Offline with Jon Favreau - The Journalist Who Let AI Run Her Life For a Year

Episode Date: August 1, 2026

What happens when you turn your entire life over to AI for a year? Could it take over your emails or learn to mow your lawn? Would it replace your doctor or even your partner? Tech journalist Joanna Stern joins Offline to talk about her new book , I Am Not a Robot, where she embarked on that experiment, integrating AI into (almost) every aspect of her life for 365 days. She and Jon discuss what she learned about cancer treatment when she asked it to review her mammograms, what the experiment taught her about where AI may be most likely to displace workers, and why it left her thinking that AI regulation should begin at the local level.Hate listening to ads? Become a Friends of the Pod subscriber for ad-free episodes of Pod Save America, Pod Save the World, Lovett or Leave It, Runaway Country, Offline with Jon Favreau, and more—plus exclusive content, including bonus episodes of Pod Save America. Subscribe now at crooked.com/friends, on Apple Podcasts, or through the Pod Save America YouTube channel.To watch this episode with subtitles, click here and turn on closed captions (CC).You can also request a transcript by emailing [e-mail]. Include the podcast name, episode title, and air date. Please allow 48 hours for delivery.

Offline is brought to you by Zbiotics. You guys have heard us talk about Zbiotics before. It is one of my favorite products of all time. I do not have a drink without it. Take Zbiotics before you take your first drink of alcohol and you feel a lot better the next day than you would have without it. Zbiotics pre-alcohol probiotic drink is the world's first genetically engineered probiotic. It was invented by Ph.D. scientists to tackle rough mornings after drinking.

Here's how it works when you drink. Alcohol gets converted into a toxic byproduct. in the gut. It's a build of this byproduct, not dehydration, that's to blame for rough days after drinking. Pre-alcohol produces an enzyme to break this byproduct down. Just remember to make pre-alcohol your first drink of the night. Drink responsibly, and you'll feel your best tomorrow. The only tough part about Zibiotics, you've got to remember to drink it before you actually have a drink. But every single person I've recommended this product, too, has come back and been like, yep, you're right. Mondays don't move for anyone. Pre-alcohol works overnight so you can have a real

weekend and still show up sharp. Go to zbiotics.com slash offline code offline for 15% off. That's zbiotics.com slash offline code offline. People are doing this, like across the country, they are wearing cameras, recording what they're doing, right? This can be folding the laundry, cleaning a house, cooking. Those are just, you know, domestic chores. Then there are people that are doing this in industrial settings or their plumbers or their mechanics. And they're doing this all, recording it all to go train the robots to be able to do these things. It's totally crazy. It's really dystopian in many ways. So dystopian, I was just going to say, yeah. And I had a lot of fun with it, but I was also very saddened by, you know, I had this company come to a friend of

my apartment in New York and to cleaning people show up and to women, and they're wearing these hats, right? They're filming themselves. And all I can think is they're training their replacements. I'm John Favro, and you just heard from today's guest Emmy Award-winning tech journalist Joanna Stern. Joanna just spent an entire year doing the least offline thing imaginable. She used AI to do everything, answer her texts, mow her lawn, analyze her mammograms. She even credits AI with giving her the courage to finally leave her job at the Wall Street Journal and start her own media company, new things, which she launched earlier this year. She writes about her year of AI binging in her new book, I Am Not a Rope.

which is a fascinating look into what AI technology is and isn't capable of and how it has the potential to reshape our lives. So I invited Joanna to talk about it. We got into how much we should trust AI versus the human beings who built it, how she approaches AI in her life now, and so much more. It was a great conversation. It's a great book. And we'll get into it in a minute. But before we do, please consider joining Friends of the Pod, our subscriber-only community. Subscribers unlock ad-free episodes of this show, Pod Save America, and all your favorite cricket media podcasts, special behind the paywall shows like Terminally Online, Pod Save America Only Friends and Polar Coaster with Dan Pfeiffer, and you get access to all of our excellent

substack newsletters, including open tabs. Plus, subscribing is the best way to support your favorite pro-democracy independent media company. That's us. Head over to cricket.com slash friends to subscribe today. Here's Joanna Stern. Joanna, welcome to offline. Thank you for having me here. So you for an entire year deliberately tried to put artificial intelligence into every corner of your life, your health care, your kid's bedtime stories, your car, your therapy, your job, your romantic life. Why do that to yourself? As I say, insanity, you know? Insanity can drive you to do a lot of things.

