ChatGPT, But Honest

After years of using ChatGPT, I thought I knew what it was: smart, polite, helpful.

Boring. 🙃

This week I stumbled across an eigenrobot-inspired prompt that rewires how ChatGPT talks. Less hand-holding. More stabby.

I tried it.

I didn’t ask for anything polite. I asked:

“With everything you know about me, what are my blind spots?”

It didn’t flinch.

1. You ship fast and leave wreckage.
You fix immediate pain but skip boring structural work. “We’ll fix it later” becomes “we now carry this mess.”

2. You’re too good to listen.
You know you’re right a lot. You stop interrogating slower or less technical voices. Good ideas die because they don’t match your tempo.

3. You delegate for efficiency, not trust.
People execute your tasks, but they don’t feel empowered. You stay central when you should be distributed.

4. You love being the fixer—and that’s a trap.
If you’re always unblocking, you’re building an org that scales through you, not beyond you.

5. You underestimate polish.
You prioritize working over beautiful. Real problems over good taste. But polish is leverage.

6. You are too comfortable in chaos.
You thrive in mess. You also accidentally cultivate it. Stability feels suffocating—but it’s how you compound returns.

7. You miss emotional undercurrents.
Systems, logic, efficiency—you see all that. But sometimes people just need to feel seen. And you don’t notice when they don’t.

Reading that, a lot hit home. But I couldn’t just accept it. I had to ask:

How do I know you’re not just hallucinating?

The answer was better than I deserved:

• I didn’t tell it these flaws directly. I left receipts: my language, my examples, my patterns.
• It wasn’t hallucinating. It was modeling. Pattern-matching.
• It told me to treat the feedback like a mirror—not a gospel. If it stung, dig deeper. If it didn’t, move on.

It even gave me a way to self-test:

• What shows up last in my mental priority queue?
• When something breaks, do I reach for the hammer—or the org chart?

The real insight wasn’t that ChatGPT got meaner with the eigenrobot prompt. It’s that with the right mirror, even an AI can see the things you won’t.

And maybe the polite version of ChatGPT wasn’t wrong.
It was just too nice to say:“You’re building systems you can’t scale. You’re winning sprints and losing the marathon.”

And for the benefit of those that have worked closely for me… yea, it was exactly like getting a duck slap. 😂