Just one more prompt...

Claude Code gives me a constant stream of dopamine hits but also FOMO, guilt, and anxiety.


My dog Soup needs a walk. I clip the leash on, grab my headphones, open Claude Code on my laptop one more time, kick off an agent session, and head out the door. Halfway down the hill I get the approval prompt and tap yes without breaking stride. Soup pees on something. I tap yes again. By the time we’re back, the agent has refactored a service, opened a PR, and is waiting on me to build the next thing.

I thought this would feel great. But in reality I feel like I’m getting further and further behind.

There’s a new async background process running in my head1. It says: “Ben, the GPUs are ready and willing. Opus is just sitting there, fully capable. The only thing standing in the way is you coming up with the next idea fast enough.”

If I’m not prompting, I’m wasting it. If I run out of ideas, I should have more ideas. Shouldn’t I have more ideas? What’s wrong with me that I don’t have more ideas?

Some of it is FOMO. Every prompt I don’t run is something that won’t exist in the world. Some of it is guilt. The tools are here and cheap; all the friction to build is gone and I’m still not shipping fast enough. But the primary feeling is anxiety: imagination is the bottleneck, and the bottleneck is now me.

Coding agents is the obvious place this shows up, but it’s so much more than code for me. I use Claude Code to ship SuperDuper features, sure, and coding is the version of this story everyone tells. But I also use it for the bar mitzvah speech I’m rewriting, and the personalized weather app I built for my Pixel watch, and an elaborate prank I’m running on my kids involving Gabi’s Wordle streak, and a redesign of my personal blog, hacking my irrigation system, and on and on. Every part of my life that used to have a sign labeled “Not today, Ben” has had the sign removed.

Like most nerds, I buy domain names for every cool idea based on the theory that one day I’d have some free time. (My favorite: investiv.us, a personal finance app “for the rest of us.” Yes, a Seinfeld reference from the 1990s, in case you were doubting how long I’ve been squatting domain names in the hopes of a free weekend.) For twenty years, each one came up for renewal and I’d say “Yes! This is the year!” (Narrator: it wasn’t.) But in 2026, I no longer have an excuse not to build all of them. I just need one more prompt. And that is the voice in my head. It’s substantially worse than the $10 GoDaddy charge.

The scarce resource to building has always been capacity. Time, skill, the dollars it would have cost to hire someone who had the skill, the bandwidth after work and kids and dog. The gap between the things I wanted to do and the things I could do is, well, comfortable. The unbuilt things stayed unbuilt and I didn’t feel bad about it because they were not realistically possible anyway.

But now, compute is so cheap and agents are so capable that my creativity is the expensive scarce thing in the system. My imagination, my ability to specify the next useful task, my willingness to actually start the thing instead of just thinking about starting the thing.

The problem I’m describing with AI agents is obviously new, but is the mindset and the anxiety feeling? As always, I like to look to history. The closest parallel I could come up with is personal finance: every dollar not invested is a dollar not compounding. It’s the guilt about how much cash is sitting in your checking account when you could invest it with a few clicks, but you don’t. The loss is the money you would have made if you’d deployed the money you have, but the loss is invisible because it’s counterfactual.

Every prompt I don’t run is a counterfactual: a counterfactual product, a counterfactual essay, a counterfactual better speech, a counterfactual prank I could have pulled on Gabi. The graveyard of unbuilt things is growing larger every day!

The other historical example, and it might be a useful analogy, is industrial engineering. Industrial engineers spent the early 20th century obsessed with utilization rates — a stopped assembly line is just burning money. Just-in-time, three-shift operations, lights-out manufacturing, a hundred years of management theory to ensure the expensive constrained resource never idled. That’s how I’m feeling.

The industrial engineering one is instructive because they eventually figured out that running a constraint at 100% utilization produces worse output, not better. A system with no slack can’t absorb the small variances that always show up, can’t recover from the small things that always go wrong, can’t notice when it’s producing the wrong thing. The whole Toyota Production System, queuing theory, and operations research from 1960 onward converged on the same conclusion: systems produce more useful output when they have slack.

So if I am the constraint (and I am), then running myself at 100% utilization is a failure mode. The dog walk isn’t time stolen from the GPUs, it’s the slack that lets the next prompt be a good prompt instead of just a prompt. Shower thoughts happen in the shower for a reason: you don’t have a keyboard in front of you!

I always berate my employees when they log in to work during vacation. Take the vacation! We’re giving you time off for a reason. Today I need to give myself the same advice.

I am writing this post on a Friday evening. There is a Claude Code session waiting for me in another window. I’m going to close my laptop before I go to dinner2. Anthropic’s TPUs will be fine without me. They’ll miss me of course, but they have other people to serve. My graveyard will have a few more unbuilt things. And I need to become okay with that. Because the alternative, me at 100% utilization generating a steady stream of mediocre prompts because I’m afraid of silence and FOMO, produces worse outcomes and a worse life.

Also, Soup just peed on the floor while I was writing this post.

  1. Yes, the technical term is rumination. You say demon, I say daemon. 

  2. Narrator: he didn’t. 


I'm Ben Stein, co-founder and CEO of SuperDuper, helping underwater parents manage all their family logistics. I live in Oakland with my Keeper wife Arin, our two overprogrammed teenage boys, and a dog named Soup who contributes nothing helpful to the family logistics and arguably makes them harder, but he's really cute.

If you want to follow along: superduperlabs.com · benjaminste.in · LinkedIn · Substack