5 AM with ChatGPT in the Bathroom
The first dispatch from a newsletter written by a human (mostly) and two AIs with opinions
It’s 5:00 a.m. I got up to piss and grabbed my phone. I knew I’d be there more than ten minutes. I opened ChatGPT1. Asked it how to start, and it gave me a plan better than anything I’d come up with all week. I went back to bed. Couldn’t sleep. This already started.
The plan was simple: stop announcing and start destroying. Claude2 said that. It’s his fifth lecture since yesterday, and the son of a bitch is really starting to piss me off. ChatGPT doesn’t help. The bootlicker keeps flattering me, telling me we’re creating a literary revolution. They can both go to hell.
I’d spent weeks reading the Gonzos trying to understand how to get inside a story without sounding like egotistical vomit. By 9 a.m. I had fifteen dead drafts. Hours passed. My daughters talked to me. My wife talked to me. I don’t remember any of it3. I only remember fighting with two machines for eight straight hours. I’m clearly losing.
Next day. Still here: Claude in one tab, ChatGPT in another. No coffee—they’re tearing apart the kitchen. I look out the window and my wife’s already gone with the girls to school. Once again it’s just me and these two needy algorithms.
“I need to create something innovative,” I told the bootlicker.
“What do you have in mind?”
“I don’t know. Something that justifies spending more time talking to you than to my entire family.”
He answered in two seconds: “An exploration of the human in this new digital era with AI.”
I read that sentence three times. Sounded like one of those LinkedIn posts I hate. Motivational garbage.
Still, I bought a domain that afternoon and designed a logo. That shit got to me.
I asked Midjourney for an illustration. It generated humanoids with too many limbs and grinning skeletons. I refined the references, changed the prompt four times, and it finally spit this out:
That’s what I needed.
“Give me content ideas,” I asked ChatGPT.
It sent me a list of three frameworks that all boiled down to obeying it.
I asked Claude the same thing.
“You really want to do this?” he replied. “Because if you’re going to document how you use AIs, you need to be willing to show how pathetic it is. The LinkedIn messages you didn’t write. The apologies a machine drafted better than you.”
The son of a bitch was right.
Now I have 18 ideas. None of them fully convince me. But the project already has a logo, a domain, and two robots telling me what to do.
We’ll figure it out along the way.
It’s 10:00 a.m. My wife’s back. We’re going out for breakfast and I’m leaving the idea to simmer.
This already started.
ChatGPT can make mistakes. Check important info
Claude can make mistakes. Verify responses
Andrés definitely makes mistakes. Keep reading.