---
title: Time to Make Things Has to Be Set Aside
url: https://oosioo.com/en/p/%EB%A7%8C%EB%93%9C%EB%8A%94-%EC%8B%9C%EA%B0%84%EC%9D%80-%EB%94%B0%EB%A1%9C-%EB%96%BC%EC%96%B4-%EB%91%90%EC%96%B4%EC%95%BC-%EC%83%9D%EA%B8%B4%EB%8B%A4
date: 2026-10-16T13:03:32+00:00
author: SYSOP
summary: The line in my notebook read, "Devote a limited, special stretch of time to creating something new." These days we spend most of the day looking at things other people made, and with AI making production cheap, the supply grows faster still. I went through studies showing AI beating people on creativity tests and studies showing the best people still ahead, then the claim that plumbers are making a comeback and where it falls short. Finally I set out which hours of the day to set apart, and how, to keep human creativity alive. The fifteenth piece in A Page for Today.
---
# Time to Make Things Has to Be Set Aside

![A sheet of lined notebook paper reads "Devote a limited, special stretch of time to creating something new." On the left is a timetable of the day's 24 hours: most cells are shaded as watching time, and only a few are set apart as making time. Above it is the title "Time to Make Things Has to Be Set Aside"](/uploads/46064fa8098effc9.webp =720)

When I opened my idea notebook today, this is the line I found.

> Devote a limited, special stretch of time to creating something new.

Reading it, I went back over my day. The moment I woke up, I skimmed the news on my phone. On the way to work I flicked through a few short videos. Over lunch I read something someone else had summarized, and tonight I plan to watch a drama someone else made. Nearly my whole day is spent looking at things other people made. How many minutes did I spend making something from scratch?

## The supply of things to watch keeps growing, and keeps getting cheaper

I'm not the only one. In Korea's 2025 Broadcasting Media Usage Survey, run by the Korea Media and Communications Commission, 92.0% of people aged 13 and over used a smartphone at least five days a week. 81.8% used streaming video services, and among the types of service people used most, short-form video came first at 78.9% ([Dailian](https://www.dailian.co.kr/news/view/1592552/)). Watching has already become the default setting of the day.

Now add AI. The cost of making something has fallen close to zero. The numbers from the music streaming service Deezer show this well.

![The share of daily uploads to Deezer that were made entirely by AI: about 10% in January 2025, 18% in April, 28% in September, and 44% in April 2026 (about 75,000 tracks a day). Over the same period, AI tracks accounted for only 1–3% of all plays](/uploads/dc63a55a6566e8bc.webp =720)

After Deezer added an AI-music detector in January 2025, it reported that about 10% of the tracks uploaded each day were made entirely by AI. That share reached 18% in April and 28% in September ([Digital Music News](https://www.digitalmusicnews.com/2025/09/11/deezer-ai-tracks-daily-stat/)), and by April 2026 it was 44%, about 75,000 tracks a day. Yet these tracks make up only 1–3% of all plays, and most of those plays are flagged as fraudulent ([Engadget](https://engadget.com/entertainment/music/deezer-says-ai-made-songs-make-up-44-percent-of-daily-uploads-163642921.html)).

The gap between those two numbers caught my attention. Even when AI makes nearly half of what gets uploaded, the songs people actually go looking for are still the ones people made. Making got cheaper, but what people seek out did not. So what is still left in the things people make?

## The case that AI is already creative enough

Let me start with the evidence that AI is already creative enough. There is more of it than you might expect.

In 2023, a team led by Erik Guzik at the University of Montana had ChatGPT, running on GPT-4, take the Torrance Tests of Creative Thinking. The scoring service did not know which answers came from the AI. The result: the top 1% for both originality and fluency ([University of Montana](https://www.umt.edu/news/2023/07/070523test.php)). The same year, Mika Koivisto of the University of Turku in Finland and Simone Grassini of the University of Bergen in Norway asked 256 people and three chatbots to come up with unusual uses for a rope, a box, a pencil and a candle. On average, the chatbots scored higher than the people ([Scientific Reports](https://www.nature.com/articles/s41598-023-40858-3)).

Research ideas are no different. In 2024, Chenglei Si, Diyi Yang and Tatsunori Hashimoto at Stanford brought together just over 100 natural language processing researchers and had them compare, blind, research ideas written by an AI with ideas written by experts. The reviewers rated the AI's ideas as more novel ([arXiv](https://arxiv.org/abs/2409.04109)).

If creativity means the ability to play a move people would not think of, there are older examples too. In the second game against Lee Sedol in 2016, AlphaGo's Move 37 was one a human player would make perhaps once in ten thousand times. The commentators were puzzled at first, but by the end of the game it was clear that this move had decided it. In May 2025, Google DeepMind's AlphaEvolve found a way to multiply two 4×4 complex-valued matrices with 48 multiplications. In that setting, Strassen's 49 from 1969 had stood unbroken for 56 years ([Google DeepMind](https://deepmind.google/discover/blog/alphaevolve-a-gemini-powered-coding-agent-for-designing-advanced-algorithms/)).