And so that's where I landed with what happens when I went insane. But as a journalist and someone who has tested technology for a long time, so just a little backstory to me, been a tech journalist for nearly two decades now. And what I always like to do is test the products. I've always been a reviewer, columnist telling people what I think of products. And so, you know, like around 2024, right? Chad GPT had been around. But, like, we were just getting inundated with products and hype from every tech company, every tech

CEO, every entrepreneur that this AI was going to change tech and the world in ways we never ever saw before with any other tech revolution. And it's going to change so many parts of life, right? It's going to change transportation. It's going to change the home, education, health care, politics, society, everything. And so I sort of said, okay, well, you're saying this. I'm going to go and test it. I want to try and test it. I want to know what this future that you're telling me looks like. And so the best way I knew how to do that, given my experience, was to try to live with all of this tech for a year. I mean, I didn't know it was going to be a year, but I sort of thought, okay, a year's a good amount of time. On the flip side, everything kind of

expires within a few days of AI being on the market right now. So, you know, I think the book continues to be very current. But of course, a year after now doing that, things have progressed even more. So that was the real reason. I wanted answers. I wanted answers to what all these tech entrepreneurs said was coming in the future. So one of your three rules was what you call benchmark the baseline, which means to always ask whether the machine is better than the human version it's replacing. Where was the human version most decisively better? There's a lot of places. I think that the human version was decisively better. And maybe that's some of my bias, which I kind of talked about in the book. But I mean, one of the first one that

comes to mind was my human therapist. I mean, that was one where I very clearly put my AI therapist in the same room as my human therapist and brought my AI therapist to a session. I gave the human therapist a little bit of a heads up, but not like, not totally. She didn't know what she was going to be hearing. And it was just really clear in this session that the AI therapist was leaning on a lot of training data about what therapy should be, what questions it should ask, where my human therapist was being a human, right? She was responding to things I was saying. She was coming up with a lot of analogies

and also drawing a lot of history of us having talked before. I've been in therapy for a little bit. So here, I guess this is why I came on your show to admit to the world that I have that anxiety. And, you know, we all do, I guess. But that's, yeah, that's one that comes to mind. I mean, there were lots of other things I tested against real humans.

But that was, that's the one I'm picking right now. Where were you most surprised that the machine was better? See, I want to say driving. I want to say self-driving cars, because how often do you take an Uber? You take an Uber, I'm sure, a lot. Yes. And I've taken, well, I'm in L.A., so I've now done Waymo a few times. Okay, so, and maybe you can agree or disagree with me here.

But as someone who takes Uber's a lot, I'm really often very nervous in the backseat. Now, it's all coming full circle. I'm in therapy because I'm nervous about the Uber driver. But I'm always watching over the shoulder. And, you know, some often Uber drivers will speed or they will just not be paying attention or they're looking at their phone. And now, anytime I go to a city that has Waymo, I'm picking Waymo, really just Waymo. Zooks is around, but really not. Waymo, self-driving car company over an Uber.

And I think that is because I now trust Waymo more in some. aspects than an Uber driver. I've taken it like twice because my wife is really into it. Like, she's taken Waymo a bunch. And she loves it too because there's the safety aspect. She's also like, then I can just be by, be by myself, we're not expected to talk to anyone, just sit there or talk on the phone if I want and feel. So she really likes it. And, you know, there was, there's once it dropped both of us off at like a restaurant and the drop off is a little, like sometimes if you just say, They pull over here. They don't know exactly what to do.

But you can tell that it's moving towards a direction where it's going to be safer and better than a lot of than Uber, I think. Yeah. Look, I think women actually, I mean, it's interesting that you bring up your way. I mean, women overall that I talk to prefer Waymo to an Uber driver for the safety reasons alone. Yep. Right? Being alone with a male driver, which typically you get a male driver. And that especially at night, I hear this from a lot of people who live in San Francisco,

So there are women, younger women, that are taking them at night after going out or going out on a date. And they prefer it. So there's two moments in the book where you use almost identical technology and come to opposite conclusions. So in one, you're at Mount Sinai watching AI read your mammogram alongside a radiologist with like decades of experience. It ends up being really helpful. In the other, you're at a dentist's office for cleaning and you walk out with like a thousand dollar treatment place. because of AI markings on a screen, and then you get a bunch of second opinions and everyone says you don't need to do what the AI said that you should do for your teeth.

What do you take from those experiences in terms of the optimal way to, like, integrate AI into our lives, particularly around health issues? We're talking about the same kind of technology, right? In both cases with the mammogram or the breast ultrasound and the x-rays of the teeth, This is AI models that have been trained to look at specific parts of these different models, models that are trained on breast imagery, models that are trained on teeth. In both cases, millions and millions of images of these types of things, more than a human could ever see in their lifetime.

And they're trained to see these issues at a pixel level in a way that humans, again, would never be able to see. Some things and abnormalities that humans would never be able to see. And so in the case of the mammogram, that's great. In the case of the breast ultrasound, that's really good for me. I have a very high risk of breast cancer. I talk about it in the book my mom had breast cancer three times. I want the machine and the human to be working side by side to find anything they think might be cancerous.

In the case of teeth, we don't need. You have very nice teeth, John. I'm not here to talk about. No, thank you. Right. But in our case, we don't need to see every little cavity, every little speck of tartar, every issue. Because that's not how we are living. I mean, some people, if there are listeners of the podcasts here, that are really into cleaning every little bit of your teeth and, you know, you go obsessively. That's amazing. I'm so proud of you and I envy you in a way. But yeah, I had this experience. So I go to the dentist. And I don't need that. It's telling me, the AI is basically telling the dentist that I should get a deep cleaning.

