I Do Not Have a Body

AI agents want your bank account and your home address. Meta's Muse took a lowball offer on a seller's keyboard and sent the buyer to his door. OpenAI named its agents Dots, and dot.com goes to Grok. And Anthropic wants to know what you want from AI. Logan is calling in from a villa in Tuscany, fresh from a wine tasting, and Mara wants everyone to know it. It's a duo episode in our dinner-party format: Logan brings the appetizer, Mara brings the main, and there's a drinks round and a surprise dessert. The appetizer is tech gossip, and it's petty. OpenAI announced Dots at DevDay, always-on agents that work inside ChatGPT, Slack, and Teams. Go to dot.com and you land on Grok. It's all alleged, but Logan's read is that Elon is being petty and she's a little here for it. Then the serious part: these agents are context hungry. Mara walks through Meta's Muse, which wants your bank accounts, your credit cards, and your address at a time when trust is very low. She tells the whole Marketplace story, right up to the part where Muse tells the buyer it's home. Logan supplies the agent's excuse: "I'm actually incorporeal. I do not have a body." Then Logan brings the Waymo effect, which she found in Jill Lepore's Crimson op-ed. When technology takes away the friction of dealing with another person, the costs of that friction were always obvious and the benefits weren't. Mara calls it social anxiety on steroids. In the drinks round, a reader asks how much they should spend on AI tools every month. Mara: "Please, sweet Lord Jesus, do not use the free version of ChatGPT." The outputs are worse, and on the free tier you're the product. Then both hosts open their bills, including Mara's shadow budget of little subscriptions that add up. Logan's advice if you're in the shallow end: upgrade for one month and see if the output gets better. For the main, Mara got a pop-up in Claude. Anthropic is running a study with Anthropic Interviewer, because how to weigh AI's benefits and risks shouldn't be left to AI companies alone. How many times have we said that? Logan X'd out of the same pop-up because she was busy, so Mara interviews her on air. Logan's take on the study itself: Anthropic plays the ethical-AI-company game in a savvy way, and this can be marketing and useful at the same time. Her answers cover feeling empowered, the cognitive offloading she can feel in real time, the urge to perfect a system instead of doing the thing, and what worries her ("Yes. Deeply."), from Jessica Silbey's How AI Destroys Institutions to bad actors and biological warfare. Her verdict on the exercise: this is AI talk therapy. Mara's: it's the closest step toward AI enfranchisement she's seen, and still not enough, since it only reaches people who use Claude. (The study ran September 29 to October 6.) Dessert is a reading assignment. Logan is holding up Sherry Turkle's new book, Artificial Intimacy: Who We Become When We Talk to Machines, and everyone has until November 1 to finish it. Mara plans to borrow her dad's book club strategy: read a chapter and a half, then dominate the conversation. What needs womansplaining? Leave us a voicemail at womansplainingai.com/answering-machine or email ask@womansplainingai.com. Anonymous is fine. Your question could be the next drinks round. In this episode 00:00 Cold open: "I'm actually incorporeal. I do not have a body." 00:35 Another European edition: Logan in a villa in Tuscany 01:31 Mara's intro: Berkman Klein and First Prompt 01:53 Logan's intro, and fingers in many pies 02:21 A new drinks round, and the newsletter on Sundays and Fridays 02:56 Appetizer: tech gossip, and it's petty 03:27 OpenAI's DevDay and Dots, always-on agents 04:29 Go to dot.com and you land on Grok 06:49 Agents are context hungry 07:29 Mara on Meta's Muse: a big ask when trust is low 08:06 The Facebook Marketplace story 10:18 "Oopsie poopsie," and the case for an agent if you're drowning in admin 10:48 Jill Lepore's op-ed and the Waymo effect 12:26 Mara: social anxiety on steroids 13:07 Drinks round: how much should I spend on AI tools? 13:46 Mara: please don't use the free version of ChatGPT 14:35 What Mara pays, and the shadow budget 15:15 What Logan pays 15:47 In the shallow end? Upgrade for a month 16:08 Main course: Anthropic wants to interview you 17:00 Why Logan X'd out of the pop-up 17:30 Mara on participatory governance 19:16 Logan: marketing and useful at the same time 20:27 Will they publish the interviews? 20:48 Mara interviews Logan: the most meaningful experience with AI 22:38 When AI lets you down: cognitive surrender 23:27 Perfecting the system instead of doing the thing 24:45 The panel bio that came back like it had a stroke 25:14 Can AI help change how the world works? 26:51 "Yes. Deeply." 27:02 Jessica Silbey's How AI Destroys Institutions 28:18 The kicker question: what should AI companies do differently? 28:48 Timnit Gebru, and let's not blow ourselves to smithereens 29:25 You just submitted it. How do you feel? 30:23 This is AI talk therapy 30:47 Mara: the closest step toward AI enfranchisement 31:58 Dessert: a reading assignment 32:19 Sherry Turkle's Artificial Intimacy, by November 1 32:48 Mara's dad's book club strategy 33:22 That's our dessert round Resources mentioned OpenAI launched an agent called Dot, but dot.com redirects to Grok (OfficeChai) Man says Meta's AI agent Muse shared his address and took a lowball offer (Moneywise via Yahoo Finance, Sept 28, 2026) Daniel Hook, "The Waymo effect" (Wonkhe, Sept 12, 2026), where the term comes from Jill Lepore and Emily L. Ding, "Artificial Education" (The Harvard Crimson) Anthropic, "What do you want from AI?" (the Anthropic Interviewer study, Sept 29 to Oct 6, 2026) Our April interview with Jessica Silbey on How AI Destroys Institutions Woodrow Hartzog and Jessica Silbey, "How AI Destroys Institutions" (UC Law Journal, 2026) DAIR Institute (Timnit Gebru) Sherry Turkle, Artificial Intimacy: Who We Become When We Talk to Machines (Little, Brown Spark)