Some interesting things to read this weekend in March
Image created with Midjourney.
Dear Friends,
Last Sunday morning, a photograph in The New York Times Magazine caught my eye. It was stark and black and white, captured by the brilliant Ashley Gilbertson. A man, soaked like a sailor in a storm, was crossing a river with his arm out. A little branch stretched above him, shaped like a crossbow. It’s an extraordinary combination of action—the moving river, one body in motion—along with stillness. But something else struck me: That man looks familiar. I read the caption and realized that one of my old roommates, from 29 years ago, was now a character in a story about a truly remarkable race.
The Searching Smarter section below is sponsored by Elastic.
First the story: It’s a gripping piece about the Adventure Racing Association National Championships in West Virginia, as a team of three tries to solve clues and navigate the forests of West Virginia with nothing but a compass under the remnants of Hurricane Helene. Two of the three eventually finish 30 hours after they begin. The author, Doug Bock Clark, writes, “Navigating by the stars, I felt synchronized not just to the map and the earth but to the galaxies beyond. The blinding terror of the storm had been replaced by a sense of proportionality.”
Now the friend: He’s a remarkable man named Jedediah Britton-Purdy, who survives a treacherous tree fall but ultimately bows out of the race as he convulses in the cold and nears hypothermia. Jed and I were roommates in the spring of 1996 when we both won the same fellowship our junior year in college. I’ll never forget that first afternoon when we all stood in a circle and were asked to describe one surprising fact about ourselves. “I didn’t wear pants until I was nine, and I didn’t wear shoes until I was 13,” Jed said.
Shortly thereafter, he wrote a beautiful book, For Common Things, which is also excerpted here, that argued for Americans to take the world a little more seriously and to avoid the easy irony that lets one smile wryly as the things they care about burn down. It was a story, too, of Jed’s childhood as the home-schooled son of brilliant parents in rural West Virginia. “I cannot help believing that we need a way of thinking, and doing,” he writes in the book, “that has in it more promise of goodness than the one we are now following.”
He followed up with remarkable reportage from around the world, in a book titled Being America, then led a career as a law professor. I highly recommend this essay from 2014, “The Accidental Neoliberal,” a searing self-examination of the sort that I have rarely, if ever, seen. The young Jed, the 40-year-old Jed wrote, had let identity politics of a sort define his early work and was partially complicit in the system he was criticizing. There have been other excellent essays as well, including in my favorite magazine. He’s a remarkable writer and person. And though I haven’t seen him in a while, I am quite relieved that he survived the falling tree in the West Virginia storm.
Meanwhile, it’s almost baseball’s opening day! I highly recommend Kaitlyn Tiffany’s piece about a short stint she spent playing for the Yankees, or at least for a fantasy camp run by the team. “There’s no crying in baseball, and there are no women in baseball either,” Tiffany writes. It’s an excellent story about the history of sport and why the author has such love for it, despite baseball’s best efforts to not love women back. And while we’re on baseball, I can’t leave out one of my old heroes and office-mates, Roger Angell. If you have an hour or two, read his profile of Bob Gibson, which describes a game Angell watched at age 48. If you have just a minute or two, read his final blog post, which describes a game he saw at age 98.
Only two episodes have dropped, but I’ve been entirely captivated by a new podcast called The Redefector, from Julie Cohn. It’s the story of Vitaly Yurchenko, a Russian double (or maybe triple!) agent who came to the CIA in 1985 for reasons that have never been entirely clear. I hope to understand it better as the series goes on. And there’s a piece in Air Mail about the lawyer Tom Goldstein. I had known him only as a blogger, scholar, and frequent Supreme Court litigator. Apparently I missed a few things! Here’s a representative paragraph from the tale:
“Goldstein apparently got so stoned he left his brand-new Ferrari on the Las Vegas Strip with the keys still inside before staggering into the Bellagio’s poker room and begging Bilzerian for money so he could continue gambling. The following month, Goldstein would appear before the Supreme Court for the 24th time, successfully overturning a Vermont law that restricted the sale of physician prescription data for marketing purposes."
I’d also like to give a shout-out to this Paul Graham essay about schlepping and to Erika Hayasaki for this story in Men’s Health about two men desperately searching for answers to help cure them of Long COVID, a disease many people claim doesn't exist. I also appreciated this Teen Vogue essay about Vivian Jenna Wilson, Elon Musk’s estranged daughter who is fighting for trans rights. And to my aunt, Heidi Nitze, who has an exhibit of her lovely landscape paintings at the Carter Burden Gallery on West 28th Street in New York. She’s in her 90s, but the opening last night was filled with people in their 20s. “Do you know her? The colors remind me of my home,” one man said in the elevator on the way out.
This section is sponsored by Elastic, the Search AI Company. With the Elastic Search AI Platform, you can create custom GenAI applications to enhance resilience, fuel innovation, and boost productivity in your organization. But with seemingly endless use cases, where should your business start on its generative AI journey? Read the executive’s guide to operationalizing generative AI.
Perhaps the most important lesson for generative AI users is to focus not just on what you ask a LLM to do, but how you ask it—how you prompt the model. I know so many people who tell me: “I tried using AI and it just wasn’t that good.” I then explain how to prompt properly and tell them to go back. They’re usually amazed at how much better the model gets.
My first technique is to get the model to start role-playing. If I’m asking for journalistic advice, I start with something like, “You are a brilliant editor, trained at The Atlantic, who has closely read Strunk & White.” If I’m asking for advice on my daily videos, I’ll begin, “You are an extremely smart chief of staff to an executive at a media company. You care deeply about accuracy and you know a lot about AI.” I then give the model very clear instructions and specific examples that can guide it.
This sounds weird! But the models are weird. As research has shown, they act in spooky ways like humans. That makes sense: they were trained on lots of data created by humans. For example, if they read about violence and pain, they get rattled and inconsistent. If you tell them to calm down, they do so. Like humans, they are also good at certain things and terrible at others. I use models all the time to check whether something I’ve written is accurate based on the transcripts of an interview I’ve done. But I never use them to write anything for ethical reasons, for legal reasons, and because they are just bad at writing. You should also be very careful about each model’s privacy settings.
Another important step is to learn good ways to use the persistent memory of the model. AI models have what are called “context windows,” which define how much information about a conversation they can store at a time. If you’re uploading files, you can run out of space pretty quickly. But if you are just chatting, you can go on for quite some time. For several months, I have uploaded information about my training and diet to the same conversation in ChatGPT. It returns daily—useful!—advice about what I should eat to properly recover from one workout and prepare for the next. AI companies are also getting smarter about identifying and saving the most salient details from your previous conversions. To see it for yourself, check out the “memory” tab of your account settings when you’re signed into ChatGPT.
The most important point, though, is a simple one: These models are really smart. Try them and learn by doing. But don’t spend so much time looking at them that you miss the photograph of your old friend that just happens to pop up in a magazine at the breakfast table.
Cheers * N