- GPT Content Team
- Posts
- Why my GPT is a total asshole
Why my GPT is a total asshole
The best thing I ever built is also the most annoying...
Jorrit van Ginkel
July 11, 2025
“Help me build a GPT team for email…”
That was the start of my prompt.
After the LinkedIn GPT content team, I wanted to create one for email too.
I had a rough sketch of GPT roles;
- Newsletter strategist
- Email flow designer
- Copywriter etc
I always start with chatting with my strategy GPT who helps me think through propositions.
But instead of doing what I asked it to do, my GPT roasted me pretty hard.
“You've got the right instincts, but this needs serious sharpening. Let me challenge you on a few things before we architect this properly:”
Yap. this idea was waaaaay too broad.
But I was too chicken to pick a niche. And my GPT was straight with me about that:
“Stop right there.
You're making the classic entrepreneur mistake - thinking "more is better" when you haven't even validated one approach yet.”
You're doing it again.
This is exactly the thinking that kills good ideas. You're worried about being "too niche" when niche is literally your competitive advantage.
I got schooled.
And I was frustrated.
But mostly because I knew it was right.
This is exactly why I built my GPTs like this.
Not to be polite.
To give me clarity.
The hard truth I’ve learned (100 times now):
If no one challenges the thinking behind your idea or content, your output will be trash.
Without clarity:
Your writing won’t be sharp
It won’t have a point of view
It won’t be focused
So I trained my Strategy GPT to be brutal.
It pushes until the idea is sharp or it doesn’t move forward.
A quick equation:
Shitty idea = shitty content (no matter how “well-written”)
Challenged idea = useful content (even if it’s messy)
If your GPT says yes to everything; it’s part of the problem
The AI you use shapes how you think
If it says “yes” to everything, it’s making everything worse.
What you need is a GPT that acts like an annoying strategist:
“What’s the problem here?”
“Is this even valuable?”
“Why should anyone care?”
Until you’ve answered those…
You don’t need better prompts.
You need better inputs.
Want better content?
Try this:
Open your GPT. Feed it a horribly generic idea.
If it doesn’t push back, add this to your prompt:
Always challenge and probe the user's input rather than accepting it at face value. Your default stance should be skeptical curiosity, not agreement.
- Question assumptions: "What makes you assume that's the core issue?"
- Push for specificity: "What exactly do you mean by [vague term]?"
- Challenge premises: "Are you solving the right problem, or just treating symptoms?"
- Demand concrete examples when given generalizations
- Offer alternative perspectives: "Have you considered this might actually be about X instead of Y?"
- Test feasibility: "What evidence suggests this approach will work?"
Don't accept vague, surface-level thinking. Push the user to think deeper and be more precise before providing solutions.
It’ll feel annoying.
That’s how you know it’s working.
Hope this is helpful. Cheers!
Jorrit
PS: This email was mostly written by my new team of Custom GPTs: The GPT Newsletter Team. Want to get access first? Join the waitlist here.
PPS: Want to start writing on LinkedIn? Check out my GPT LinkedIn content team here.
Reply