Is It Still Worth the Time?

In my last post, I wrote about how I realized that I spent way too much time during the week interacting with my AI agent. That, however, doesn’t line up with how I envisioned I’d use it. It should make my life easier, so I can be out in nature meditating while it does the annoying stuff.

But what exactly is the annoying stuff? I think listing items on online marketplaces falls into that category. You have to come up with a description, take photos, figure out a price, shipping included, and find the correct category for the item. That’s a lot to do given that I’ll either charge €2 for it or give it away for free. I’d rather send a bunch of photos to my AI agent and have it handle all of that.

So that’s what I did. I had the AI agent create a reusable skill. To my surprise, that took way more iterations than I had anticipated. For example, it initially failed to upload images with the listings. It also didn’t consider the condition of an item when estimating a price, and sometimes made up details about the condition it couldn’t verify from the picture alone. At one point, I thought we had it all sorted out and gave it a picture of a hacky sack. It listed it as a juggling ball.

That’s when I stopped. With the time I spent validating this skill, I could’ve listed everything 10 times. It reminded me of an old xkcd comic with the very fitting title “Is It Worth the Time?”¹ It shows how much time you can afford to spend on an automation given how much time it saves you. For example, if you save 1 hour on a task you do once a year, building that automation should take less than 5 hours for it to pay off over five years.

That chart was usually shown to software engineers when they had spent or were about to spend too much time automating a short and rare manual task. Most of the engineers I know, including myself, would do it anyway. It was fun, and you usually learned something useful along the way.

Nowadays, it seems like we can automate anything with just a few instructions to an AI agent. But that’s not the whole story. You still have to validate its output, which is time-consuming. I wouldn’t mind that work if it were at least fun or educational, but it isn’t. Ironically, now I find myself seriously asking whether it’s actually worth my time. This time, the answer is no. So the fastest way out into nature is to just list the hacky sack myself.