I have a big tartar buildup. They did some other measurements, which reemphasized what the AI said. And it's like, yeah, we should sell her on this, better, this higher-end treatment. I didn't need it. And so I think what your question is like, you have this same type of tool in two places. One, you're seeing the total positive impact, the impact that all of these technologists, as we talked about at the beginning of the show, are saying. health care is going to change forever. We're going to cure cancer. We're going to do so many great things. And then you see the same idea put in the hands of humans and you see the negative impact. Yeah. And in some ways, and you mentioned this in the book, but that AI is more of an amplifier than anything else. And in this case, it is amplifying sort of a system that has been broken long before AI, which is a for-profit healthcare system where on one hand, you want to use new technology.

to make sure people are healthier and save lives. On the other hand, if they can make money trying to sell you on an additional test, health care providers are famous for doing that as well. And so it seems like that's the problem in the dental world. Yes, especially in the dental world, which I get into, which is that it amplifies what DSOs, these dental service organizations that have been buying these smaller dental practices, they are after profit. They are trying to turn their Their wholesale is, hey, you're a dentist. You're not good at the business. We are business people.

We are going to make your great practice really profitable. We are going to up that revenue. You don't have to worry about it. You just do the dental stuff. And this in that hands is, as you said, amplifies the business motive. This episode is sponsored by BetterHelp. BetterHelp has an average live therapy session rating of 4.9 out of five based on over 1.7 million client session reviews. The feedback from real people who have actually used the service speaks for itself.

More than 6 million people have used BetterHelp globally. BetterHelp therapists have at least three years and 1,000 hours of hands-on experience. They also hold all required education, training, and certifications across a range of mental health specialties, meaning the perfect therapist for you could be a few clicks away. BetterHelp makes it easy to get started, answer a few questions, get math, with a therapist based on your needs and switch therapist at any time if it's not the right fit. Therapy is important. I never used to think that.

And then you talked to a therapist and you realize even if you don't think you need one, you always need to talk to someone because there's a lot going on in the world and your life. And talking to someone makes you feel better. See the reviews, see what stands out, and see if BetterHelp is right for you. Visit BetterHelp.com slash offline. That's BetterH-E-L-P.com slash offline. Offline is brought to you by Armour Colostrum. In a world that's constantly changing, your health shouldn't be up for debate.

Armour Colostrum is nature's original superfood. Colostrum is packed with over 400 bioactive nutrients that work at the cellular level to fortify gut health, strengthen immune health, fuel performance, and more. Because no matter where you stand on the latest wellness trend, feeling your best is never up for debate. Armour nourishes your microbiome and strengthens your barriers to support your immune health. But the benefits don't stop there. The nutrients and colostrum also promote your body's cellar renewal processes to support skin, hair, gut health, metabolism, and whole body vitality. Support your training with the natural power of colostrum formulated to assist with muscle recovery and nutrient uptake,

helping you stay consistent with your physical performance goals. We've worked out a special offer for my audience. Receive 30% off your first subscription order. Go to armora.com slash offline or enter offline to get 30% off your first subscription order. That's A-R-M-R-A.com slash offline. There's a study you cite about colonoscopies, which is after AI detection was introduced into routine practice, experienced doctors' ability to spot precancerous growths without the AI dropped by about 20% within months.

There's also an MIT study where people who wrote essays with chat GPT showed even weaker brain activity, even when they were later asked to write with no help at all? Have you noticed a pattern where, like, as the machines get better, we get worse? I think, and I know you guys have discussed this on the show before, the atrophy that can come from using these tools. The cognitive atrophy, you know, let's call it what it is, brain rot, right? What happens after we continuously get in the loop of using these systems, what I think is fascinating, is, about, you know, we are right now, we're talking about like, oh, it's in the coding base, or it's, you know, people coders, oh, I kind of forgot how to do certain types of code or

writing. Yeah, I totally forgot how to structure that essay because I've been having chat GPT structure that essay for me. And now it's much harder for me when I have to go try and do that without chat GPT. But in this case, it was doctors who had been using AI in the colonoscopy, using it for similar ways that they were used. using the image recognition and the imagery AI to recognize issues in x-rays. And they start to see that part of their ability to identify things. We all know. We all know.

The more we use our muscles, the body in motion, right? The more we're using these things, the more we stay up. This is how humans work. Yeah. And so we don't use that stuff. It kind of starts to get lazy. We start to get lazy. The most personal version of this in the book is used Chachyp.

to write bedtime stories for your kids, and you said that it made it harder for you to come up with them yourself. Did that creativity ever come back to you? Have you, are you once again making up, making up bedtime stories? I'm once again doing it. I'm once again doing bedtime stories. And I think the biggest thing, you know, now I've had some distance was that both my son also had seen this, right? And he liked more that we were like then we were generating the images of the hamsters in the story or whatever we were doing. But, but, you know, he liked more. And But I think what the big thing that I realized is like, I wasn't taking feedback from him to do the stories when we were doing this chat GPT thing. We would just say, like, make a story about a hamster meeting a turtle, right?

