AI made me faster at almost everything. Then, after two weeks of preparation, I sat down with an expert and couldn't explain my own work.
I used AI to write this. Claude drafted it from my notes and a long conversation about what happened, and then I edited dozens of times. That matters here more than usual, because this piece is about what happens to the work you don't fully make yourself.
Two weeks for one meeting
I spent two weeks preparing a deck.
The goal was simple. I needed to explain a proposal to a stakeholder from one of my accounts at work, and get their feedback. A meeting like that used to be the most normal thing in the world for me. I've had hundreds of them.
We sat down. We started.
And I went blank.
Not nervous. Not a little foggy. Blank. I knew I had done good work. I just couldn't find it in my head.
It wasn't a lack of preparation
That's the part that still bothers me.
I know how to work with AI. I never accept what the first prompt gives me. I actually start with pencil and paper. Ideas, connections, arrows between them. I let that sit until I've digested it.
Only then do I start the ping pong with the AI. I give it context. I question what comes back. I go back to my notes and compare. I make assets with it, and then I fix them by hand.
I want to be inside the loop.
So I wasn't the person who pastes a prompt and ships the output. I did the work.
The confession
Here's the other half, though.
Over the last months I've connected AI to almost everything. It's basically my assistant. Answer this person on Slack and explain where we are on that issue. Leave me a draft of an email to these three people about this topic.
At the same time, my responsibilities grew. Several accounts. Different projects running in parallel. Dozens of people every week, each with a different puzzle.
Leaning on technology to handle that is normal. It's what I'd recommend to anyone in my position.
And it works. It feels like I cover ten times what I used to.
But my sense of ownership dropped by about the same amount.
Those numbers are a feeling, not a measurement. The feeling is real, though. I got to a point where I couldn't explain my own work to other people without an outline and a summary the AI made for me.
How the meeting ended
Clumsy. That's the honest word.
I had a detailed structure in my notes. It didn't help, because the expert took the conversation somewhere else. They asked things in their own order, from their own angle. That's what experts do, and it's exactly why I wanted their feedback.
My structure was a script. Once we left the script, I had nothing underneath it.
So I dropped it, and I talked in basic mode. Loose words. Half ideas. Trying to stitch them together in real time, with a lot of effort.
It wasn't a disaster. We got somewhere. But it hit me hard.
And it was the first time I understood that the answer was not to go faster. It was the opposite. Even though, technically, I could scale ten or a hundred times past what I used to do.
Why I think it happens
This is a hypothesis, not a finding. One case, mine.
Before AI, the slow part of preparing something was putting it together. Deciding what mattered, writing the ugly first version, rewriting it until the argument held. That part was painful. It was also the part that left the project in my head.
Now, even when I start on paper and review everything, a lot of that assembly happens on the other side of the screen. I bring the judgment. The AI brings the stitching.
Psychology has a name for part of this. The generation effect (people remember what they produce better than what they only read) was described in the late 1970s. And a recent study from the MIT Media Lab found that people who wrote essays with ChatGPT had a harder time quoting their own essays afterwards, and felt less ownership of them. The researchers called it cognitive debt.
Reviewing is not the same as building. My name was on the deck, but I hadn't fully built it in my head.
What I haven't seen described much is my version. Not a student. Not a beginner. Someone experienced, careful with AI, doing good work, who can't explain it live a couple of weeks later.
In my notes I started calling it synthetic memory. What I "remember" of a project is a reconstruction the AI hands me when I ask. It can be completely accurate. It still isn't mine.
What I do now
I force myself to explain things to myself before the next call.
No deck open. No summary. Just me, out loud or on paper, answering: what is this about, why did we decide what we decided, and what would I ask if I were on the other side.
If I can't do it, I'm not ready, no matter how good the deck looks. It's uncomfortable, and it takes time I don't feel I have. It's also what gives me back the feeling that the work is mine.
And I've started treating every task as something I need to be present for, even the small ones, even the ones a process handles for me. Every person and every topic deserves that, with or without automation.
The FOMO part
There's a lot of noise right now telling us the opposite.
Magic prompts. Everyone is using them. If you don't get on the boat, you'll be left in prehistory.
I'm on the boat. I'm not getting off. But speed and the feeling of infinite possibilities make it very easy to end up in chaos. And when you're fast, more than ever, you need to stay connected to what you're doing and why you're doing it. Otherwise you get lost completely, and you don't notice until someone asks you a question you should be able to answer.
Back to the meeting
If that meeting happened today, the deck would look the same.
The difference is what would be underneath it. Not a script. Something I explained to myself the night before, so I could follow the expert wherever they wanted to go.
AI can make me ten times faster. Owning the work is still my job.
So here's my question: when did you last have to explain your own AI-assisted work live, without your notes, and how much of it was still really yours?
Originally published on Level Up Coding.
Top comments (0)