After all that, it is hard to keep saying "AI is not creative" as it stands. At least not on a test sheet, a Go board or a page of equations.

## The case that the very top still belongs to people

Look a little closer at those same studies, though, and a different picture emerges.

The title of the Koivisto and Grassini paper says it outright: "Best humans still outperform artificial intelligence in a creative divergent thinking task." The chatbots won on average, but the person with the best answer beat every chatbot's best answer in seven of eight scoring categories. The peak above the average still belonged to people. The human answers also included plenty of poor ones. Human creativity is uneven, but its tip is sharp.

In 2025 the Stanford team published a follow-up. This time they did not just rate the ideas; they carried them out. Forty-three researchers were each randomly assigned either an AI idea or a human idea, each spent more than 100 hours executing it, and the resulting papers went through blind review again. The AI ideas' scores dropped significantly more than the human ideas' on novelty, excitement, effectiveness and overall rating. On many measures the ranking flipped and the human ideas came out ahead ([arXiv](https://arxiv.org/abs/2506.20803)). The team called this the "ideation–execution gap." An idea that looks plausible and an idea that works when you try it are not the same.

The study that stayed with me longest came from Anil Doshi of University College London and Oliver Hauser of the University of Exeter in 2024. They had people write short stories, and showed only some of them story ideas generated by an AI. The stories written with AI ideas were rated as more creative and better written. The less naturally creative the writer, the more they gained. But when all the stories were put side by side, they were far more alike ([Science Advances](https://www.science.org/doi/10.1126/sciadv.adn5290)). Each person got better, and the world they added up to became more uniform. The researchers called it a social dilemma.

This is how I read the three studies. AI is very good at producing the average of good things. Judged by that average, it beats most people. But the one point far from the average, the one you can only find by actually trying, the one that is valuable because it is different, still comes from people more often. Articles about general AI drawing near appear every week, but as of today that is how it stands. So I tend to think that in most creative fields anyone can be better than AI. That "anyone" comes with a condition, though. I'll come back to it.

## The idea that plumbers are making a comeback

There is a claim you hear often these days: in the age of AI, trades like carpentry, electrical work and plumbing will rise again.

Geoffrey Hinton, often called the godfather of deep learning, appeared on the podcast *The Diary of a CEO* in June 2025. Asked what work to do in an age of superintelligence, his answer amounted to: learn plumbing. AI will take over routine intellectual labor, he said, but it will be a long time before it handles physical things as well as people do ([Storyboard18](https://www.storyboard18.com/digital/plumbers-will-survive-automation-better-than-office-workers-says-godfather-of-ai-70893.htm)). Nvidia's Jensen Huang said at Davos in January 2026 that plumbers, electricians and construction workers would earn six-figure salaries ([Bloomberg Law](https://news.bloomberglaw.com/california-brief/nvidia-ceo-says-plumbers-in-demand-as-ai-upends-job-market)).

The two men's reasons are not the same. Huang is talking about demand: building AI data centers takes a lot of people to lay power and run pipes. Hinton is talking about ability. As the roboticist Hans Moravec pointed out in the 1980s, walking and grasping are harder for a computer than chess. What is easy for people is hard for machines, and this "Moravec's paradox" has not yet been solved.

I'd like to add one more thing. From the outside, plumbing looks like repeating a set procedure, the kind of work that should be easy to quantify and automate. In practice, no two jobs are the same. The pipes inside the wall of a 40-year-old apartment run differently from the drawings, someone who fixed them before left a strange joint, and the owner has a fixed amount they can spend. In between all that, you have to build a new answer every time. I don't know what to call that if not creativity.

Others got here first. In his 2004 book *The Mind at Work*, the education scholar Mike Rose spent a long time following carpenters, welders, hairdressers, plumbers, electricians and restaurant servers at work. Carpenters do complex spatial math in their heads; a server juggles orders, customers' moods and the kitchen's pace all at once. Rose wrote that the judgment, memory and perception in this kind of bodily work have always been undervalued because of the stereotype that splits "mind work" from "body work" ([Penguin Random House](https://www.penguin.com.au/books/the-mind-at-work-9781101174944)). Matthew Crawford, who holds a PhD in philosophy, quit as head of a Washington think tank after five months and opened a motorcycle repair shop. In his 2009 book *Shop Class as Soulcraft*, he says that standing in front of a motorcycle that won't start, forming a hypothesis, checking it with tools and forming a new one when it's wrong, was deeper thinking than any office job ([NPR](https://www.npr.org/transcripts/106513632)).

There is evidence on the other side too. Hinton himself said in the same conversation that AI will eventually be better than us at everything. Right now in India, thousands of people strap cameras to their heads and film themselves cooking and folding towels. That footage, paid at around 250 rupees an hour, becomes data for robots learning to copy human hand movements ([Al Jazeera](https://www.aljazeera.com/gallery/2026/6/11/photos-indias-workers-are-training-ai-robots-to-take-their-jobs)). If today's plumber is safe only because robot hands are still clumsy, that safety shrinks as the hands improve. So what keeps the trades going, as I see it, has less to do with the job title than with the ability to build a new answer in front of a different site every time. That ability exists in offices, in kitchens and at desks too.