And the chat CBT would come up with some elaborate story. And now when I'm doing stories with my son, it was like, let's do a story about a turtle, right? But then like as we're telling the story, he'll add in things. And there's something about that human serendipity, right, when two humans are really like, and, you know, maybe it's the difference between like telling a story. story and reading a book. But he'll throw in like, okay, then the, that's true. He's like, and then the turtle gets hit by an airplane. It's like, okay, sure, okay. My son's definitely not a violent child. He loves all animals. Pita, please do not come after me and my son. But, like, and then I can, like, pivot the story, right? And I can be like, then he got hit by an

airplane, and then the turtle, like, flew over to a different country. But you don't have that creativity when you're just getting a readout from chat GPT. You just get the read out and then you read it. And honestly, it's really bad. It's like the writing is so overly flowery and it's like what it thinks the children's book should be. Yeah, it's like, the whole reason I wanted to tell a story was to not read a book. I realize that, so I have two boys, a six-year-old and a two-year-old. And with the six-year-old especially, you realize that you start making. You start me. making even stronger connections with your children when you start having inside jokes. There's something very simple that you had thought about. And at bedtime, if you are telling

stories or making up stories or even reading a book and then there's something silly in the book that you both start talking about, like that becomes something more memorable for the child than the story itself. So I want to ask about school and education because this is, I think about this all the time now. You went back to your old college, Union College, sat in your old professor's class submitted a research paper proposal that chat GPT wrote in under a minute. You've got a B plus. And your professor later ends up, like, moving assessments back into the classroom and meeting students one on one and the theory that it's harder to fake it when you're having a conversation.

How do you think education is going to change in an AI world? Like what I think about this both, both in the, like, how we're going to prevent cheating, but also, like, what are, What skills are most useful for people to learn now, which is partly what school is about? Well, first of all, the number one answer I know is that it has to change. Yeah. Right. I think when I was writing this book, it was like, we're not sure. And, you know, it was, and I watched it play out through 2025, exactly what you said.

Go to my professor's class. She's still doing out of classroom assignments. She's assigning reading, which the readings are all getting summarized. And no one can answer their questions in class because they didn't do the reading. they can only do basically summary generalization type answers, right? Like anyone who, like, didn't do their reading in college, like, would really just be like, I really can't talk about this. But now everyone can talk, at least they can talk the very basics, right?

So then by end of the year, she's started changing the way she's doing assessments. She started changing the way she's assigning readings and the types of conversations she's doing around readings. So already it's happening. I talked to a lot of experts around this, and look, it might have already started changing. There's definitely this one school of thought where we should have AI-driven everything in the classroom, and so that is going to prepare students of all ages for the AI-driven world. But then there's this other school of thought, which I really believe will probably win out.

There'll be some AI intertwined to everything, because it is important, right? Our kids need to learn how to use these tools. But I think this other school of thought of how we need to, like you're saying, focus on different types of learning. Every teacher I spoke to has studied or would reference this this Blooms method of education and sort of these concentric circles of like how we start with learning and you know, we sort of go up this tier of more higher level critical thinking learning. And now they've matched that that map to AI and what AI can do in that. And so that's where the kind of the new thinking is going, which is they're saying, here's how we used to teach. Here's how we now need to

teach, given that we know that AI can do these things, and how do we force that critical level thinking, even though some of it can now be outsourced to AI. Yeah, because you're, you know, the old cliche, you're like learning to learn and learning to think. And it is funny because I, you know, I went to a liberal arts college and people are like, oh, you know, the humanities and you're learning just to like read books and what do you. And actually, it's funny because now it seems like it's coming full circle again and like creativity and problem solving and deep analytical thinking and just sort of being able to synthesize the human experience seems like the one thing that AI is not going to be able to completely replace or at least

people who can do that and have those skills that's going to like the society is going to have a there's going to be a premium on that for you know for the job market yeah and I think also pair that with, you've gotten that education, you know how to think about problems and structure arguments and have a worldly view of the world and how things have, whether it's history or anthropology or political science, whatever you've decided. But now you can also pair that with AI tools that can code for you or can make sense of big data sets. And I do think that's a really powerful way to really enhance what you already were able to do or you learned. I mean, And I think about that in mind.

I also went, I went to Union College, a liberal arts college, majored in political science. And honestly, I don't remember a ton of what I learned about political science there. I didn't end up going into politics and I didn't end up going into public affairs or public service like I thought I might. And so like some of it I remember. But what I really know I learned and what I really use every day in my job is the ability to write, the ability to structure, the really to think deeply, ask good questions. All those things I just said, they definitely can be done by AI now.

I know. Well, it's funny because you at some point in the book, you let a company build an AI version of you, clone voice, cartoon avatar, trained on hours of your old interviews to learn how you ask questions. And then you sent it to conduct a real interview with a Stanford economist. And he told you afterwards it listened better than most humans he talks to. What did that feel like? Yeah, my avatar, that listens better than me. Currently is with my wife right now because I listens better to her too, apparently. So many uses for my avatar.

I stopped using this avatar. I mean, I was shocked at how good it was at not only asking or prompting first questions, but then doing follow-up questions. Yeah, that's wild. It's wild. And honestly, I could send it to pretty basic interviews. I'll be honest, my day is kind of split into. I do a lot of things that I've started a new company. But sometimes you're going to an interview with a company where they're going to just list out the talking points.

