AI weather, Campbell's soup, and quirks
Lux Recommends #464
AI weather reports, Campbell’s soup meets the trade war, and why you need to embrace your weird. Also, China’s consumer class and nuclear war.
From Lux Capital
Exciting news from Lux portfolio company Sakana. In an announcement on X, the team rolled out their transformative new method for teaching LLMs. Most current advanced models follow a two-step learning process: first, the model must learn to solve a problem from scratch and is rewarded for correct answers and second, the model trains “students” in its methods. But Sakana is flipping that around. It is tasking its model with generating clear step-by-step explanations for solving problems first. The models are rewarded not for solving the problems, but for how helpful their explanations are. The goal is smaller, more efficient models and easier training processes.
In other Lux portfolio news, Delphi, an AI startup that lets individuals create chatbots based on themselves, announced a $16 million series A round, led by Sequoia Capital.
From around the web
1. The AI in the sky
You might have missed my article earlier this week on how, without much notice, we’ve suddenly gotten much better at predicting the weather. If you did, it is worth a read, especially in light of Google’s recent jump into the fray. Wired’s Boone Ashworth reports the company wants to put 50 or more satellites into orbit to monitor the ground in real time and then use AI to better predict breakout fires in their early stages.
The FireSat operation plans to launch three more satellites in early 2026, and eventually work up to its final number of 52 satellites over the course of the next few years. At full capacity, the satellites should be able to detect a fire as small as ten square meters and then collect updates on the spread every 20 minutes or so. The goal is that the window between updates is short enough to give them the types of information first responders can actually put to good use.
Want your accurate weather predictions without all the fuss and dramatization of today’s weather television? Our scientist-in-residence Sam Arbesman directs you to Weatherstar by Matt Walsh. The project “aims to bring back the feel of the 90s with a weather forecast,” Walsh writes, “but available in a modern way.”
Finally, weather balloons can help provide better data for modern predictions. But they may also help solve our heat problems — at least according to this cartoon Laurence found in XKCD.
2. The great mall of China
When diagnosing the Chinese economy, economists typically put a lot of weight on consumer underconsumption. Although the proportion of consumption to investment does look low by international standards, according to Ruchir Sharma underconsumption is largely a myth. He explains the real story in the FT. H/t editor Katie Salam.
So far this century, in real terms, private consumer spending in China has grown more than 8 per cent a year, faster than in any other economy — by far. Over the past few years, consumer spending growth has slowed in most countries, due to ageing populations and falling real incomes, and it has fallen in China as well to 5 per cent a year. But that is still higher than in any other major economy except Turkey, where consumption was boosted by a credit boom and refugee inflows.
3. Chicken soup for the protectionist soul
When you think about the products that will be affected by tariffs, you probably aren’t thinking about Campbell’s soup. But you should be. I highly recommend Ed Conway’s deep dive in into how the company got caught up by Trump’s trade war, trade agreement negotiations between the United States and Britain, and the decision to close a blast furnace plant in Wales.
The first thing you need to know is that making a tin can is hard, surprisingly hard. I realise this probably sounds a little counterintuitive. If I asked you what you thought the most challenging steel in the world to make was, you’d probably say: weapon grade steel, or maybe the kind that goes into a submarine hull or into the landing gear on an Airbus A380. But actually in some ways, it turns out making the steel that goes into tin cans (they’re mostly steel - the tin is just a plating) is even more exacting.
4. Odd jobs
In time for the wave of college students hitting the job market, Sam found this excellent job search advice from Adam Mastroianni in . In a nutshell: embrace your oddities.
Do you not realize that, to me, and to almost everyone else, you are all completely nuts?
No, you probably don’t realize that, because none of us do. We tend to overestimate the prevalence of our preferences, a phenomenon that psychologists call the “false consensus effect.” This is probably because it’s really really hard to take other people’s perspectives, so unless we run directly into disconfirming evidence, we assume that all of our mental settings are, in fact, the defaults. Our idiosyncrasies may never even occur to us.
… This is why people get so brain-constipated when they try to choose a career, and why they often pick the wrong one: they don’t understand the craziness that they have to offer, nor the craziness that will be demanded of them, and so they spend their lives jamming their square-peg selves into round-hole jobs.
5. From Tehran, with love
Finally, Katie recommends checking out Aaron MacLean’s interview with journalist, author, and television producer Annie Jacobsen for the podcast. In their discussion, they game out the use of nuclear weapons — especially relevant in light of the war in Iran.
