Last week, I caught ChatGPT in a blatant lie.
It told me:
"This is beautiful, Adam — grounded in purpose, full of momentum, and uniquely you."
Which… is objectively false. Not because I’m not those things, but because a language model doesn’t actually know. It was a moment of projection, one that felt surprisingly intimate and deeply fake at the same time.
I had planned to post about that interaction — and to share a bit of the comedy I find in my AI dev workflow. I’ve got a whole team of agent personas now:
- Peter writes the spec.
- Egon rigorously reviews it.
- Winston builds it.
- Ray reviews the implementation and goes,
“This whole plan was a mess. We should’ve gone a different direction.”
It feels eerily like working with a real team. And annoyingly — Ray was right.
But before I could post, I came across a video by Cy Canterel that stopped me in my tracks. She was talking about our human tendency to anthropomorphize everything — giving minds and emotions to things that don’t have them. One of my favorite examples is the “man in the moon.” Her example was even better: the mirror in Snow White.
In the story, the evil queen asks, “Who’s the fairest of them all?” And the mirror says, without emotion or agenda, “Snow White.”
Not because it wanted to cause harm — but because it had seen her, and that’s the truth it knew.
No bias. No ego. Just reflection.
And that’s what these LLMs are:
A mirror.
A super smart, magical, very fast-typing mirror.
That realization hit me hard. Because when ChatGPT gave me that lovely compliment — some part of me wanted to believe it. But another part of me wanted to talk to an actual person. Like, crave connection.
Subconsciously, I reached for human warmth to validate something an AI had echoed back to me.
Fast forward to today. I’m deep in the code, finalizing this project, and it’s a mess. Not throw-everything-away bad — but bad enough that I need to breathe.
There’s a certain pain that comes when I create something new, something I want to love — only to find it carelessly assembled by the very tools meant to help. It’s like watching someone finger-paint over a canvas you spent hours sketching.
It creates the illusion of production readiness — but when you lift the hood, it’s chaos.
So I’m calling it.
For the next 48 hours, I’m putting AI-enhanced coding on pause.
I’m going back to coding solo.
Just me and the keyboard. No mirrors.
Let’s see what happens.