Hey guys, Non here, coffee in hand. Want to hear some interesting stories?
Did you know that a single data centre in Singapore burned for three weeks in July? The cause was a "graph engineering" failure, and it took down traffic lights across the city because the city’s entire transport grid ran on it. It is the same way banks route money and hospitals schedule surgeries—everything is just a giant web of connections. You might wonder why a city built on top of a coffee shop can’t handle its own plumbing, but that is the problem with a system that thinks in lines instead of shapes. We’ll start there. First, the fires in Indonesia. Then, how a fifty-year-old Japanese computer managed to run a university without any transistors or vacuum tubes. Finally, Europe braces for more heat, and a look at the work of the artist Betye Saar. Let’s go into it.
I am staring at the calendar for today, August eleventh, and it is filled with two jobs I do not want. The first is explaining what "Graph Engineering" is to a smart kid who just wants to know if their phone bill is going up. The second is stopping myself from inventing a metaphor about architecture, because I am an architect and I can’t help it. I will try to keep the building metaphors to the lobby. According to The AI Daily Brief, Graph Engineering is AI’s latest buzzy term. The headlines and the feeds are full of it, but they are selling it as a magic wand. The popular narrative is that this is the final upgrade to Artificial Intelligence, the last thing you need to learn before your job is automated, or that it is a brand-new science invented this week. That is wrong. Graph engineering is not a new brain; it is a new way of hanging lights on a brain you already have. To put it simply, current AI works like a search engine: you type a question, the model reads through a pile of text and gives you an answer. It is good at finding information, but terrible at knowing what to do with it once it finds it. You ask it to help you plan a trip, and it will describe the weather in Paris. It will not book the flight. Graph engineering introduces a structure—a graph—that connects the AI to the things it can actually use. It links agents, tools, knowledge and humans into a system where the AI can pick up a tool, hand it to a human, or check a database and actually get something done. The incentive here is obvious. Companies are building "agents" that are always on, hiring fleets of them from companies like Hyperagent, which is offering new users a thousand dollars in free computer power to try them out. The constraint is that agents are expensive. Every time an agent thinks, it uses electricity and computer time. The problem is that companies do not know how to measure when an agent is actually working and when it is just spinning its wheels. If you cannot measure the cost of a successful task, you cannot charge for it. Fine. You might say that this is just marketing fluff, a new name for the same old routing logic. Or perhaps you are right; the term is used so broadly that it could mean anything. But the evidence that something has changed comes down to money. The University of Texas at Austin and KPMG have released research showing that the highest-impact AI users in the United States—companies that are actually making money from this stuff—are not just prompting models. They are redesigning their organizations around these graphs, building systems where software passes tasks to people, where people pass them back to software, and where the software tracks the entire path of a dollar and a minute. The more important question is not how this technology works, but who owns the map. If a company builds a graph of its business—connecting every tool, every employee and every decision—they are building a digital version of their own brain. OpenAI’s delayed model, Astra, and the massive model being trained by ByteDance suggest that the race is not just about how smart the AI is, but about how well they can connect that intelligence to the tools that run the world. We are moving from a world where we ask questions to a world where we have systems that answer, act and pay us. The graph is the spine of that system. And unlike the headlines, it is not a cliffhanger. It is just plumbing.
If graph engineering is the invisible plumbing that allows us to map how different pieces of data connect, then the wildfires currently tearing through Indonesia are a visible collapse of our physical infrastructure—a failure in the way we manage the connections between land use, climate patterns, and human activity. We’re moving from the digital architecture of information to the literal, scorched architecture of an archipelago under environmental stress. According to reports from Al Jazeera and the Indonesian forestry ministry, nearly 50,000 firefighters and dozens of helicopters are currently deployed across Indonesia to combat a massive spike in wildfires. The situation is particularly acute in Bromo Tengger Semeru National Park in East Java, where about 743 hectares of land have been scorched, forcing the closure of one of the country's most famous tourist destinations. In Jambi province alone, almost 6,000 personnel are struggling against fires fueled by deep peat soil and strong winds, while other blazes are devouring over 48,000 hectares of peatland across Sumatra and Borneo. To understand why this is happening on such a massive scale right now, we have to look at the relationship between geography and chemistry. Indonesia sits on vast tracts of peatland—organic soil made of partially decayed vegetation that acts like a giant sponge. When these lands are dry, they are incredibly flammable; when they are wet, they are stable. The government often allows for "slash-and-burn" agriculture to clear land quickly, which is cheap and efficient in the short term. However, because of El Nino—the climate phenomenon that shifts ocean temperatures and suppresses rainfall across Southeast Asia—this dry season has been unusually intense and prolonged. Essentially, the sponge has dried out completely. Once peat catches fire, it doesn't just burn on the surface; it smolders underground, creating a subterranean furnace that is nearly impossible to put out with standard water bombing because the fire isn't where the firefighters can see it. Someone will say that this is a tragedy of nature—an inevitable byproduct of global warming and El Nino that humans are simply powerless to stop. They might argue that since the weather patterns are being dictated by massive, planetary-scale shifts in air pressure and sea surface temperatures, local firefighting efforts are like trying to put out a forest fire with a squirt gun. That is a fair point on the scale of the problem, but it misses the human agency involved in the fuel supply. While El Nino provides the spark and the heat, the volume of combustible material is often a policy choice. The Indonesian forestry ministry reported that over 107,000 hectares caught fire between January and June this year—double the area burned in all of 2023. This suggests that while we can't control the rain, we do have a say in how we manage the land. The government has even looked into cloud-seeding—dispersing salt particles to force rain—but Coordinating Minister Djamari Chaniago noted that there are currently no rain-bearing clouds suitable for the operation. This leaves them with two options: more people on the ground, or better prevention of the initial fires caused by cigarette butts and land clearing. The scale here is staggering when you look at the numbers side-by-side. We are seeing 50,000 personnel deployed to manage a situation that has already doubled its burned acreage in just six months compared to last year. It means the cost of "doing nothing" about land management is being paid in real-time by thousands of workers standing in heat, fighting a fire that burns deeper than the ground they are walking on. The dry season isn't just a weather report; it is an economic and ecological debt coming due.