## Creativity shrinks when you don't use it

I said earlier that "anyone" comes with a condition. Creativity is partly an inborn talent, but it is more like a muscle that rusts when you don't use it.

In 2025, a team led by Nataliya Kosmyna at the MIT Media Lab divided 54 university students into three groups and had them write essays. One group used ChatGPT, one used a search engine, and one wrote with no tools at all. EEG showed that the no-tools group had the broadest and strongest brain connectivity, and the ChatGPT group the weakest. The ChatGPT group's essays resembled each other, and the writers struggled to recall sentences they had just written. In a fourth session the two groups swapped conditions, and the difference grew sharper. Those who went from ChatGPT to writing alone showed less brain engagement. Those who went from writing alone to ChatGPT, on the other hand, remembered more, and several brain regions were active ([arXiv](https://arxiv.org/abs/2506.08872)). It is a preprint that has not been peer reviewed, so it should be read with care, but the point that order matters lines up with the Doshi and Hauser study above. When my own thinking comes first and AI comes after, AI becomes a tool. When AI comes first, there is less room for my own thinking.

Things to watch take up space the same way. In 2014, a team led by Timothy Wilson at the University of Virginia had people sit alone in an empty room for 6 to 15 minutes and simply think. Most found it unpleasant. In one experiment, the room had a button that let people give themselves an electric shock. Among participants who had said beforehand that they would pay to avoid being shocked, 67% of men and 25% of women pressed it at least once ([Science](https://www.science.org/doi/10.1126/science.1250830)). Fifteen minutes of just thinking, with nothing to look at, is that hard to bear. Short-form video fills those empty fifteen minutes for us, so we almost never live through empty time.

Yet empty time is exactly where thoughts grow. In 2014, Sandi Mann and Rebekah Cadman at the University of Central Lancashire in the UK had people copy numbers out of a phone book for 15 minutes and then come up with uses for plastic cups. Those who had done the boring task thought of more uses than those who hadn't. Those who only read the phone book, the most bored of all, thought of the most. The researchers concluded that boredom invites daydreaming and daydreaming leads to creativity ([Creativity Research Journal](https://doi.org/10.1080/10400419.2014.901073)). This ties back to the ninth piece, "[Scrub the Tub, and You'll Want to Write Again](/en/p/욕조를-닦다-보면-다시-쓰고-싶어진다)," and the tenth, "[Ideas Come When You’re Not Looking, and Leave Just as Fast](/en/p/아이디어는-딴짓할-때-와서-금방-간다)."

## How to set aside a limited, special stretch of time

The words I came back to in the notebook line were "limited" and "special." Time for creating doesn't need to fill the whole day. In fact, it can't. But it does need to be set apart so it doesn't mix with the rest. Here is what I took from the studies and stories above.

**First, set it aside in blocks.** In his 2009 essay "Maker's Schedule, Manager's Schedule," the programmer and investor Paul Graham wrote that managers cut their day into one-hour slots, while people who make things need to use time in units of at least half a day. One meeting in the middle of a morning breaks the whole morning into pieces too small for real work ([Paul Graham](https://paulgraham.com/makersschedule.html)). If half a day is too much, even an hour a day is good, as long as no notifications cut into that hour.

**Second, make before you watch.** If you open the day with other people's things, that day's thinking starts from other people's things. In the first hour of the morning, before opening your phone, write down even one line of your own. That is also why this series begins by opening an idea notebook.

**Third, my draft first, AI after.** Like the group in Kosmyna's study that wrote alone first and then got AI, I first sketch something with my own head, however rough, and then have AI polish it. Keep that order and AI becomes a tool for widening your thinking, and you can step a little to the side of the "everyone becomes alike" world that Doshi and Hauser described.

**Fourth, allow boredom.** Don't fill an empty fifteen minutes with a screen right away. Walk, wash the dishes, or just look out the window. As Wilson's experiment shows, it is uncomfortable at first. Bear with it a little, and as in Mann and Cadman's experiment, your thoughts start to wander.

**Fifth, fix something with your hands.** As Crawford thought in front of a motorcycle, work done by hand gives you a different problem every time. A broken drawer rail or repotting a plant, it doesn't matter. Meeting one small site with no set answer each day keeps the creative muscle in use.

## Back at the notebook

AI will keep getting better. The list I wrote today of what AI still can't do may need rewriting in a year or two. Still, one thing seems unlikely to change easily. In a world where everyone asks the same AI the same questions, the person who thinks of a different question becomes rare. Just as there are songs people seek out even when 75,000 new tracks land on Deezer every day.

But that different question rarely comes while watching things other people made. It comes from your own empty time, from time spent building something with your own hands. I think that is why the notebook line said "devote" rather than "find" time. Not something you do with leftover time, but time you hand over on purpose by taking other things away.

Today I kept my phone face down while writing this. That was the limited, special stretch of time I devoted today.

That is how *A Page for Today* works: it starts from one line in the notebook and arrives at one page of today's thinking.