The Most Interesting Reads: August 15, 2025

Dear Friends,

I want to start with a note of praise for one of the greatest magazine writers of my lifetime, William Langewiesche, who died earlier this summer. One personal favorite is his harrowing recounting of the end of the Estonia, a ship carrying nearly a thousand people that sank in a Baltic storm. Each detail is perfectly placed and precise. “Survival that night was a very tight race, and savagely simple,” he writes as he describes the many tragedies and dramas that unfolded. As with the best magazine stories, it’s not just a gripping narrative, though it most certainly is that. It’s a story with details about how we love, how we understand fear, and ultimately how we act when death is at hand. Read it now if you haven’t yet. Unless you happen to be on a boat. 

The most peculiar piece I read recently was this story in Nautilus about a brilliant physicist named Peter Putnam who tried to uncover a new theory of the mind, which now has particular relevance to work in artificial intelligence. He was haunted by the burdens of wealth inherited from his family, as well as his sexuality. At the end of his life, he worked as a janitor and night watchman, even with tens of millions of dollars in the bank. He bequeathed the world a theory of the mind, left behind in old filing cabinets, as well as a giant collection of public art—including the iconic Gay Liberation Monument that can be seen both in Greenwich Village and on my old campus. 

The most intense story I read was Anne Applebaum’s dispatch from the civil war in Sudan. She traces the roots of the conflict, the perplexing outbreak, and the utter dissolution of values and morals as the war has continued. She traveled twice to the country in the past year, to both sides of the front line. “This wasn't merely the sound of artillery, but the sound of nihilism and anarchy, of lives disrupted, businesses ruined, universities closed, futures curtailed. No international laws, no international organizations, no diplomats, and certainly no Americans are coming to fill it.”

The news came out this week that alcohol consumption has hit record lows. This reminded me of one of the best podcasts I’ve heard recently, which was a debate on Boundless Life about whether one should completely abstain. It starts rather ominously, and feels a bit like you’re about to be thrown into a bro-pseudoscience salad. But gradually, it turns into the most thoughtful and interesting debate I’ve heard on the subject. (Spoiler: I left it continuing to believe that it’s entirely fine to drink every now and then in moderation.) And my favorite detail from any story came from this primer on the greatest mountain runner of all time, Kilian Jornet, and his training this spring. Here’s the bit I loved: “A month before he toes the start line in Olympic Valley, Jornet ran a 50-mile workout to simulate the end of the race. After dropping off his two older girls at kindergarten, he had six-and-a-half hours before needing to pick them up. So he threw on his shoes … and immediately got out the door.” That is an A+ way to balance parenting and elite ultra training.

Last: my book, The Running Ground, is coming out soon. In advance, here is a lovely seven-minute video—featuring my middle son destroying me in a hill workout—about why I run and what my father taught me about the sport. Also: please pre-order the book. That sends a signal to bookstores that they should stock it, which of course is something I hope many of them do.

The Most Interesting Things In Tech This Week

GPT-5 has been out for about a week now, and if you judge the model purely by its reception online, it’s a clear disappointment. GPT-5 is an improvement over GPT-4 in many ways, but it’s not a breakthrough. It certainly isn’t AGI or ASI. (Here’s Gary Marcus; here’s Casey Newton; here’s Ethan Mollick; here’s Joanna Stern.)

I’m still using the model every day, and, in some ways, slowing down progress might actually be a good thing. When a technology makes a seismic leap every few years, as ChatGPT has done, it’s difficult to integrate into a company. The operators could overestimate its capabilities, assume it will replace a whole set of functions, and suffer when it fails. Or, they could decide to wait to adopt it because they figure it’ll be twice as good a year from now. Steady, marked improvement is easier to track and plan for. I don’t think this is as good as AI is going to get; but I do think that a slower pace of change would help the world catch up. 

Meanwhile, the brilliant economist Erik Brynjolfsson wrote an op-ed in the Wall Street Journal making the case AI is worth more to the US economy than conventional metrics suggest. He and coauthor Avinash Collis say we need to look at consumer surplus: how much value we get from AI. By surveying people on what they would need to be paid to give up AI programs—an important distinction from how much they would pay for them—they found the technology contributes $97 billion a year. Their analysis is a good reminder that technology’s true worth isn’t always captured in traditional economic statistics.

The deal between Nvidia, AMD, and the US government is … unique. I fully support the government finding novel ways to reduce our debt-to-GDP ratio, but this one sends a troubling message. Is the argument now: Selling advanced chips to China is a dire national security threat—unless the government gets some cash, in which case it's OK? Also, export taxes are illegal under the US Constitution. The administration says this isn’t one. We’ll see if the courts agree.

Listen

I very much enjoyed chatting with Pete Pachal on The Media Copilot about a range of AI/media topics—including the economic models that might emerge with the rise of the agentic web. 

We’ve begun posting episodes from The Most Interesting Thing in AI podcast on YouTube. You can catch my conversations with the “Godfather of AI” Geoffrey Hinton, digital privacy pioneer Eva Galperin, and PwC Chief AI Officer Dan Priest on the Atlantic Re:think YouTube channel.

We’re releasing an episode a week; please check it out. 

This Week’s Question

I’m going to try to make this newsletter more regular, and one feature I want to add is more responses from readers—and also from commenters on LinkedIn and Facebook. So here’s today’s question: What is your favorite William Langewiesche story and why? I’ll write about it next Friday. You can email me here.  

Best * N

I hope you enjoy this newsletter, which, thanks to your support, has more than 475,000 subscribers. Please continue to forward it to anyone else who might enjoy it. They can sign up here.