Hello again from the Bay Area! I’m here for a bunch of things I can’t wait to tell you about, including our first New Things dinner and a Microsoft event.
But first: computers. I think about computers a lot. Lately, though, I’ve been thinking about the computers our AI agents use—and why many of them run Linux. Then Googlebooks finally arrive, ChatGPT gets more ads and Ring gets a smart lock you have to crank. Plus, I discover why I probably need to buy thenewthings.si.
Some people bike. Some knit. Some (fine, me) spend their free time watching a computer play computer games. Hobbies!
Yes, I’ve been watching my OpenAI Dot agent play Solitaire on its own virtual computer. I also made it paint a portrait of me in GIMP, its image-editing program. Why? Because I can!
AI agents now have computers of their own so they can work 24/7 for us. As you all know, I’ve been testing Instinct, Meta’s Muse and OpenAI’s Dot. We’ve talked about what these things can do—make appointments, monitor the web, buy things.
But underneath all that, these agents have virtual machines (or VMs) running in the cloud, complete with their own operating systems, browsers, storage space and memory. And right now, they’re largely navigating computers built for us.
Which raises a much more interesting question: If agents are going to become computer users, why do their computers need to look anything like ours?
Fun fact: The word “advertise” is derived from the Latin word “advertere,” which means “to turn towards.” In 2026, it feels more like the ads are turning towards us.
This newsletter was written and curated by Joanna Stern and Adele Lowitz.
|
