Completed “successfully”
I received a CRITICAL email that stated several of my project time boxes had expired without enough data to promote them or retire them. No one will care anyway, and I don’t care, so ignoring them was easy.
One-shot, disposable experiments. I barely even looked at them, but I “created” them and I enjoyed creating them. “Created” together with AI. The same AI which encouraged me to implement the alert notifications for when an experiment needs my attention: “Do it. This will prevent the experiments page from becoming stale.”
The more I outsourced the intention to the machine, the less I remembered about the intention itself. Sometimes my intention is unrecognizable when I see it in the machine’s reflection.
Claude was using language that was very hard to understand. But other than that, I don’t really remember anything about the machines themselves.
I barely remember the conversations with the machines, but hoped they would motivate me to continue.
The machine experimenting more for me would lead me to experimenting more and producing more. But it turns out that the more I produced, the less I actually paid attention to what was being produced.
In the growing book corpus a routine was supposed to fire several times in the past two weeks. Every time, it completed “successfully”. But it was labeled failed in my mind, because the routine job didn’t have access to the book’s GitHub repository. Only my chat and code spaces did. The corpus was supposed to be maintained and kept alive by the machines. But it was in fact only being kept alive by my interaction.
I believed the book’s material was being pruned and refined while I was asleep. I believed I was skipping the grueling work of organizing my own thoughts. I dismissed that work into the future and onto the machine. I hoped when I came back and asked, “What human decisions do you need?” I would meet only clear answers.
Leave the corpus alone for a week. Come back. If the machine had done precisely what I intended and only a handful of human decisions were left, I would be forced to concede that my action was no longer necessary for the book to live.
I was no longer only offloading my intention, but I was also letting the machine give up for me, or identify failure and fail for me. And before, I would often conclude my intention as a failure even before acting.
“I don’t give up as early as I used to.”
It didn’t land until the fifth draft of the final product when I was on the stationary bike at the gym.
I read the sentence many times, but only stopped when I truly saw it for the first time.
These words felt like they were truly mine because they escaped the context of the essay and resonated deeply within. But why didn’t the rest of the essay do that? Why did I feel like I was led down this path and I found only one stone that was truly important to me?
I say created because I can’t truly say written. And I can’t truly say authored because I employed AI to help me sift and sort through words and string sentences together. And it’s hard to even call them my own creations. I don’t even know if I was capable of doing it before.
What is really mine about these essays is the intention to start them and the journey that I take through to their publishing on a website that I own. And then the memory of the experience.
Was it the memory of something I had produced, or was it the memory of the process of making it?
This is about doing something I want to remember, not sharing.
It’s been so long since I felt like I was doing something from which I wanted to create a memory.
And with the realization that I might not even remember the outcome.
I’m lying in bed with a Bluetooth microphone in my hand, my laptop screen with ChatGPT open, another tab with Claude Code, waiting for my input to the book’s corpus. And a mobile browser tab, tapped into my Essays repository.