I looked over Chris’s shoulder and my jaw dropped when he showed me his list of scheduled tasks in Claude. We were sitting in the dining room of a friend’s house, talking about all the cool things we were doing with AI. I didn’t want to “brag” about the agents I was building. And it’s a good thing I didn’t.
I stood behind Chris, who was seated on a cushioned bench (when had those become so popular?). The standard centerpiece of desiccated flowers acted as a silent witness to the group of three of us looking over his shoulders as he navigated his laptop.
Chris pulled up his favorite “super script” and explained how much it had helped him. But it was not that specific job that had me so impressed. What stunned me to silence was the sheer number of “scheduled tasks” he had in his account.
My Claude account had exactly zero.
“What the heck are all those?” I asked, barely masking my surprise.
“Oh, they’re cowork tasks,” he explained casually. “Claude helped me figure out how to solve a problem, then I scheduled them so they run every day.”
“Do those run when you’re not at your machine, like when you’re traveling?” I asked, a small amount of fear squeezing my heart.
“Absolutely,” Chris said without looking back. “They even run when my machine is off. When I log back in, the work is completed, and the results are in my account.”
I took a step back. While Chris continued to show off his awesome automations, I moved to a chair at the head of the table and sank into it. A knot formed in my stomach. My Claude account had exactly zero scheduled tasks. Chris had solved the problem I was trying to solve in a completely different way. And his solution was, no doubt, much easier than mine.
How had I missed that?
I had been working with frontier AI since 2023. Yet, this had flown right under my radar. How?
Just to be sure, I pulled up my phone and asked Gemini to confirm what Chris had just told me. I use other AI agents to investigate each other, and Google’s super bot confirmed my fears. Everything Chris said was true. Claude had the ability to execute work tasks regardless of whether you were at your computer or not. Using their own infrastructure and virtual computing, a scheduled task could do work while I was traveling.
Shit. How. Had. I. Missed. That?
I had done the thing I try to avoid when I’m learning or coaching. I had glommed onto a “bad teach.” A bad teach is when you find a good-sounding idea that is actually wrong, and then you act as if it’s true without questioning it. For example, there are many ways to teach kids how to pitch. Most of them are wrong, and a few do more harm than good. There are lots of well-meaning coaches in youth baseball who are setting kids up for arm injury and don’t know it. That’s a bad teach.
I had grabbed a bad teach about what Claude could and couldn’t do, and I felt my face flush with embarrassment. I had made the classic mistake of developing a mental model about a technology, but never revising that model. In the early days Claude could not execute tasks when you were away. Believing I was an advanced AI user because I used Claude Code on the command line in a terminal, I disdained the desktop application. When Anthropic launched Cowork, I judged it to be a watered-down version of Claude Code, for non-technical folks. I was partly right, which helped justify my belief. It was easier.
What I didn’t realize then or later was that Cowork continued to evolve. I blissfully ignored the significance of what Anthropic created. To make matters worse, I had ridiculed people who become entrenched in fixed beliefs about technology. (How many people still talk about seeing AI images with six fingers, or bad text?) Yet there I was making the same mistake.
Anthropic had created the infrastructure needed to execute scheduled tasks regardless of whether or not I sat at my computer. Hell, my laptop didn’t even need to be on!
The Claude Desktop application had become a browser of sorts. It was a window into the world the AI worked in. The application mediated my communication with Fable, Opus, or Sonnet.
How had I missed this?
I slowly shook my head. Instead of the euphoria of learning something new, I felt the heartburn of getting something wrong. I knew then that what I knew about AI had become intermingled with my identity. I was a smart AI guy. So when I got something wrong, I was wrong, and that can be a very uncomfortable place to sit.
How do you keep up with everything going on in AI? It seems like Anthropic, OpenAI, Google, xAI, and now Meta release updates at a breakneck speed.
Taking a deep breath, I realized I had a needle to thread. On the one hand, I did not want to give in to the temptation to defend myself and my ideas. I had seen plenty of people double down on a belief and refuse to accept new information. I did not want to do that. On the other hand, I did not want to throw out everything I had done and learned. This is the other extreme. It is human nature to seek certainty, and one easy way to do that is to fall into an extreme position.
I could feel my own mental energies gather around wanting to argue that what Chris was doing could not be as good as what I had built. At the exact same time, I felt like everything I had built was garbage and a waste of time. That extreme thinking was the red flag. The truth, as they say, was somewhere in the middle. And the best thing I could do now was lean into the discomfort and get curious.
I stood up, fully expecting Chris to give me a jab or two for being so wrong about AI. I mean, he could have, right? But when I put my hand on his shoulder and asked him, “Hey, can you share that prompt with me? The one you used to create that workflow?” Chris looked up and smiled and said, “I’m happy to.” And he meant it.
At that moment I realized he was enjoying the same positive feelings I cherish when I help someone. While it might cause me some discomfort to receive a little bit of help, who was I to deny him the joy of sharing what he had learned?
Besides, it was good practice for me to practice a little humility. I still had a lot to learn. Who better to teach that lesson than a friend?