I Choose You, GPT!

When I first started using GPT, it felt like wandering into the grasslands and encountering my first Rattata and Pidgeys.

Cute.
Manageable.
Something to catch and deal with, but nothing to write home about.

This first exposure to GPT was like fumbling with a new Pokédex — exciting but clumsy.
It was already way better and faster than Google, giving me specific, direct answers without needing to scroll past pages of SERP.

It was great. A convenient little helper that worked at my pace.
But deep down, I knew this was going to be something more.

I had barely scratched the surface of AI.

Stage 1: A Better Google

Most people go through a few stages in their use of GPT, starting with GPT as a search bar with better grammar.

Likewise, I’d use ChatGPT in place of Google.
I'd toss out simple questions and get simple answers.

"What's 'good morning' in Japanese?"
"What are some foods that dogs can eat?"
"Explain blockchain to me like I’m five."

It was like battling a starter Pokémon with nothing but Tackle and Growl, good enough to scrape by, but nothing that made you feel like a real trainer.

The tool was impressive but the way I was using it was pretty basic.

Stage 2: Evolution and Growth

One day, it started to make sense.
This GPT wasn’t just a trivia bot.
It could take more complex instructions, even act with nuance if I fed it the right energy.

I started giving additional prompts like:

• "Write in an active and persuasive tone."

• "Repurpose this into a LinkedIn post and a tweet thread."

• "Summarize this text and highlight the three key shifts you made."

It felt like unlocking my first real moveset.
Suddenly, I wasn’t just slinging Quick Attacks; I was strategizing elemental matchups, playing the right Pokémon for the right opponent.

At the same time, I started to learn about its limitations.
Just like how you can't Thundershock a Geodude (it's ineffective), you can't throw just any task at GPT and expect it to land.

Sometimes it refused. Sometimes it hallucinated. Oftentimes, the outputs were monotonous, boring, and so AI.

But that was part of the process of learning, part of the grind.
You can’t rely on Rare Candy, you’ve got to farm the XP, tweak your moveset, and put in the time. I spent hours refining prompts, tweaking commands, and hitting these dead ends that made me want to smash the GPT up.

Most people get stuck here.
I’ve been stuck more often than I can count, especially when I ask it to help with something out of my expertise e.g. coding.
But I’ve also come to realize that I can’t expect the AI to spit out Hyper Beams on command, especially if I’m not prompting it the right way.

Stage 3: Commanding Your Pokémon

The real evolution came when I stopped treating GPT like an instruction machine and started treating it like a battle partner.

I began talking to it like a seasoned trainer commanding mid-battle:

"Here's my goal. What information would help you generate a better output?"
"List three directions this could go, and explain the pros and cons of each."
"Why did you suggest that approach? Can you show me another angle?"

Suddenly, it wasn’t just answering.
It was thinking with me.

I still run into many dead ends, but slowly and surely, I’ve learnt to guide it by:

• Setting the parameters

• Layering context

• Catching the hallucinations

• Asking it more WHYs

Like evolving from a wild skirmish into a full gym battle with strategy, counters, synergy.
Like Pikachu learning Thunderbolt.
Like Charizard finally obeying Ash’s commands.

Stage 4: System Architect

Then I learned that I can layer GPTs.

And I started to move beyond prompts.
I don’t just want outputs anymore.
I want to design systems.

In a system, there are parts of your workflow that AI can handle, and there are parts that still need human oversight or override.
Some day, maybe AI will bridge even more gaps, take over even more humans.
But today, the challenge is knowing where AI fits in, and where we still need the human spark.

The GPTs are getting more powerful, and whatever system we design has to keep pace.
This means recognizing what the AI can do now, anticipating how they’ll evolve, and constantly refining how to integrate them into our process and systems

AI is progressing fast.
And we have to keep up, not just in prompting, but in how we think about collaboration, system design, and creative command.

“To catch them is my real test, to train them is my cause.”

I’m still learning to design better systems and just like everyone else navigating with AI.
And I believe the future is boundless.
We’re barely getting started.