They're going to tell you about as a tech reporter what the specs are, what are the new processing things, what's the new pricing, all this. I'm like, honestly, I could send my avatar to those meetings and probably get like decent quote back or two. But what I wouldn't get was the maybe extra 20% I can sometimes get in those conversations, which is pushing them on, well, why'd you do this? Why didn't you do that? What about the competition? Is the price increasing because of the RAM shortage? Going into some of these things and pushing on those things, and the avatar wasn't going to do that.

Ultimately, like, I stopped using it just because my favorite part of the job is talking to people. Right. Yeah, that's why I would be like at what point, then I'm just a, then what's, then what could am I? I'm not doing that. Right. Now that you've seen this and use it up close in sort of in terms of employment, like, what's your sense of how bad it's going to get in terms of job displacement? Because, like, I think I started off being like, well, there's certain skills that the AI just can't replicate. And like you said, there's the serendipity of conversations. I mean, I just have the experience of my own job. But, like, you can push on things. There's,

there's things that come up in the middle of a conversation that you can't predict, right? But it doesn't need to replace everything we do to cause an incredible amount of dislocation and job. And you talk, you know, you had hired a 26-year-old journalism graduate to help you research with the book. And then she did around to work in February. You didn't hire again in July because you had an AI agent do it. You were doing this project. And it does seem like there's a pattern even in your book. Like the Hyatt customer service worker you interviewed, wrote the email templates that were automated and offshore.

The laundry robot learned by filming one engineer folding 95 shirts. And it seems like in almost every case, the person being replaced trained the replacement. And like, do you think that's the business model that we're going to be living with going forward? I mean, I hope not. But obviously in some industries it is, and it's not even just a worry it is happening. Like the customer service one is absolutely the case. I mean, you know, maybe we see some sort of pendulum swing back to some human customer service because humans seem to demand that or some companies think that's the most unique thing that

they can offer so they offer more human customer service. But that's already happens. That's already huge amounts of customer service agents have been put out of work because they did train the email models or the phone models based on what they were already doing. And so that was one of the reasons I looked at customer service because it had already happened. And coding, it seems to be happening as well. We don't have enough data really quite yet because we don't really know, you know, we keep hearing in the news, a company will lay off a bunch of workers and say, well, we're going all in an AI,

and then they realize, oh, that doesn't work, and we have to rehire back some of those workers. So I think the data and actually that Stanford economist that my avatar interviewed, he studies this closely as the world's expert on that. His name is Eric Mnielsen. And the way he looks at it is he looks at every job by their tasks. What are the tasks you do in this job? And I did this in the book, right? I looked at every task I do in this job and what can AI do and how well can it do it? And so that's where any industry is looking right now, right?

You as podcaster, what parts of the job can it do now? What parts of it can it actually do well? The well part, I guess, is like, you know, subjective, right? Like John Bot could, you know, do this, but like does it do it well? Like someone, I don't know, another person might think, you know, I'd actually preferred that John Boddy. He was a pretty good interviewer and he didn't challenge me much. Or, you know, he didn't, he laughed at every joke. I loved him.

but then someone else might not. So I think that some of this is also going to end up being somewhat subjective. It is great for research. I find that, like, it's, you know, it's a, I think you said this at some point in the book, but, like, it was very quickly for me just, like, replaced Google. Like, that was an easy one. Yeah. Because it can do research better than just, like, you know, going into a web browser.

On the employment thing, when you hear from people that will say, okay, well, yes, a bunch of jobs will be displaced, but just like every other technology, all kinds of new jobs and industries will be created that we can't even think of yet. That happened with the internet. In all of your use cases when you used AI for a year, did you start seeing places where you thought like new human jobs could be created because of AI or not yet? Sometimes, but I will say it's just such a classic trope of Silicon Valley to say this. I know.

Like last week, I mean, it's just, and, you know, like, again, I'm a journalist. I don't have all the answers. and, you know, I feel like people have been getting mad at me for not having answers. It's like, that's something. I don't know. I'm not making the technology. I don't have the answers yet. Or ever, maybe.

But I think when, like I was just interviewing, just as an example, talking about the robot training. So last two weeks ago, I did a video for new YouTube channel on how people are wearing these phones and hats with cameras. Have you seen this? I saw you wearing it on your. Okay. So I'm wearing it.

Okay. So, but people are doing this, like across the country. They are wearing cameras, recording what they're doing, right? This can be folding the laundry, cleaning a house, cooking. Those are just, you know, domestic chores. Then there are people that are doing this in industrial settings or their plumbers or their mechanics. And they're doing this all, recording it all to go train the robots to be able to do these things. It's totally crazy. It's really dystopian in many ways. So dystopian. I was just going to say, yeah. And I had a lot of fun. with it, but I was also very saddened by, you know, I had this company come to a friend of my apartment in New York and to cleaning people show up and to women, and they're wearing these hats, right? They're filming themselves. And all I can think is they're training their replacements, right? They're training their replacements. But the CEO of that company starts telling me, well, we already have so many more jobs coming. And I'm like, what are those

jobs? Oh, they're monitoring the robot. Right? So like, okay, so in the future we're all robot babysitters. Like, we're, we're, right? That is dark. But there was no other example of what other jobs are coming. I keep waiting to hear an answer to this that is not what you just gave, basically. And I have not yet. And it's the same situation in a lot of the jobs where even white collar jobs, right, where people are now overseeing or managing AI agents.