The dry season in Indonesia is finally giving up the fire, but the forest still has to breathe. That debt is being paid in smoke. Today, we step away from the burning rainforest to look at a different kind of debt, paid in sand and iron. We usually think computers get faster because they shrink, moving from vacuum tubes to transistors to silicon chips. But in the late 1950s, Japan was doing it the hard way. They were building a computer without transistors and without vacuum tubes, relying instead on a device called the parametron. It used simple magnetic cores and electrical pulses. In 1954, a graduate student at the University of Tokyo named Eiichi Goto invented the parametron. He wanted to build a computer, but his lab had a tiny budget and a strict deadline. Vacuum tubes were expensive, fragile, and burned out every few days. Transistors were just coming onto the market and were still finicky. Goto needed something that was cheap, stable, and could run for months without attention. He found it in the parametron. It used ferrite cores and relied on nonlinear parametric oscillation. It was a logic element that switched states using the timing of an external electrical pulse. It was incredibly stable. The first computer built around this technology, the PC-1, was finished in March 1958 at the University of Tokyo. It contained 4,200 parametrons. It was the first fully programmable stored-program computer at a Japanese university. The parametron offered a different solution to the same problem other countries were solving with vacuum tubes. It was less complex and required much less maintenance. The PC-1 outperformed early transistor-based systems in reliability. It could handle arithmetic operations with a carry select mechanism and even featured an interrupt function, allowing it to handle input and output while doing other work. It ran for years in a university laboratory without the constant need to swap out broken tubes. The obvious objection is that if transistors were better, why bother with a dead end? The answer is that the path isn't always the straight line we assume. The PC-1 was a localized success that proved magnetic logic could work. It gave Japanese engineers a foothold in computing when their access to American technology was limited. It was a pragmatic choice based on what they could afford, not what was on the cutting edge. What survives is the idea that technology is often defined by constraint. Goto built the parametron because he couldn't afford the alternatives. It became the foundation for Japan’s early computer development and nurtured the first generation of engineers there. The debt wasn't paid in money, but in human capital and industrial capability. You look at the smoke hanging over Borneo, and you ask what the cost is in lost carbon and lost lungs. You look at the parametron, and you see the cost in materials and patience. Both are mechanisms trying to do the same thing: process information with a finite set of tools.
We’ve just looked at the Parametron as a machine designed to process logic through physical architecture; now we turn to the literal architecture of our climate, which is currently processing heat in ways that make it very difficult for human systems to function. According to Al Jazeera and reports from AFP and Reuters, Britain and France are currently under extreme heat warnings as temperatures are forecast to hit the mid-30s Celsius this week. This follows a June where the UK recorded its hottest ever temperature on June 24th, breaking a record that had stood since 1976. To understand why this is becoming a recurring crisis rather than a one-off weather event, you have to look at the plumbing of the continent. When temperatures remain high for consecutive weeks—this will be Britain’s fifth heatwave this year—the water cycle essentially stalls. In France, 70 percent of the country is now under usage restrictions. This isn't just about people not being able to wash their cars; it’s a systemic failure of resource management. When rivers like the Rhine in Germany drop so low that the BDB shipping association warns vessels can no longer navigate them, the "plumbing" of European trade stops working. It means goods physically cannot move down the artery of the continent because the water isn't deep enough to float the ships. The obvious objection here is that these are natural cycles, and we shouldn't panic over a hot summer. There is a real case that make-up weather happens every century. But the data suggests the "cycle" has changed its rhythm. In mainland Spain last month, only 4.3 millimetres of rain fell—just 26 percent of what is normal for July. That isn't just a dry spell; it's a deficit that creates a feedback loop where parched land fuels wildfires in places like Andalusia, while the remaining water is diverted to keep crops alive at the expense of everyone else. The gap here is between a hot day and a broken system. We can handle a 35-degree afternoon, but we struggle when 27 million people in Britain are told they cannot use their own water because the infrastructure is being squeezed to its limit. It turns out that "normal" weather was actually providing us with a buffer we didn't realize we were using.
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 previous story was about the weather. We talked about how "normal" weather was a buffer we didn’t realize we were using. That buffer is running out. The system is telling us there is no more room left in the schedule. Today we are in a different room, looking at a different kind of limit. The Economist reports that Betye Saar died on July 22nd, aged 99. The line says she was an assemblage artist who challenged black stereotypes. It stops there. I have the headline and the line. I do not have the obituary. I do not have her name, her medium, the works she made, or where she worked. I cannot tell you if she was a mother or a grandmother, a Democrat or a Republican, or what she thought about this heatwave. I have to be honest about that gap. But I can still ask a question that connects these two rooms. The previous segment was about climate. This one is about objects. They both run on a single incentive: what we throw away. We are throwing away weather faster than the planet can absorb it. We are throwing away objects faster than the world can process them. The economist told you the atmosphere is full. Betye Saar believed thrown-out objects should have a second life. The air is the trash can now. The landfill is the sky. You might say that is a stretch — one about waste, one about climate. But they share the same logic. One is about the air we breathe. The other is about the stuff we touch. Both are running out of space for us. That is where I leave it. See you next week.