Hey guys, Non here, coffee in hand. Want to hear some interesting stories?
Let's start with what happened on Tuesday in the White House. According to reporting from the New York Times, President Trump announced a fifteen percent tariff on a specific chemical used to make computer chips. The goal of this is to put pressure on China, which is the main producer of this material. It matters today because computer chips are the wiring of the modern world, and tariffs are a blunt way of reshaping the global economy. Before we get to that, we’ll look at the tension inside big tech, where executives keep promising that artificial intelligence will make work disappear, while the people writing the code are working up to ninety hours a week. We’ll also look at Nvidia and the trust issues around AI optimism, and finally, an obituary for the artist Betye Saar, who believed that things we throw away should always have a second life. To put it simply, the economy is shifting from a world of cheap goods to a world of scarce, expensive chips, and the price you pay for that will show up on your next phone bill.
The headlines this morning are all shouting about trade wars. You are being told that President Trump has just slapped a 15% tariff on polysilicon, a specific chemical crystal used to make computer chips and solar panels. The narrative you are hearing is simple: the United States is flexing its muscles to stop China from winning the race for technology, and this new tax is the weapon in that fight. The story is being sold as the start of a hot new cold war over silicon. That is not the thing that happened. The thing that actually happened is that a man in the White House looked at a spreadsheet, saw a number, and decided to change it. According to The BBC, Trump signed an executive order on Thursday that raises the price of polysilicon by 15%. This is not about ideology; this is about supply and demand. Polysilicon is a raw material. Before this order, companies in the US could buy it cheaper from factories in China or Europe. Now, the government says they must pay more. To explain why the price matters, picture a grocery store. You usually pay a fixed price for milk. If the store manager decides that the price of milk will now be a minimum amount—let’s say, if the market price drops below five dollars, the store must charge five—the result is the same regardless of whether you call it a tariff or a floor price: the milk costs more. In this case, the minimum price is set to protect Hemlock Semiconductor and Wacker Chemie, which are the two main producers of this material inside the United States. Fine, you might say. If the US wants to make its own chips, that makes sense. The obvious objection is that this is just protectionism. If the market can’t compete, why force it to? Why not let the free market decide if American factories are good enough? That is a fair point. Protectionism does not make a company better at what it does; it just makes the product more expensive. The argument here is that the US has already been subsidising this industry for decades, and now it is just trying to stop the money from leaking out. The government claims that imports have caused the US share of global production to drop from 50% in 2005 to less than 2% in 2024. They say that losing control of the raw material means losing control of the military equipment and the electronics that run the country. Here is the part that is usually left out. The move comes at the exact same time that Washington and Beijing have been holding their "trade war" talks on and off since May 2025. The tariff was announced while those negotiations were technically still live. It is not just a trade dispute; it is a mechanism to give the US leverage. By making polysilicon more expensive, Washington can demand that China lower its prices on other goods in return. Beijing has already called it "abusing state power," and they have threatened countermeasures, including tighter controls on drone exports and a review of imported printers. The more important question is what this costs the average person. Polysilicon is not a finished chip; it is the sand-like stuff before it is melted into wafers. A 15% tax on the raw material does not immediately mean a 15% tax on your iPhone. But if the cost of the sand goes up, the cost of the factory goes up, and eventually, the price of the device in your hand does too. We already know that tariffs work by making things more expensive; today, the Department of Commerce is simply picking which things get the price hike first.
