Three things I read this past week or so that I wanted to share. There’s a theme in here, I suppose. But that’s where we’re at right now.

PS these are all by people I know, but they’re very worth your while.

Legacy Media Wants to Hire You (No, Not You)

In Scratch, Rahawa Haile absolutely nails the situation for that many writers and editors are facing. Want to get paid? Leave your morals behind… or at the very least tell them to wait outside in the rain while you go inside to have some uncomfortable conversations and a bunch of free snacks. The wild, wild difference between job listings at the most heavily-capitalized businesses or whale-sized media outlets feels more extreme than ever. “If you want to be an employed journalist—at the best-resourced publications with the biggest reach—it increasingly feels like you’d better get ready to throw your ethical qualms out the door or pack up your laptop and leave,” she says.

⌘

Why the Legendary Erdős Problems Are Falling to AI

Konstantin Kakaes wrote an engaging look in Quanta examining how computers have started coming up with novel solutions to Erdős problems, the vast list of conjectures made by the prolific Hungarian mathematician Paul. Kakaes shows how the answers emerged as well as how they didn’t emerge, and looks at not just what it might mean for mathematics—and mathematicians. (One answer: they’re often working for AI companies.)

⌘

What does AI look like?

Proof that the most straightforward questions are often the most helpful, a simple query about trying to visualize what AI is unlocked a whole universe for David Zax. Along the way, he manages to deliver a surprising explanation of how LLMs work that was both different and more useful than most of the other explainers I’ve seen. “A modern AI system like ChatGPT… is a series of spreadsheets filled with numbers,” he writes. And really, in some important sense, that’s all there is to it. Hundreds of spreadsheets with many many thousands of values, which—when connected together in the right way, at least—are able to generate some form of illusory magic. “All those numbers in all those spreadsheets stand there in columns like DNA, bits of crystallized wisdom. Every conversation you have with a given AI model sends your query ricocheting through the same vast warren of identical, frozen numbers. The only thing that changes is what you input, and somehow that’s enough to conjure the system’s infinitely variable output.”