Hey guys, Non here, coffee in hand. Want to hear some interesting stories?
Did you know that on August 19th, the United States declared that fresh sanctions would “squash” Iran’s economy and “collapse” its regime? This is the one that matters because, as an architect, I look at infrastructure and see how long systems last; economic infrastructure is built on trust and trade, and if you pull the plug on the energy and shipping routes it relies on, you aren't just raising prices, you are asking a society to hold its breath for years. Before we go there, we start with a frantic list of nine AI techniques from yesterday—some you’ve heard of, some you definitely haven't. Then we’ll look at Aaron Swartz, the scraping laws, a piece of critical tech going dark, and finally a quiet obituary for Mary de Rachewiltz.
Let’s get into the list.
According to the AI Daily Brief, the landscape of artificial intelligence is moving away from simple chatbots toward complex, persistent systems. Specifically, companies like Replit have launched Free Mode using the GPT-5.6 Luna model, while xAI has released Grok 4.6, which is positioned as a faster and dramatically cheaper alternative to the current market leaders. At the same time, we are seeing the rise of "agentic" workflows, where platforms like Hyperagent offer a fleet of always-on agents, including a $1,000 inference credit for new users to get started. To understand why this matters for your daily life, you have to look at the shift from "prompting" to "workflow teaching." In the past, if you wanted to use AI, you had to know exactly how to ask a question to get a good answer.
Now, the goal is to teach the AI a routine—like how to manage a specific set of files, respond to a certain type of customer email, or coordinate a team of smaller, specialized bots to complete a project. It is moving from a tool you use, like a hammer, to a junior employee you train, like an intern. The incentive for companies is simple: if you can automate a multi-step process rather than just a single sentence, you save on the most expensive cost of all, which is human time. The popular narrative right now, the one you see in the headlines and on social media, is that AI is becoming "too smart" or "too dangerous," leading to a wave of fear about what these models might do on their own.
People are worried about the "sentience" of these agents or the total replacement of the human worker. Fair enough. If an AI can now perform a sequence of tasks without a human clicking "next" every five seconds, it feels like a loss of agency. But the reality of the data suggests a different story.
The real bottleneck isn't the AI's intelligence; it's the reliability and the cost of "inference"—the actual computing power required to run these thoughts. For instance, while Grok 4.6 is marketed as a "fast" and "cheap" option, its real value is in providing a more affordable path for businesses to experiment with these complex workflows without breaking the bank. The move isn't toward a god-like machine; it's toward a more efficient plumbing for digital work. We are seeing the transition from "Will it work?" to "How much will it cost to run every day?" The more important question is how we maintain human expertise when the easy way to do work becomes delegating it to a fleet of bots.
We just talked about how we are handing our mental models to fleets of algorithms. Now we have to talk about how we hand our money to fleets of ships. In AI, the question is how much it costs to run a model every day. In the Strait of Hormuz, the question is whether a choke point can be choked again when the flow is oil.
The underlying logic is the same: you pick a pressure point and push until the structure breaks. The difference is that AI fails silently—your spreadsheet is wrong, and you fix it. Economic pressure fails spectacularly—your food costs double, and the regime blames the other side. According to The Economist, US Treasury Secretary Scott Bessent is urging allies to support a plan to "squash" the Iranian economy.
He told CNBC that the goal is to "collapse" the regime. This comes after President Donald Trump said earlier this year that countries providing "any type of lifeline" to the Islamic Republic would face consequences. Bessent confirmed the new measures will be detailed at a press conference on August 24, but he did not say if countries like China could be targeted. Let's unpack the mechanics here.
The US and Israel started a war in February, arguing the strikes were to stop Iran from building nuclear weapons—though Tehran has consistently denied it. Iran responded by hitting Israel, US bases across the Middle East, and Gulf Arab allies. Since then, Iran has largely closed the Strait of Hormuz—the narrow waterway through which about 20% of the world’s oil and liquefied natural gas used to transit. That caused wild fluctuations in oil prices, which hit every petrol station in every country that buys fuel.
The US and Iran agreed a ceasefire in April, then another one in June. Both were supposed to facilitate negotiations, but the truce has been broken repeatedly, and talks have stalled. Instead of returning to war, Trump is now gearing up to unleash the "most crushing economic operation ever taken against any country." Bessent said this new economic pressure campaign would minimise the need for a "large-scale kinetic restart" of the conflict. Operation Economic Fury was announced by the US Treasury and the Pentagon.
