Hey guys, Non here, coffee in hand. Want to hear some interesting stories?
Want to hear some interesting stories? Let’s start with a death. Zhu Rongji, the man who cleared the cobwebs out of the Chinese economy when he was finance minister, died yesterday at ninety-seven. That matters because the country he helped open to the world is now tightening the door, and the people left inside are wondering if they’re locked out for good.
After that, a weird thing for sale in Tokyo—people paying strangers to keep their food cold. Then, I’ll look at what happens when AI starts doing the job of the middle-class engineer. Finally, the artist Betye Saar, who taught us that trash is just a thing with the wrong address. Let’s go into it.
You are looking at a piece of history that is already being wrapped in the language of a funeral, but the only thing dying here is a system that ran on a different set of rules. Zhu Rongji, the man who turned China into the workshop of the world, has died at ninety-seven. He was the premier from nineteen-ninety-eight to two thousand and three, and he is the reason you can buy a toy or a phone today with a brand name stamped on the side, rather than a hammer and sickle. He got China into the World Trade Organization in two thousand and two, he moved power over local taxes to Beijing, he shut down the factories that did not work and sold off the state firms that were bleeding money, and he encouraged people to buy their own houses.
He is remembered for double-digit growth, for calling corrupt officials "tyrants" and for telling the country that if they did not fix their banks and their dams, the Yangtze River would swallow them whole. The official line in Beijing calls him a "revolutionary" and a "fighter"—a loyal Communist who quietly made the Party richer while the country got bigger. That is the headline, and it is tidy. It frames Zhu as a technocrat who saved the Party by fixing the economy.
But the more interesting story is the one nobody in the headlines is explaining right now: that he never stopped the Party from ruling, he just handed it a much bigger piggy bank to do it with. He is a technocrat who spent his career turning the Party into a massive landlord and a massive conglomerate, so it could pay for its own survival without asking the people for permission every time. He did not liberalise the political system, he simply stripped the provinces of their money and their power, then told them they had to perform or lose their jobs, which is a way to keep a dictatorship stable that does not require elections, only growth. He is praised for opening the country to trade, but he did it by putting a leash on the local bosses who ran their regions like fiefdoms, forcing them to answer to Beijing and to the world market at the same time.
He is celebrated for cutting the state sector, but he left the Party itself untouched, letting it own the assets he sold off. There is a real case to be made that he was a necessary evil, and that without his pragmatism the country might have collapsed into chaos earlier. But if you want to see what he actually built, look at the two million people who lost their jobs in state firms between nineteen-ninety-eight and two thousand and three. He was willing to break the social contract of the factory floor to buy the Party a few decades of growth.
We do not have a clean number for the inequality he unleashed, but if you look at the wealth gap in China today, you are looking at the shadow side of his reforms. He died as a hero of the state, but he left behind a system where the Party owns the assets, the provinces fight over the crumbs, and the people who actually built the economy are just renters in their own country.
We just looked at how Zhu Rongji helped build a state-owned economic powerhouse in China, but while that story is about the macro-management of national assets, this next one is about the very micro-level physics of survival—specifically, how people are trying to stay alive when the climate shifts from bearable to dangerous. We are moving from the halls of power in Beijing to the sidewalk heat of Tokyo. In the last two months alone, more than 53,000 people in Japan have been treated for heatstroke, and 79 have died. To address this, a company has begun rolling out units called Do Hiemon, which roughly translates to "The Cooling Box," into public spaces and selling them to businesses.
These are essentially "human fridges"—enclosed spaces where people can step inside to drop their body temperature rapidly. To understand why these are appearing now, you have to look at the limitations of traditional cooling. In a city like Tokyo, air conditioning is the standard for indoors, but there is no indoor equivalent for a person walking down a street or standing in a train station during a peak heatwave. If you are outside, your body is constantly fighting to dump heat into an environment that is already hotter than your skin.
When that equilibrium fails, your core temperature rises until your organs begin to struggle. The Do Hiemon works by creating a localized, high-intensity cooling zone. It was inspired by freezer vending machines—the kind you see in convenience stores where the glass stays cold and the interior stays chilled even when the door is constantly opening. By applying that same logic to human scale, the company is trying to create a "thermal sanctuary." It’s not just about blowing cold air on someone; it's about moving them into a different environment entirely for a few minutes to reset their internal thermostat.
