT in a way that’s actually aligned with your goals.

So, the secret to getting ChatGPT to do what you want? Know exactly what you want it to do, and then ask for that.

The Clarity Checklist

When you want ChatGPT to act as your assistant, the following 7 points are what I suggest you get clear on, and then consider adding to your prompt:

1. Your background, familiarity with the topic, or level of expertise
2. All relevant information about the audience
3. The main take-away and/or objective for what is being created
4. Supporting information and/or concepts you want included
5. Specific description of the desired output (including tone and style)
6. Specific description of how you want ChatGPT to interact with you
7. Upload any relevant or supporting documentation and reference it in the prompt
   • This can be with the prompt or as part of a project (episode 6)
   • Don’t need to upload a zillion files – pick great instead of picking a lot

My suggestion here is to write out all of the clarity items from above and then literally paste that into ChatGPT and see what you get in terms of output quality.

Bonus Strategies

You can also:

• Ask ChatGPT to interview you to help you clarify your ideas
  • Ex: “I’m creating X, can you please ask me questions one at a time to help me clarify what I want.”
• Use it to put words to a process you have that you can’t fully articulate yet
  • “Can you please interview me one question at a time to help me put words to a process I use but don’t fully understand yet?”
• Ask if you missed anything—after you’ve added your info
• And when in doubt? Ask ChatGPT for help. With anything. Just type it in. It’s literally that simple.
  • Ex: “I know I want something but I’m not sure exactly what. Help!”

Real Talk on Output Quality

Something super important to note is that I don’t ever ask ChatGPT to just make something from scratch. No matter what the task, it’s always working from material I’ve already written or uploaded.

If I’m asking ChatGPT to “create” something, it’s more like summarizing copy that I wrote and have pasted in, or it’s synthesizing something using a significant amount of resources that I created and uploaded.

That’s how I maintain my voice and style, and it’s what makes editing easier when I do need to tweak something.

If you want outputs that feel and sound like you, you need to give ChatGPT more you (or, MOAR you, for those of you from the Maestro fam) to work from.

Technical Nerd Break: Reverse Prompting

If you’re into the more “advanced” stuff, don’t forget about reverse prompting (covered in depth in Episode 6).

Reverse prompting is where you give ChatGPT the input and the final output and ask it to write the prompt that would have created that exact output.

But What If ChatGPT is Still Not Doing What You Want?

Here’s where we talk about limitations.

The 7 clarity items above are designed to make you less of the limitation. But, as dope as ChatGPT is, it still has its limitations. Like we discussed last episode, AI still isn’t what the movies promised us.

Technical abilities aside, if we circle back to our earlier discussion that prompting is merely communication, I do think that it can serve as a jumping off point for a discussion about what I think most people want ChatGPT to be: A mind reader.

We want software that understands us without us asking. Software that doesn’t just respond, it anticipates. We want maximum alignment with minimal input.

To borrow from the ever-wonderful James Olivia Chu Hillman, we want relational fuckery. Stay tuned for my thoughts on this and what I believe contributes to what is now being referred to as “AI psychosis”.

For now, let’s just remember the nerding out we’ve done in earlier episodes and recall that AI is software that guesses, and ChatGPT (which is a type of AI) is math, not magic.

How I Used ChatGPT Recently

Each episode I include a section where I briefly discuss how I used ChatGPT that day/week.

This week I’m not sharing a specific task, and instead highlighting that I’ve been switching between the different models based on the task at hand.

• For creative tasks (like writing this very Companion), I’ve been using GPT-4o (remember, the current default model is GPT-5-auto). The outputs feel more aligned, and mirror what was being generated prior to the GPT-5 rollout.
• For everyday tasks? GPT-5 is fine.
• But when we’re co-creating and voice matters, GPT-4o is what I’m going with.

This is not to say that you need to be messing around with all the different models and switching back and forth. Just sharing how I’ve been using ChatGPT recently!

That’s it for today’s episode. Always grateful for you.

Questions, comments, concerns, additions, subtractions, requests? Hit reply or head to the website (​chatgptcurious.com​) and use that contact form. I’d love to hear from you.

Catch you next Thursday.

Maestro out.

AI Disclaimer: In the spirit of transparency (if only we could get that from these tech companies), this email was generated with a very solid alley-oop from ChatGPT. I write super detailed outlines for every podcast episode (proof here), and then use ChatGPT to turn those into succinct, readable recaps that I lightly edit to produce these Curious Companions. Could I “write” it all by hand? Sure. Do I want to? Absolutely not. So instead, I let the robot do the work, so I can focus on the stuff that I actually enjoy doing and you get the content delivered to your digital doorstep, no AirPods required. High fives all around.

Did someone forward you this email?
​Click here to join the ChatGPT Curious newsletter family.​

Stay curious.​​