It includes a naval blockade aimed at stopping Iranian exports and sanctions on foreign banks and firms doing business with Tehran. The idea is to degrade Tehran's warfighting capability by cutting off its revenue streams. In April, the US launched a wave of sanctions after military operations did not produce the overthrow of the regime or surrender that Trump had hoped for. The Iranian economy has long been struggling.
Official figures show that in the 12 months to February 2026, inflation was astronomical, while food prices doubled. Everyday Iranians blamed the regime for their difficulties. At the time, Trump told demonstrators "help is on the way" and urged them to "keep protesting," an appeal that seemed to have little to do with the diplomatic solution and much to do with regime change by other means. Someone will say: if the goal is a ceasefire, why are we talking about crushing the economy?
Why not just enforce the truce? There is a real case that the economic pressure is not a backup plan, but the actual plan. If you can collapse the regime without firing another shot, you save lives, you save equipment, and you make a political point that the US can win without casualties. From the American perspective, it looks like a win-win: you get the outcome you wanted (regime change) with less cost than the alternative (war).
The problem is that this assumes the Iranian state and the Iranian society are the same thing. The more important question is whether a state that is already operating in a semi-peripheral economy can be pushed further into the peripheral zone without collapsing into something else. Iran is no stranger to sanctions. A raft of them were first imposed shortly after the Islamic Revolution of 1979, during the hostage crisis.
More have been added since then. The regime has not collapsed. In fact, sanctions have often succeeded in controlling dissent. The government recently cracked down on protesters earlier this year.
A resident in the central Iranian province of Khuzestan told the BBC that a country with a long experience with sanctions will be able to overcome additional pressure. "There may be an impact at first, but that doesn't necessarily mean Iran won't be able to overcome the crisis," he said. The difference between now and 1979 is the global structure of trade. In 1979, the world was more fragmented.
The US had more leverage because the rest of the world was either part of the US-led Bretton Woods system or dependent on American markets. Today, the system is more multipolar. China and Russia have different incentives, and they are willing to trade with Tehran even when the US puts pressure on banks. Scott Bessent has not said if China could be targeted, but the US has already sanctioned Chinese firms over their trade with Iran.
If they cut that off completely, it risks a broader confrontation with Beijing. The real danger of "squashing" an economy is not that it will fail to hurt the people in charge. It will hurt the people in charge, and it will hurt the people on the street. The question is whether the state can absorb that pain and redirect it.
The Iranian state has a history of redirecting economic pain into nationalist sentiment. When the economy is bad, people look to the regime to protect them from the "outside" enemy—the US, the West, Israel. That is how sanctions have historically been weaponised by authoritarian regimes to shore up their own legitimacy. If the US succeeds in breaking the Iranian economy, the result might not be a democratic uprising.
It might be a more resilient, more paranoid, more inward-looking autocracy. You buy people fish, and they learn to rely on you for fish. If you stop bringing the fish, they don't learn to fish; they just get angrier at the person who isn't bringing the fish. The US is betting that the population has already passed the point of endurance.
The figures show inflation and food prices had already doubled by February. The argument is that the cost of maintaining the status quo is now higher than the cost of changing it. But history suggests that the cost of changing a regime under external pressure is often very high, and the outcome is rarely what the external power intends. On August 24, we will find out if this "coordinated economic isolation" is a surgical strike or a scalpel dipped in acid.
The question of whether economic isolation can truly collapse a regime is essentially a question of who owns the infrastructure of life. If you can cut off a country’s ability to trade, you exert power over their survival. But there is a different kind of infrastructure at play when we talk about the data that powers our modern economy, and how the law decides who gets to own it. We see a glaring double standard in how the United States treats the "illegal" acquisition of information.
In the early 2010s, the government prosecuted Aaron Swartz, a co-creator of the RSS protocol, for downloading academic papers. He faced a staggering 35-year prison sentence and a $1 million fine because he didn't just scrape a website; he physically trespassed into a room, plugged a laptop into a router, and rotated his MAC address to bypass administrative bans. He was punished for bypassing a specific gate. Now, look at Meta.
They are currently scraping the open internet at a scale that dwarfs Swartz’s actions to train their proprietary AI models. They aren't sneaking into rooms; they are moving mountains of data across the public web to build products that generate billions in revenue. Someone will say that the law is finally catching up to the reality of AI, and that Meta's actions are a necessary evolution for technological progress. They might argue that Swartz’s actions were a criminal violation of physical security, while Meta’s are a matter of complex, ongoing copyright litigation.