The price tag for this survival reflects its niche. Each unit costs 1.5 million yen, which is roughly 9,407 dollars or 6,973 pounds. That is a significant investment for a piece of public furniture, and it suggests that the solution isn't just about "beating the heat" as a minor inconvenience. It’s an infrastructure response to a systemic problem.
You might be right to wonder if this is just a high-tech band-aid. A critic could easily argue that spending nearly 10,000 dollars on a single "cooling box" is a waste of resources when the real issue is urban heat islands—the way asphalt and concrete trap heat in cities. They would say it’s far more efficient to plant trees, use reflective在 roofing, or improve city-wide ventilation rather than building individual pods for people to duck into like they're hiding from a storm. There is a valid argument that these boxes are a reactive luxury for those who can afford them, rather than a proactive fix for the environment as a whole.
However, there is a gap between long-term urban planning and immediate mortality. You can plant a forest to cool a city, but that takes decades of growth and millions in municipal budget; you cannot tell a person currently suffering from heatstroke to wait twenty years for the shade of an oak tree. The Do Hiemon fills the space where policy has failed to keep up with the speed of the thermometer. It’s about the "now." The more important question is what happens when these boxes become standard.
If 53,000 people are being treated for heatstroke in just eight weeks, we aren't looking at a bad summer; we are looking at a shift in the baseline of human geography. When a business pays 1.5 million yen for a box, they aren't buying an amenity like a coffee machine; they are paying for the ability to keep their employees conscious and functional. It turns "coolness" into a commodity with a literal price point. I don't know if these boxes will eventually replace traditional shade or if they will remain a desperate, expensive emergency measure for the hottest hours of the day.
But for now, they are the newest architecture of survival in a warming world.
Yesterday we looked at Japanese businesses selling "human fridges" to keep staff awake, an architecture of survival built on the assumption that your body is just a resource to be managed. Today I want to talk about a different architecture: the one replacing the person who used to check that resource wasn't being overused. The story comes from a blog post written in 2026 that found its way onto Hacker News, where people are discussing the death of the middle-class software engineer. The premise is simple but brutal.
According to the author, someone who used to be a senior developer now opens their computer on a normal Monday morning and finds 7 pull requests waiting for review. One of them has changed twenty-four thousand lines of code and deleted almost four thousand, all generated by an AI assistant. The team has rewritten more of the system in a single day than they used to produce during a two-week holiday. Here is the thing.
For a long time, software had a natural friction. Writing code is hard. It is expensive and slow. That friction acted as a filter.
You could not just slap a half-broken feature together and ship it, because the cost of getting it right, of testing and reviewing the work, meant that only projects with a real value could survive. That friction protected the middle class—the people who could read code, make it cleaner, and translate between what a business needed and what a computer could do. It meant there was a point where a project became too cheap and too risky, and management would say no. Now, that friction is gone.
As one commenter on the thread puts it, bad engineers can now amplify their bad engineering by ten times. They do not need to understand the system to add a service, or denormalise a database, or wrap a serverless function around something that already works. They type a prompt, wait for the code, and open a pull request. To the untrained eye, the work looks finished.
It works. So they keep going, layer after layer of logic that no one can hold in their head, a system built like a luxury car purchased on a credit card. The car looks great, and you can drive it, but the debt is invisible until the wheels fall off. Someone will say that the code quality standards have always been low, and that bad actors were always a liability.
That is fair enough. But there is a difference between a messy codebase maintained by people who understand the mess, and a codebase maintained by people who simply paste it in. The tragedy, as the article describes it, is not that the code is bad, but that it is easy. It has become so convoluted that fixing the mess would require a colossal amount of work that management cannot justify paying for, and so the team just opens the next pull request instead.
This is not just about code. It is about the cost of pretending that a machine can do the thinking part of a job. When you remove the person who knows why the system exists, you do not just get cheaper code; you get an architecture that cannot be understood, and therefore cannot be fixed. And unlike the human fridges, you cannot put this one back in the box.