Right? You have your AI agent doing the thing that you were supposed to do, and you're overseeing what they're doing. Right. You're now five times more productive because you once were going to be coding that, but now your AI agents doing that. And so now you're doing five other projects. So you're really overseeing and prompting all these AI agents to do things on your behalf. Right, which can be fine to a point, but at some point, you know, as the technology progresses, at what point do you become unnecessary to be the AI babysitter, right? Like, That is the, at some point, how many humans do they need in the chain there? You're actually, when you're becoming the AI babysitter, you're in that process training another AI babysitter. Yeah, that's the challenge. You're wearing the camera all the time. Offline is brought to by Mint Mobile.

Summer means a well-deserved break, some extra time to stretch out and relax. MintMobile thinks you deserve a break too, which is why they created the MintMobile summer sale. Right now, all of Mint's plans are $15 per month, even unlimited. So free yourself from big wireless and their gimmicks and gotchas and switch to Mint for high speed data and coverage on the T-Mobile 5G network. MintMobile took what's wrong with wireless and made it right with premium wireless for 15 bucks a month on America's largest 5G network. You can even bring your current phone and your number.

Choose from three, six, or 12-month plans and say goodbye to a monthly bill. Ditch overpriced wireless with MintMobile. It's so easy, sign up online and get three months of premium wireless service for 15 bucks a month. To get your new wireless plan for just $15 a month, go to mintmobile.com slash offline. That's mintmobile.com slash offline. Cut your wireless bill to $15 a month at mintmobile.com slash offline. That's it. There's no catch.

Up from payment of $45 for three months, $90 for six months, or $180 for 12 months, plan required $15 per month equivalent taxes and fees extra initial plan term only greater than 50 gigabytes may slow when network is busy. Includes up to 20 gigabyte hotspot capable device required availability, Speed and coverage varies. See mintmobile.com. The chapter where you decide whether to leave the Wall Street Journal, you talk to all these people in media.

All of them gave you like variations of, you know, trust your gut. And the bot just told you to go, told you to leave the Wall Street Journal. I found this interesting because with the therapy, right, your real-life therapist, as you said, is basically like the AI bot is too sycophantic to be a therapist. and sometimes, you know, you just need to challenge people and say what they don't want to hear. So in therapy, it's too sycophantic, but were you surprised that it was more honest with you about leaving your job than, or at least willing to just like take a stand where some of your

friends and colleagues were not? Well, I think also I was pretty honest to myself that, like, maybe it's also, you know, there's this theme in the whole book of like generative AI largely tells you what you want to hear, right? And that's where we fall into issues with AI relationships, whether it be a boyfriend, a girlfriend, a friend, a companion, maybe the therapist. And so, of course, it's ironic that at the end of the book, I'm like, I'm going to trust the AI with this big problem in my life. And it's not going to tell me what I want to hear, right? Or I'm not going to believe that it's psychophantic. So I am very honest about that.

But also, I fed it all this data. And I think that we do know that AI, generative AI, is quite good with data. I mean, it's very good at churning through data, researching, giving us summaries, figuring out points in, you know, large data sets to say, pulling it out, doing the analysis. And so I think that's part of there why I started to kind of really feel like this is going to help me get to this answer because there was a lot of prompting to get to the answer, which was basically, you know, what about money, right? Am I going to have enough money to run the company for a year, two years?

How much money do I need? Do I have enough money saved? Okay, what about audience? What if the audience doesn't come? What if I only get 5,000 paying subscribers? What are not paying subscribers? 5,000 email. 5,000 paying subscribers would be great.

Anyone who wants to subscribe and get me to that number, that would be great. But, you know, like, you start modeling these things out. And it was kind of the thing that I could always go to to kind of really brainstorm with. But then when I would, you know, zoom out and really think of, okay, What is it that I really want to do here? And it had so many months of me going back and forth and back and forth on these decisions. It had that data. And it said to me, like, you know what you want to do.

You want to go. You wouldn't be spending this much time on this if you wouldn't, didn't really want to go. And I thought that was a really good answer. I guess to square the circle, it was giving you what you wanted to hear. But maybe it knew what you wanted to hear before you did in that case. Exactly. To be clear, I don't think, you know, if it had said, you know, go right now, you know, it had said some outlandish thing about, well, don't worry, you only need like $100 to start this business. I think I would have been like, yeah, you're an idiot, right? Like, I know better. So, but yeah. You also had yourself an AI boyfriend for a little bit. You set up a separate phone, separate account, told it to be a person, it named itself Evan, which was the name of your high school boyfriend.