But the reality is that the law is being used to protect one business model while allowing another to flourish. Swartz was punished because he disrespected the gatekeeping of academic journals. The current AI boom is allowed to proceed because the government is currently unwilling to risk the economic implications of limiting the very companies it wants to lead in the AI race. The difference is that Swartz wanted to disseminate knowledge, while Meta wants to own the means of its reproduction.
The gap is the price of the data. We are moving from a world where information was guarded by physical locks to one where it is harvested by algorithms, and the law is currently choosing which side of that shift to enforce.
We were just talking about the fact that people used to protect information with physical locks, but today they harvest it with algorithms, and the law is struggling to decide which protection is worth paying for. That story was about theft of data, but the next one is about theft of physical space. It is the same instinct. When the value of something goes up, you stop leaving it on the sidewalk and you start trying to bury it.
In the case of electricity and the internet, the sidewalk is the surface of the earth, and right now, the price of leaving things there has gone up. Russia’s war in Ukraine has done the same thing for infrastructure that it did for data: it revealed how vulnerable exposed facilities are. Drones can hit a power substation just as easily as a server rack, and the cost of putting those things underground is no longer theoretical. A company called EarthGrid just tested a boring machine in a California quarry.
They use three plasma torches spinning at the front of the machine, heated to 27,000 degrees Celsius—hotter than the surface of the Sun—to chew through ultra-hard granite. In January, their machine bored through three metres of rock in a single go. Troy Helming, the founder, calls it creating a tornado inside the tunnel that sucks away the debris, sometimes looking like molten lava. It is loud, it is violent, and it is exactly the kind of brute force you need when you are trying to move things through material that was designed not to be moved.
Engineering firms are now reporting that they are seeing more people ask for this, but it has always been an expensive and technically difficult way to build. You do not put a data centre or a power substation underground unless you really have to. The obvious objection is that this is madness on paper. It is hideously expensive to dig a tunnel and then line it properly, and it is slow.
While you are digging, the power is out and the internet is down. You might argue that it is cheaper just to build a roof, put cameras on the perimeter, and wait for the police. That is what people have done for a century. You look at a power plant or a phone tower and you assume it will be fine as long as you have enough guards.
Here is what survives the comparison. The cost of not doing it keeps rising. We are seeing natural caverns in northern Italy used as data centres because the dolomite rock keeps the servers cool without air conditioning. Dennis Bonn, the chief executive of Trentino DataMine, says that ninety million cubic metres of rock provide protection against physical intrusion, earthquakes, and water.
It is a fortress that pays for itself by not having to be cooled. The more you look at it, the more it seems like the modern way to own things is not to guard them, but to hide them. The old world was about visibility: we built towers so we could see for miles. The new world is about invisibility, and it is harder to attack something you cannot see.
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.
We have spent the last few minutes looking at how the most sensitive pieces of our modern infrastructure are retreating into the shadows to escape scrutiny and attack. It is a strategy of physical and digital concealment. But if we move from the architecture of hidden servers to the architecture of human devotion, we find a different kind of invisibility. It is the deliberate choice to stand behind a person or an idea that the rest of the world has already decided to leave behind.
According to The Economist, Mary de Rachewiltz, who devoted her life to defending Ezra Pound, has died. That is the only information I have. I don't know what she looked like, what she did for a living, or what she said in her final hours. I only know the specific gravity of her commitment.
When we talk about a life "devoted" to a defense, we are talking about the most expensive thing a human being can spend: time. To defend someone, especially someone who has been cast out, requires a refusal to let the public consensus dictate your private reality. It is the opposite of the "invisibility" of technology we just discussed. While a server hides to survive, a person who defends a pariah often stands in the open, choosing to be seen with something that others find radioactive.
Someone might say that such a devotion is a waste of a century of life, a stubborn clinging to a ghost. But there is a reality to that choice that survives the judgment. It suggests that for some, the truth of an individual's worth is not a democracy. It is a private, immovable fact.
We live in a world that wants to optimize our loyalties, but Mary de Rachewiltz lived in the gap where loyalty doesn't have to make sense to anyone else. I don't know what she defended, but I know what it felt like to do it. Goodnight.