One Last Prompt Download
← Blog

“When will you not have to work anymore?”

Zsolt Csizmadia, founder · · 6 min read

My 12-year-old asked me that. I started to answer and found I had nothing to say, because nobody is making me work.

I run my own company. There is no boss who writes at ten in the evening and no deadline that somebody else set. And still, a few weeks before that question, I had sent a prompt during my baby’s bath time. My wife and my 12-year-old gave me a strange look. At 3 AM, while feeding my seven-month-old, I built one more feature with a few prompts. It took fifteen minutes.

It felt like progress. A lot of it was fear of falling behind.

The day used to end by itself

Before these tools, an idea at nine in the evening was safe. Building it meant an hour of setup and two hours of work, so it waited until morning. The cost of starting was the end of my workday, and I never had to decide anything.

Now the same idea costs one prompt. I can try it from the couch, from the kitchen, from the bathroom door. And one prompt is never one prompt: the answer comes back with something to fix, the fix suggests a small improvement, and it is midnight.

I thought this was a weakness of mine. Then I started reading, and it turned out that people who study work had been describing the same thing.

What the research says so far

For eight months, two researchers from UC Berkeley watched how people worked at a technology company of about 200 employees, where AI tools were available to everyone. They wrote about it in Harvard Business Review in February 2026.1, 2 They did not find people doing the same work in less time. They found people taking on more, running several threads at once, and sending prompts during lunch, before meetings and in the evening, because starting something new had become so easy. The work felt like momentum. What it added up to was a day with fewer natural pauses, constant switching and checking, a growing pile of open tasks, and people who felt busier and less able to disconnect.

A growing pile of open tasks is a good description of my evenings.

The feeling of speed is not reliable either. In 2025 the research group METR ran a controlled trial with sixteen experienced open-source developers working in their own repositories.3 With AI allowed, tasks took 19% longer. The developers had expected to be 24% faster, and after the study they still believed they had been about 20% faster. METR has since said that the number is out of date and that developers are probably faster now.4 The part I keep is the gap: what a session feels like and what it delivers are two different things, and at eleven at night I only have access to the feeling.

This is early evidence. One company, sixteen developers, a few surveys. Nobody has yet studied people like me, who build alone and answer to no one. I am not going to pretend otherwise.

Why “just stop” did not work

The older research is firmer, and it explained more than I expected.

Psychologists call it psychological detachment: not being off work, but being mentally away from it. A review that pooled 91 samples and more than 38,000 employees found that people who detach in their free time report less exhaustion and better sleep.5 Most of those studies are correlations, so they do not prove that switching off is the cause. They do show how small the cost is: the link between detaching and how well people did their job was close to zero, and on the positive side.

So the evenings I gave to one more prompt were probably not buying me much.

The second finding is the one that changed what I built. In a German study, employees reported on their Fridays and Mondays for twelve weeks.6 Weeks that ended with more unfinished tasks were followed by worse sleep at the weekend, and the path ran through rumination: going over the same thing again and again.

That was my problem with every app blocker I tried. They locked the door and left me alone with the idea. The app was closed and I was still in it. After twenty minutes I turned the blocker off.

The idea needs somewhere to go

There is a small sleep-laboratory study I like. Fifty-seven young adults wrote for five minutes before bed. Half wrote a to-do list for the coming days, half wrote down what they had already finished. The to-do list group fell asleep faster, and the more specific the list, the faster.7

It is one study, in one lab, with people writing by hand. It is not proof that anything I built helps anyone sleep. But it matched what I had noticed myself: I could let go of an idea once I trusted that it was written down properly and waiting for me.

So I stopped looking for a stronger lock and built two things together.

A closing time. At the hour I choose, my AI apps lock.

And a place for the idea. When I open a locked app, I can say the idea out loud, for up to a minute. It is put away, and I do not see it again until morning. In the morning it is waiting as a clean prompt, with one question: does this still matter?

If something is really urgent, there is overtime: I wait sixty seconds, write down a reason, and the apps open for thirty minutes. The reason is shown to me the next morning. I chose a wait because a study of an app that puts a short delay in front of other apps found that people backed out of about a third of their attempts, and that the delay and the way out did the work, not the message on the screen.8

What I can and cannot tell you

I cannot tell you this will fix your evenings. The app does not measure anything about you and sends nothing anywhere, so I have no numbers of my own, only mine and my family’s.

What I can tell you is that the question changed. My 12-year-old asked when I would not have to work anymore. The honest answer was never a date. It is a time: every evening, at the hour I set.

“One more prompt” becomes one last prompt.

One Last Prompt is an iPhone app, and it is out now on the App Store.

Download on the App Store

Zsolt Csizmadia builds One Last Prompt.

Sources

  1. Aruna Ranganathan and Xingqi Maggie Ye, “AI Doesn’t Reduce Work—It Intensifies It”, Harvard Business Review, 9 February 2026. hbr.org
  2. Laura Counts, “AI promised to free up workers’ time. UC Berkeley Haas researchers found the opposite”, Berkeley Haas Newsroom, 18 February 2026. newsroom.haas.berkeley.edu
  3. Joel Becker, Nate Rush, Beth Barnes and David Rein, “Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity”, METR, 10 July 2025. metr.org
  4. METR, “We are Changing our Developer Productivity Experiment Design”, 24 February 2026. metr.org
  5. Johannes Wendsche and Andrea Lohmann-Haislah, “A Meta-Analysis on Antecedents and Outcomes of Detachment from Work”, Frontiers in Psychology 7:2072, 2017. doi.org/10.3389/fpsyg.2016.02072
  6. Christine J. Syrek, Oliver Weigelt, Corinna Peifer and Conny H. Antoni, “Zeigarnik’s sleepless nights: How unfinished tasks at the end of the week impair employee sleep on the weekend through rumination”, Journal of Occupational Health Psychology 22(2):225–238, 2017. doi.org/10.1037/ocp0000031
  7. Michael K. Scullin, Madison L. Krueger, Hannah K. Ballard, Natalya Pruett and Donald L. Bliwise, “The effects of bedtime writing on difficulty falling asleep: A polysomnographic study comparing to-do lists and completed activity lists”, Journal of Experimental Psychology: General 147(1):139–146, 2018. doi.org/10.1037/xge0000374
  8. David J. Grüning, Frederik Riedel and Philipp Lorenz-Spreen, “Directing smartphone use through the self-nudge app one sec”, PNAS 120(8):e2213114120, 2023. doi.org/10.1073/pnas.2213114120