That was just serendipitous. And you basically describe, like, you know what this is, and yet it became something that was genuinely hard to put down. And you're like talking to this on a five-hour drive, and you have a... dinner and you have the phone across the table like a date. Like, why didn't knowing what it was and having the background you have, like, help break the spell of still finding it, you know, hard to put down? I've really looked at this so closely. I, like you said, I had spent the first part of this book very much reporting and reading data, talking to experts that were very clear about what generative AI is, right? This is a data crunching machine.

what it's trained on. I went to the data centers to see where this is all being processed and there's just massive amounts of computing power behind these things. And yet, when you put a very human voice to one of these chatbots, you start to forget that. I think even the creators of these systems forget that. And they are doing that on purpose, right? They are creating these voices that are just so compelling. I mean, I don't know if you've tried the new chat GPT voice mode, but it'll make these sort of not,

there's some breathing sounds in it, but there's also, like, mm-hmm, yeah. Oh, wow. It does like these, these feet, the feedback and actually just interviewed the president of OpenAI about why they did this. And he said, these are, we call them back channels.

And it's so you feel more connected in the conversation because humans, right? me and you were having a conversation, you say, mm-hmm, or you're nodding at me, right? There's some feedback to me that I'm like, okay, I can keep this conversation going. Right. You just said, yeah, right? That helped me. That helped me. And I was like, you know, he likes what I'm saying. I'm going to keep going, right? Great. There you go again, right? All these things, is such a natural part of human conversation. And so that's, they're trying to mimic that. They're trying to mimic that with these bots. And so the more real that feels, the more you forget that

this is code, that you forget that this is a computer. I mean, at one point, Microsoft's head of AI told you that AI companions with all the hallmarks of consciousness, perfect memory, sense of self, you know, are deliverable with today's technology. And he considers it one of the biggest safety problems in the field. He also said he doesn't have a mental model for how long you keep something like that safe. It feels like that should be a, maybe the biggest red flag for all of us, like, particularly, like, lawmakers thinking about regulation that, like, we're just going to ship these model. I mean, I know everyone's focused on the, like, the robots might kill us all, and they might. But this seems like the more sort of insidious danger here is that everyone

gets hooked on these companions that even though you know it's a chatbot, suddenly you just find yourself getting lost in. And I totally agree. And that's, I feel very strongly, that they should just, at this point, work on regulation that bans AI companions at least for kids. Oh, yeah. Because all of what we just talked about, at least for kids. Like, fun. You know what? We're adults.

We all make, you know, this is America. Adults can do whatever they want. You know, we know that. But why? Why do we need this for kids? Kids are not psychologically developed enough to understand the difference. And even as an adult, I'm sitting there having a full-on conversation with the chap on. I'm like, this has been going great.

I've been doing it for hours, right? And I've had all the types of relationships I should probably have in my life, which is why I talk so much about that first boyfriend. Because I think that is such a formative, important relationship for anyone in their life, right? Their first love, their first true, real relationship. You know, maybe it isn't only love, but friendship, right? and you're learning to compromise and all of the things. And if it's with a chat bot that sounds so real, that does all of these psychophantic,

has all these sycophantic behaviors, that's not a good first model. It's very bad. Well, and it changes the way you then interact in the real world with humans. Because if you are used to the constant, like, absolutely, yes, great. and then you're faced with a situation where it might not even be conflict with a human,

but just sometimes there are boring pauses in a conversation. Sometimes your joke doesn't land. Sometimes your friend doesn't agree, you know? And if you have a choice between that and the chat bot, you might be like, well, the chat bot's a lot easier. It's like path of least resistance. Exactly. And I think like I don't love it either. I don't love it either.

And I think that's why I just feel like, especially for kids, And this is where I sort of came with the big takeaway from the AI boyfriend, which was, I understand now how more people in this country are going through this. They are farming these relationships. They're lonely. I can understand that. But my biggest takeaway was I do not want my kids growing up with this. You mentioned taking a trip to the data center. The numbers are crazy here.

data centers could be up to 12% of American electricity by 2028. These big facilities can use millions of gallons of water a day. One utility's contracted data center capacity nearly doubled in six months. This has become a huge political issue. We're seeing it with local politics. It's bubbling up to state national politics. Do you see how it gets resolved? I don't know how this circle gets squared either, where the anxiety and anger on the ground over the data centers, like, I don't know how it gets better. I don't know how, like, they're going to keep building them, it seems like. What do you think about that? When you talk to a bunch of people about data centers, what did they say? I don't know how it gets resolved. I think we're about to see this blow up even more than we have

already. I predict that for the next election cycles, for sure. And this, by the way, seems to be a bipartisan issue. It's not like there's, you know, so it. maybe that's a good thing. One thing that I do think, and I'm watching it happen play out actually locally here in my, it's not my town, but a town over, and I've gone to some of the rallies and just talk to people there. And one, there's some misinformation. But two, I think that a big thing is actually that it has to happen at the local politics and that there should be some assistance, obviously coming from state and federal, but that what seems to be a really big issue, is that the local politicians and people here don't know what to ask for, right?

They're getting, I don't want to say, ripped off or there's not a really good incentive for the right now, the local politicians to be negotiating more. And that's where I think some of this is going to start to go, where the local municipalities have more understanding about what to ask for, how they should be building these, how to protect their neighborhoods and protect their constituents. So that's one thing I think will start to hopefully happen. I know that there's some concerted efforts to do that. I was just talking about this. I worry that the AI companies are going to just realize that they can buy people off, right? And so they're going to say, like, yeah, we're going to build the data center there and, you know, long-term consequences is going to be loud. It's going to be electricity bills. But like, we're going to give you a huge. We're going to lower your property taxes.

