How To Make a Custom GPT in 30 Seconds
Link to live projects inside
Hello, Hoodies!
AI is getting extremely powerful.
In this simple walkthrough, I’m going to show you my favorite way to make custom AI agents that you can use for your studies, work, and life.
I made a custom GPT as a means to demonstrate for a previous article. (link to first gen python GPT shown below)
A simple python coding assistant. That legit took me 30 seconds to make.
In this article, we’re going to spend just a few more minutes breaking down making this new, more complete, Python Coding Assistant from scratch - called PySec. You can try it here. This the second gen one we’ll build together, right here, right now.
I’m going to breakdown a simple yet effective framework that I’ve used MANY times in my business (across multiple businesses, even) that you can use to help you with any narrowly defined facet of your life.
“OK Mr Cyber guy, why should I trust YOU for how to make AI stuff?”
Excellent point. Allow me to explain.
I spent tens of thousands of dollars on very expensive, hyper advanced AI.
In my side by side testing, I found that making SIMPLE GPTs with minimal instruction sets, allowing the core advancements in the GPT model to be 80% as effective at delivering desired results, and in some cases, exceed those of complex GPT with unique chains of thought/chains of reasonings.
So take it from the idiot that wasted a small fortune on advanced AI development, this is everything that most people need in most cases.
Prerequisites: ChatGPT Plus, and the ability to ready (congrats, if you’re reading this, you got that covered)
But first, what is a custom AI agent?
It’s a GPT that you make in an AI platform (in our case, ChatGPT) to do a specific thing so that we don’t have to give full context to a brand new AI model each time we want to do a task.
I’ll give you an example:
Say you regularly need help meal planning. But you have food allergies, macro requirements, and only like certain cuts of meat.
Instead of going to ChatGPT every week to get a new meal plan, you can put ALL of that information about you into a custom GPT one time and just open it up AND GO. It’s a huge time saver, and often, that reduction in friction can make your chat sessions even more effective because it reduces prompt fatigue (not sure if that’s a real thing, but if it’s not, I said it first).
In our case, I made a GPT to help with learning and analyzing python code for educational purposes.
If you didn’t read this and you just wanted to make a custom GPT, you may be tempted to just go into the GPT creator in ChatGPT and do something like this:
And you’d get a response like this:
With an instruction set like this:
Which is good. But it’s not great.
In my last year of testing and building custom AI apps, there’s one quick step that you can take that can double or even triple your efforts in the same amount of time.
Got to stock ChatGPT and build the instruction set there separately.
By doing this, the AI will go back and forth with you until you get all the features you want and it will be far more robust and complete than if you did it in the GPT builder.
It gave me over 20 follow up questions. After about 15, I had more than enough scope.
Then I told it to go and away we went.
In fact, this was the entire instruction set:
Compare that to the few paragraphs we had before. And the difference in using these two is night and day.
So we copy and paste this into our bot and hit publish. Then we use. That’s it.
And because of this, it knows the specifics of what YOU want and what YOUR needs are for this specific project.
AI is a whole new world. Make sure you understand how to leverage it or get left behind in the job market.
Stay tuned for more. Lots more.
Cheers,
Evan Lutz (BowTiedCyber)