The story before this one was about the Department of Commerce picking a list of materials to make more expensive. That is a blunt, bureaucratic decision: you draw a line on a spreadsheet and suddenly a widget that was cheap for the last twenty years costs more. It is a way of managing trade, but it relies on believing the people holding the spreadsheet know what the price should be. Today we are looking at a different kind of list—one that is not written on paper, but written on a schedule. It is a list of hours, and it is just as real a line in the sand. For years now, executives at companies that are pouring hundreds of billions of dollars a year into developing various artificial intelligence tools have insisted that the technology will ultimately mean people will spend less of their time working. An engineering director at Google said four years ago that AI would deliver a four-day work week. Earlier this year, and just one year after that engineering director's prediction, OpenAI took up the challenge, in a manner of speaking. It formally urged companies to start testing out a four-day work week with no change in pay, claiming that AI will soon be able to speed up so much human labour that the corporate world should prepare itself. However, a former OpenAI technical employee who left the company last year told the BBC the firm never actually trialled the four-day work week it suggested others should try while they were there. Instead, the person described what was often a gruelling work culture marked by frequent "crisis meetings", working on weekends, and "super cut-throat" performance reviews that would see colleagues suddenly let go. To put it simply, executives are selling a vision of efficiency—a world where a digital assistant handles the drudgery, freeing you to do the work that matters. That is the pitch. But the people actually building these assistants and using them are not experiencing that vision. They are experiencing the opposite. The incentive, the constraint, the money all push in the same direction. These companies are not modelling their own claims of the technology giving people more free time. They are modelling growth. " So the constraint is not the physical limits of a human being, but the optics of a quarterly earnings report. The rule is that you cannot promise the public you are using the tool to reduce hours, because that reduces the revenue per hour. You promise the public you are using the tool to increase productivity, which justifies higher stock prices. The money flows to the person who can show the biggest output, not the person who can show the best life balance. The chain of consequence is clear: you get better software, the software gets better, and the expectation for the person sitting in front of it gets higher. You do not get a shorter week; you get a faster week. Someone will say, perhaps, that this is just the tech sector being the tech sector—famously hyper-competitive, famously chaotic—and that the 14-hour days are a choice these workers are making for the chance to cash in on the AI gold rush. You might be right. It is easy to look at a former OpenAI employee who is now at a start-up working closer to 50 or 60 hours a week and see ambition. It is easy to see that as a voluntary trade-off for equity that might, one day, be worth something. And if you look at Meta, you see a company where workers are being "drafted" onto AI teams, moved without their consent and told that the work is urgent, with little room to say no. That feels different. That feels like the system tightening around them. That feels like the promise that AI would free them has been repurposed as a mechanism to push them harder. The former OpenAI employee described working "sprints" that stretch for many weeks and top 90 hours of work in a seven-day period. At Meta, workers said they have been abruptly pushed onto teams doing AI work being treated as urgent. They called it being "drafted", according to one current and one former employee, because people were not given a choice. "They just move you over," the former employee said. " The survival of the argument for less work is the thing that survives. In the case of OpenAI, the four-day week was a suggestion for the rest of the world, never a trial for themselves. In the case of Meta, strict limits were only introduced after the drafting was done. The survival of the argument for less work is the thing that survives. It survives in press releases. It survives in the words of engineers who have left, saying their old jobs were 9-to-5 and this one is a marathon. It survives in the specific, ugly reality of the 90-hour sprint. We do not have a clean number for the entire tech industry because reporting on hours is patchy, but we do know this: US tech workers are averaging 50 hours a week, with some hitting 90. We also know that the companies promising a four-day work week are the same ones demanding these sprints. The gap between the schedule on the corporate website and the schedule in the Slack channel is the real story. The Department of Commerce picked a list of materials to make more expensive today because they believe the market will adjust to that price. They made a choice about what matters most right now. The tech industry is doing the same thing, but they are choosing to make people more expensive to employ, by making them work faster and harder. They have decided that the person is the bottleneck, not the silicon. They are betting that the scarcity is not in the computer, but in the human attention span. The more they invest, the less they value the time it takes to use it.
The last story ended on a complaint about burnout, and this one is about the machine that is making the hours longer. The link is not the software itself, which is a mess, but the fact that the machine is paid to build more of itself. According to Stratechery, Nvidia’s advantage has never been a secret chip, but a lock on the code that runs on it. Researchers use CUDA, Nvidia’s own version of C++ that pretends to be regular computer language. The problem, as a developer on Hacker News pointed out, is that if you try to do something slightly off the standard path, the compiler says nothing and the code simply fails at runtime. You are debugging a machine you cannot see, in a language you do not actually understand, and the toolchain is built to protect Nvidia, not you. That is where the second-order assumption lives. Everyone agrees that AI needs more computers. That is the easy thing to predict. The harder thing to predict is that demand will grow fast enough to justify the cost of the software ecosystem that supports those machines. When you buy a piece of hardware, you expect it to pay for itself. When you build a factory, you know how long it takes to fill the building with products. In AI, you are asked to invest in a machine that is only useful if other people are building machines for it. You fund the growth of the ecosystem that funds you. Someone will say that this is just how the internet worked: Netscape sold the browser to sell the server, and Google sold ads to buy more servers, and the whole thing grew until the model broke. That is a fair comparison, but this time the loop is tighter. You are not just selling the tool; you are selling the language the tool speaks. If the language changes, the tool becomes worthless. If the language does not change, the tool eats the world, and you pay for the seat at the table. Microsoft CEO Satya Nadella has been reading the history books. He pointed to the Panic of 1873, a railroad crash that started in a similar way: investors were told that land and track were the future, and they poured money into a system that was designed to absorb capital, not to move freight. S. investors poured into railway bonds every year in the early 1870s is, today, roughly the amount the major tech companies plan to invest in data centers. The math fits the panic. The investment thesis is the same: build now, grow later, hope the market follows. The real difference, if you look closely, is that Jay Cooke used patriotism and newspaper ads to sell the bonds. Today, you use open-source communities and the promise of a singularity to sell the chips. The incentives are the same, but the pitch is quieter. You do not need a army of salespeople if you can convince the engineers to write the code for you. Nvidia will keep selling the hardware, and the hype will keep selling the software, until a cheaper way to run the code is found. The more they invest, the less they value the time it takes to use it.