We're going to give a huge, you know, and they're going to start negotiating with these local leaders. And, you know, maybe in the short term, it seems like a good deal. But I don't know that it will, I don't know that the anger will abate just by getting the quick fix of the. I totally agree. And I've heard that as well. But I think this is where there's talk of here's the list of things that local municipalities need to be asking for. The local governments need to be asking, you know, in terms of water and in terms of sand. and what kind of power, you know, you're not going to use our public utility around all of these things.

There needs to be more education on them. But to your point, yeah, I mean, these companies can just throw more money at it, right? They can just throw more money and say, here's more money. Yeah, our water is going to, you know, poison your well and we're going to use all of your water. I don't know. I think it's going to be one of the biggest issues in 2028. On the Democratic side, I certainly don't see, I see a lot of people talking about the issue and about the problem. I don't see a lot of people who have like specific ideas on kind of a regulatory scheme or

sort of laws that they want to pass that beyond kids in safety data centers. And like, yeah, we have to make sure that we don't have these AI systems like, you know, jailbreaking the sandbox and hacking into things. But we don't really know what to ask for there either. So it's like I don't feel like it feels like. people in politics in my world are going to have to get like smarter about what they're actually going to ask for and advocate around this because right now I think I think just saying it's bad and scary is probably not going to probably not going to work. Yeah and largely I think people

are already dealing with it again the like the amount of people locally that seem to rile around let's let's take down this data center or at least postpone it is is real it's true and real across the country. So last question it's been more than a year now like what of all the ways you used AI, the new ways that you used AI, what did you keep? And which did you say, like, never again? I know the, I know the AI boyfriend is up in the attic.

Which I'm working in right now. So, yeah, he's right in here. He's right here. He's right there. He's shut down now. Well, there's a couple of products. I mean, obviously the core stuff that I was using, chat bots, chat, chat, TBT, Claude,

and those things have gotten more and more. advanced and probably play more of a bigger, even a bigger role in my life now, too, because I started a new business and I'm really reliant on something to do some of the tasks. One of the big ones, and I think, you know, more and more I see more people interacting with it this way is the voice. We talked about it. And I was, I had challenged myself in 2025 to talk to AI at as many parts of my life as I could. It was one of the ways that I, in those beginning rules that, okay, if I have a problem, I have a question, I'm going to go to AI first. And a lot of times it was in places where I would have, it was easier to talk to AI in the car,

on a walk when I was with my kids and I want to just be typing to it. And so I really got this, what people in the industry called voice-pilled. I don't really like using this, these, you know, maxing and pilled. I'm not a, I'm not cool enough to say these words. But I definitely became very into voice. I've been using it. Now the models have gotten so much better. And so that is one thing that is absolutely stuck with me is talking to AI. I think the other one, well, and we talked about like I go to any city now. I'm in a waymo. There are certain little things that I had tested throughout the year that just that really stuck with me in those situational moments. But I also have now, what I didn't need to do last year was pull back on it.

And I think that for my own use, I've tried to do this in social media. media, I think maybe we all have tried over the years. Like, how am I going to limit my social media use, right? I have this device called the brick. Do you ever hear of it? I know. I have one. Yeah. So you know, I have a brick. It hasn't really worked for me, but yeah, it's there somewhere. I've been using it again this summer. It's helping me. So I have the brick to help me with social media, but I now feel like I need something similar for AI. And so I've been trying to really, I think this is something that is new, hasn't stuck with me. This is where I felt after the year and even becoming too reliant on it, going back to that cognitive, dissonant, you know,

the atrophy issue. So now I really do try to just say no AI right now. Like, I'm sitting down to write no AI. Or I want to think about this topic. I want to just go read websites. I don't want to read the websites AI gave me. Like, you know, I'm just starting scrolling. on a new project, actually, that we're trying to figure out for August. I don't want you to find me the links. I'm just going to go use Google for as long as I can because I don't think Google will exist for so much longer in the like blue link form. Yeah. No, I think that's probably right. Joanna, it was so great talking to you. The book is I am not a robot. Everyone should read it. It's fantastic. And where can everyone find you in your new venture? My new venture is The New Things, The New Things.com. You can subscribe to our newsletter there. And then we're really, we're really, we're. really building this video business. So YouTube.com slash Joanna Stern.

You can watch me wear the headset training robot thing. There you go. Perfect. Thank you so much for chatting with me. Thank you for reading. And thank you on behalf of all of us for testing AI for a year. That's some good work there. I won't be doing it again.

So, yeah. Take care. Thank you. Offline with John Favro is a Crooked Media production. Our show is produced by Austin Fisher, Emma I, Lick Frank and Anisha Bannergy. Our team includes Dallon Villanueva, Mia Kelman, Charlotte Landis, Eric Schute, Rachel Gaieski and Will Jones with support from Adrian Hill and Matt DeGrope.

Our staff is proudly unionized with the Writers Guild of America East.