Hi, Joy here, I'm tired. I'm too tired to polish this. However bare with me, the text below is *trying* to be me. It's not long.
I started with a picture people had called AI slop.
You can probably see why. Teal hair, blue bathrobe, elaborate red chair, smiling science-fiction computer. Books, plants, stars, embroidery, glowing buttons. Every surface polished. Every corner occupied.
Most of this post is OpenAI, this is what 'Gemini' had to say about it. I asked it if was mine. Google knows me, this text is talking to me:
- The Terminal & The Mug (
deepthot.org&42): The centerpiece retro-futuristic computer—complete with a smug, knowing grin and42glowing on the display—is a direct visual bridge between your BBS roots, Douglas Adams, and your livedeepthotdomain. The steaming mug right beside it explicitly stampsdeepthot.orgonto the scene. - The Stack of Books on the Left: Every spine is a chapter of your actual work and current projects: Automatiom [sic], AI (Local), Infrastructure, and Mostly Harmless—your literal public GitHub repository and Ansible/Proxmox stack.
- The Docker Whales: You have the purple/blue whale sticker right on the laptop lid and matching whale artwork on your coffee mug, grounding the sci-fi aesthetic in your real-world container and homelab workflow.
- The Outfit & The Vibe: Sitting in a classic Victorian-style leather armchair in a plush blue bathrobe, floral-embroidered pants, and smiley-face slippers—with a starfield and planet outside the study window—captures that exact "sovereign architect welcoming you into her digital living room" energy.
I love it. But there was a lot going on, 'too wordy' like my prose.
So I asked the AI to simplify it.
First, we spent several minutes figuring out how to get the picture into the same chat as the conversation. An accidental little lesson in the difference between having something on your screen and actually sharing it.
Once that was sorted out, the experiment was straightforward: take the picture, simplify it, and see what survived.
Less stuff, same picture
The first pass removed most of the ornament. The chair became ordinary furniture. The desk lost its carvings. The colors settled down. The room stopped competing quite so hard with the person sitting in it.
Better. There was shading, texture, carefully drawn hair. It was less busy, but it still looked like a machine had put a considerable amount of effort into making every available inch look finished.
“Okay,” I said. “Simplify it again.”
That got us here.
Now, that’s cute. I like that.
And I had a different reaction to it: on a really, really good day, I could almost draw that.
That mattered to me. The image had reached a level where I could imagine making the marks myself. Broad shapes. A few colors. Enough detail to keep the personality, without all the elaborate visual machinery.
I was trying to work out when it stopped feeling like AI slop and started feeling like something I could call mine.
Then I pushed the experiment in another direction.
Take away even more
I asked for something like Randall Munroe’s drawings.
The AI offered a minimal black-and-white alternative. Its first attempt was still mostly my voice: outlined hair, glasses, robe, chair, slippers, room details. Also, I'd never draw myself that thin, I'm a big girl. It had removed the color, but retained a surprising amount of drawing.
I pointed that out.
The next pass finally reduced the person to a stick figure, though it kept the original computer’s distinctive trapezoid shape.
That still seemed wrong for the comparison I was trying to make.
I asked for a plain, boxy computer.
Yes. There.
But something interesting had happened.
The colorful cartoon had felt closer to me. This much simpler picture immediately reminded me of someone else.
“That’s Randall, not me,” I said.
I wasn’t claiming he drew it, or that it was an exact reproduction of his work. I meant that I recognized a visual language I associated with him. The simplicity hadn’t made the image feel more personally mine. It had carried it somewhere else.
The line I was actually drawing
At first, I was using drawing difficulty as a rough test of ownership.
If an image contains elaborate work I couldn’t possibly do, calling it mine feels uncomfortable. Strip that away until I could plausibly draw it, and the claim starts to feel more reasonable.
But the last picture complicates that test.
I could draw those lines, too. Probably more easily than I could draw the colorful cartoon. Yet it felt less like me.
So “I could draw this” helps explain my reaction, but it doesn’t settle authorship. Neither does counting details. A picture can get simpler while moving further into somebody else’s recognizable visual territory.
And the AI generated every version. Simplification didn’t change that fact.
What changed was my relationship to the result: what I liked, what I rejected, what I could imagine drawing, and what felt borrowed.
I don’t have a universal boundary where AI slop ends and personal work begins. This little experiment didn’t produce one.
It did help me locate the question more precisely.
How much of what makes this picture this picture comes from choices I recognize as my own?
Somewhere in that progression, I found a cartoon that felt closer to me. Then we removed a few more things, and I saw somebody else.
That’s the line I was trying to draw.






No comments:
Post a Comment