The software that used to run the world and the software that now wants to run the world have a shared problem: neither has any idea how much the other costs until the bill arrives. The previous story ended with a prediction about what happens when you automate something that cannot be put back in a box: the cost explodes, and you are stuck holding the container. The same explosion is happening right now in a much quieter room, where companies are trying to figure out how to pay for the tokens that flow through their systems like water through a pipe with no meter. The problem is not that AI is too expensive; it is that the price of a token has dropped so fast and the volume of tokens has grown so wild that the old contracts and budget lines are useless the moment someone presses send.
Microsoft engineers were recently caught using third-party coding tools and burning through their company’s annual AI budget in a matter of months. Uber faced the same shock earlier this year. The explanation is simple: when you ask a large language model to write code, you do not pay for the code you get; you pay for the tokens the model burns to figure out what to write. A prompt is chopped into mathematical chunks called tokens, and every time you tweak the wording, the model recalculates.
It does not remember what it said five seconds ago, and it does not know which version of itself is cheaper. Goldman Sachs analysts expect that the number of tokens consumed by businesses will grow twenty-four times between 2026 and 2030, reaching one hundred and twenty quadrillion tokens a month, as companies stop asking models for answers and start letting them make decisions. Someone will say that the cost is only a matter of scale and that companies should just budget for it. That misses the point.
A flat fee or a one-year contract worked for enterprise software because the software behaved like a machine: it did not improvise, it did not hallucinate, and it did not change its mind mid-job. AI agents, however, work by trying things, checking themselves, and trying again, and every attempt costs money. That is why Simon Gooch of Saviynt says tying customers into a twelve-month cost model “doesn’t make any sense honestly” — it is like agreeing to pay a mechanic by the hour without knowing how long the repair will take. There is also the practical matter of prompts.
You would not send someone to the supermarket without a list; Rob Steele at iplicit says companies need to learn to be just as precise with the instructions they feed to AI, or they will be billed for the groceries the model decided to buy on its own. The survivor here is the person who stops treating AI like electricity, which you either use or you do not, and starts treating it like a service where every request is a negotiation. The bigger companies are currently flying under the radar by using personal accounts with flat fees to keep costs low, but Oliver King-Smith of smartR AI predicts the platforms will clamp down once shareholders demand profit. The gap between the software that does the work and the money that pays for the work is widening faster than the AI itself.
Before the last story: if you are getting something out of this, subscribe. The World in Twenty Minutes is on Spotify, on Apple Podcasts, on iHeartRadio, and anywhere else you already listen. It is free, it lands every morning, and subscribing is the whole reason it keeps finding people.
The tokenomics segment ended with a widening gap between the work AI does and the money it pays for. That gap is an economic design problem. What Betye Saar was looking at was an aesthetic design problem, and the gap she was filling was in the mind. The Economist reports that the assemblage artist who challenged black stereotypes died on July 22nd, aged 99.
Betye Saar believed thrown-out objects should have a second life. That is all I have. The full obituary is behind a paywall, so I am not going to reach for the usual biographical anchors—the childhood, the galleries, the specific title of a painting. I won’t name a material or a colleague or a technique, and I certainly won’t invent a last word she might have said.
I will only work with that one line and what is inside it. We are constantly told to throw things away. The economy depends on us consuming, using, and discarding. When you are an institution, you treat people and objects the same way.
You hire them, they do the work, and if their utility drops, you move on. Betye Saar looked at that logic and refused it. She looked at the garbage, the trash, the things cast aside by a society that optimises for fresh, new arrivals, and she asked what they were carrying. She treated the broken and the discarded as if they still had something to say, and she put them in a room and forced the visitor to listen to it.
Someone will say that this is just a nice philosophy, a gentle aesthetic, nothing to do with the hard, broken structures of money and power. Perhaps. But when you look at the price of a new tool versus the effort it takes to repurpose an old one, you see exactly the same shortcut. It is always cheaper to build the new thing than to fix what is already broken.
It is always easier to replace the person than to find a way to make the old skill useful again. Saar did the hard work of the repair where many of us are comfortable with the convenience of the replacement. She kept things alive long past their official expiration date. I do not know the specific objects she collected, and I do not know the exact form of the stereotypes she challenged.
I only know that she looked at the bin and asked what was still valuable in it. The answer she gave was that there is always something left over, if you are willing to wait for it to speak. Thanks for listening.