I have never, ever had workdays as strange as the ones a week ago.
This happened during the final phase of the platform and name change from Ghost to Substack and from Warp News to The Angry Optimist. A big project that absolutely could not go wrong. Over 7,000 articles, 20,000 subscribers, a large number of them paying.
There wasn’t the slightest chance I could handle all of that myself. I don’t have those kinds of technical skills.
So instead I put ChatGPT in charge of the move. Before, ChatGPT could tell me how to do things, but as of a few weeks ago it can use my computer to do them itself.
ChatGPT and I first spent a lot of time carefully mapping out everything that needed to be done, how it would be done and in what order.
A little over a week ago, we had gotten to the point of actually doing it.
A whole bunch of things had to be done, which happened in two or three parallel processes on my computer. Since I wanted to minimize the risk of mistakes (this is new, you have to be careful), ChatGPT had to check every decision with me. My job became writing different versions of “Yes, go ahead,” “No, change this,” “What’s the next step?” and so on.
Between my answers I tried to work on other stuff, but got interrupted by a new question like every three minutes. I got nothing done. Eventually I gave up and started scrolling social media.
I was like a lazy but micromanaging boss.
Even weirder
But it was about to get even weirder.
Since ChatGPT was using my browser for a lot of the work, my scrolling and the AI’s work started to collide. When I clicked around, windows ChatGPT was working in could disappear, and it had a hard time navigating.
After a while, ChatGPT asked me: “Can you take a 15-minute break while I finish this task?”
I just stared. And then I cracked up.
I had to go grab a coffee and surf on my phone while the AI finished its work.
But as we entered a critical phase of the move where absolutely nothing could go wrong, the AI needed to work completely undisturbed for longer stretches.
In the end, I had to dig out my old PC.
So, locked out of my new, expensive MacBook while I sat with a nearly antique computer from 2018.
Oh wonderful, absurd new world?
Absurd and wonderful at the same time.
Because obviously it can’t be like this, with me unable to use my own computer when the AI is using it. A teething problem that will soon be fixed, I thought.
It turned out, however, to be even easier to fix than that.
The problem had come up because of a mistake on my part.
When you use ChatGPT in an app on your computer, it can use its own browser. But then you have to log in to every site it needs to access. In a fit of laziness, I told ChatGPT to use my browser instead, where I was already logged in everywhere. Little did I know then that this problem would come up a couple of weeks later.
Now that I’ve spent a few minutes logging it in to its own browser too, the problem is gone. We can now work in parallel.
The gap between idea and execution is shrinking
Still, I’m kind of glad this happened. It sort of pinpoints the dramatic shift we’re at the beginning of. That an AI can use my computer and help me with advanced, time-consuming and boring tasks.
When I wrote a few weeks ago about the biggest AI leap since ChatGPT launched, this is exactly the kind of thing I meant. How the gap between idea and execution is shrinking.
As we roll this out across the world, it will sometimes get weird, wrong and crazy, but above all it gives us the chance to accomplish more than before. In some cases much more.
Mathias Sundin
Angry Optimist