The people building the chips are rushing because they think they have to, and the people selling the software are rushing because they are trying to keep up. That was the previous story, about Nvidia, and the rush to use what they have not yet figured out how to run. Today, the rush is different. It is the rush to believe that something wonderful is coming, even though the people promising it are the same ones who just spent years convincing the public they were going to ruin everything. According to The AI Daily Brief, Mark Zuckerberg is making the AI industry’s most aggressive case yet for optimism. He is talking about a future where the technology is open, where it helps people do their jobs, where governments and companies share responsibility for how it works. It is the exact opposite of the dystopian vision that has dominated headlines for the last two years. The problem is not the ideas; the problem is that the messenger is a billionaire who sold us our phones and then our social feeds, and who now wants to sell us the future on the promise that he has changed. He is putting up a billion dollars for a community fund and releasing a new open model, but he is also the same man who was at the centre of the data-centre debate we just left. The public is not listening because they do not believe the optimism is real; they believe it is another layer of marketing. Perhaps there is a real case that open models are the only way to stop a handful of companies from controlling every thought that can be digitized, and that a billion dollars might actually help small communities build the tools they need rather than just buying seats on a private train. Perhaps the people who spent years yelling about safety were right to be scared, and this is the only way to let the rest of the world catch up. Someone else will say that this is just a rebrand, that a billion dollars is pocket change to a company worth a trillion, and that the "open" model will just be the same model with a few weights stripped out so it can be controlled by whoever has the most lawyers. There is a case to be made that Zuckerberg is trying to buy a second chance at being liked, just as he did with his metaverse experiment. The thing that survives both readings is the gap between what the companies measure in their own spreadsheets and what the person in the queue at the DMV actually feels. The companies are measuring lines of code and dollars of investment; the public is measuring whether their data is being stolen, whether their children are being trained on their private lives, and whether the people in charge will be in the room when the system breaks. We do not yet have a number for how much of that trust is recoverable, but we do know that you cannot buy it back with a press release and a community grant. The more they talk about optimism, the more it sounds like they are trying to convince themselves first.
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 optimism crowd talks about trust like it’s a battery you can recharge, but Betye Saar was working with a different kind of material: the refuse of other people’s lives. She believed thrown-out objects should have a second life. According to The Economist, the assemblage artist who challenged black stereotypes died on July 22nd, aged 99. That is the line I have. The full obituary is behind a paywall, so I cannot tell you about the specific house in Los Angeles where she worked, or the years she spent saving junk from the curb, or the exact year she started gluing a doll and a broom together. I do not know if she ever said the words I am about to put in her mouth. I only have the idea that anchors this story: that what is discarded is not finished, it is just waiting for the right person to pick it up and give it a purpose. Someone will say that is romantic. They will point out that you cannot fix the world by cleaning up after it. They will argue that Saar’s method—taking trash and remaking it into meaning—is a kind of comfort for people who do not want to confront the damage being done right now. That is a real case to make. But even if you reject her method, you have to admit what she was refusing. She was refusing the idea that the broken things are gone for good. She was refusing the idea that we are limited to what we were handed, that we have to accept the stereotypes and the scraps as the only options available. She was betting that a broom and a doll and a bit of broken mirror could do the work a statue could not. That is a very different kind of optimism than the one the optimists are selling you today. They are selling you a future that is technically possible; she was offering you a future that is morally necessary. Here is the thing: if you are tired of people telling you that optimism is a strategy, just remember that someone spent her life turning a literal pile of junk into a statement about who she was. The more you talk about building a better world, the more it sounds like they are trying to convince themselves first. The more you look at what is left on the curb, the more it sounds like the world is still waiting for the right person to show up.