How to win — or lose — Decoder

Support for this show comes from Doppel. Maybe that ping you just got is an urgent message from your CEO. Or maybe it's a deepfake trying to target your business. Doppel is the AI-native social engineering defense platform that's fighting back against impersonation and manipulation. As attackers use AI to make their tactics more sophisticated, Doppel uses it to fight back, from automatically dismantling cross-channel attacks to building team resilience and more. Doppel, outpacing what's next in social engineering. Learn more at doppel.com. That's D-O-P-P-E-L.com. Support for Decoder comes from Adobe. Life is unpredictable, and that means you need your projects to adapt with whatever gets thrown at you. That means mastering the ability to pivot and collaborate with others to reach your goals. Adobe gets that, which is why they made a tool that's just as flexible as you are. PDF Spaces and Acrobat. Your PDF files are no longer static. Instead, they're living documents that flex with you and your project's needs. Learn more at adobe.com slash do that with Acrobat. Support for the show comes from MongoDB. If you're a developer stuck fixing bottlenecks instead of building the next big thing, then you need MongoDB. MongoDB is the flexible, unified platform that gets out of your way. It's ACID compliant, enterprise ready, and built to ship AI apps fast. It's trusted by so many of the Fortune 500 for a reason. Ask any developer. It's a great freaking database. Start building at MongoDB.com slash build. Hello and welcome to Decoder, Nilay's show about big ideas and other problems. This is Nick Statt, senior producer, and I'm joined by host, very occasional guest, Nilay Patel. Nilay, welcome back to your own show. Hello. I hate being the guest. Now, you have said that in the past, but I feel like there's also a version of you that says that that is the ideal version of this show, where you just get to not do anything and show up and talk about stuff. So I feel like there are, you're of two minds with how the show, of like, what is the ideal version of Decoder? Being a permanent guest is a level of success that is hard to attain, where other people just want you to show up because they think you will be interesting. And I would love to attain that level of success. At the same time, being the guest means you also have to be interesting all the time. Being a host, you're just in control. You're basically saying, can you be interesting over and over again for an hour, and then you see what happens. That is my job today. A few months ago, we did our annual mailbag episode, which we were thinking of as like an annual thing that would happen around the holidays, where we respond to listener questions, feedback, criticisms, suggestions. But recently, I think we thought we should just do this more often because we get a ton of great feedback and we do read all of the emails. So we're here again. So I thought we would just jump into it. Nilay, you ready? Yeah, let's do it. So by far, our most popular episode of this year was also our most contentious. It was your interview with Superhuman CEO Shrir Mahotra. That focused heavily on Grammarly's expert review controversy. We got mounds and mounds of feedback about that episode. Most of it was overwhelmingly positive. There were a lot of interesting emails and comments and feedback we wanted to highlight here. Some of them were like, damn, Nilay's questions are making me nervous, was one of our top comments. Another said we need to make tech CEOs this uncomfortable more often. A Verge subscriber wrote in to say this episode was extremely uncomfortable to listen to and absolutely the reason I became a subscriber less than a week ago. So I think to kick this all off, Nilay, my first question for you is, how did you feel about the reception to the Superhuman episode? Were you at all surprised by any of the reactions? Yeah, I was a little surprised by some of the reaction. I can get into that. You know, as Nick alluded to, Shrir was booked to come on the show well before any of the controversy. And I was really excited to talk to him. He had been both the chief product officer and the chief technology officer at YouTube. He's on the board at Spotify. He was thinking about distributing AI through Grammarly. And distributing AI is actually a really hard challenge. You're up against Google. You're up against Apple, which is going to integrate AI into iOS with Google's models over time. So, you know, there's just a lot to talk about there in the creator economy and where AI is supposed to go and how it's supposed to work. And then this thing happened. And I give Shrir a lot of credit for coming on the show. He knew what he was going to get. It's not that we give people the questions. I think it was just obvious what I was going to ask about. And my feeling was that he could take the heat because he had these big roles at big companies. And I don't like taking, like, young founders and putting them on trial for the whole industry. But given Shrir's background, his depth of expertise, his enormous network, and his ability to just sit in there and answer the questions, I felt like we could do that with that episode. Because of who Shrir was, it felt like I could ask him about the specific issues in the case as a proxy for the bigger issues with AI. And I think a lot of people were responding to that. And so the thing that surprised me was the reactions that kind of felt like, you don't understand AI. This is just how it's going to be. You don't understand what being a builder is like. And I kind of get it from one perspective. But I think my response is, A, this is what Decoder is about. What are the consequences of building these products? And how do these products actually work? And how should they actually work? And how should we all feel about them? And my sort of more important response, if we don't ask these questions, if we don't ask them sort of relentlessly, then we will never make the people building the products actually think about what the answer should be. And that was really my goal. I know Shrir is thoughtful. I know he came on because he can take the heat. And I took the opportunity to ask the questions as plainly and as bluntly as I could. And maybe that made people feel uncomfortable. I feel like everybody in the room got exactly what they knew was coming. You know? And I think it was a service to the audience because that tension right now is reflected in every conversation about AI. Are these companies taking too much from us? Are they running roughshod over the laws we have to protect things like creativity and likeness and, you know, large bodies of work that authors and creatives and other people should be compensated for when you use them again? And we're just racing forward without resolving the answers to any of those questions. So I think we accomplished what we wanted for that episode. I'm not surprised at the reaction it got. I think the thing that did surprise me is that's what we do here on Decoder. So coming to the store and being like, I don't like the product you're selling is, well, that's what we make. And I hope we continue to make it. One commenter, Brendan G., actually wrote in about the superhuman episode. He does media training professionally. And so he obviously hates listening to media-trained executives. But he said either Shrir's media training was really good or he was just smart enough to ignore it and decided to have a real conversation other than occasionally hiding behind lawyers. Brendan also said that from his perspective, it felt like you spent a lot of time grinding what felt like a personal ax. You sounded angry, although he doesn't know if that was a kind of performance that you were doing. And he said from a media trainer's perspective, I would have loved that because it just makes Shrir, in this case, seem reasonable or calm. So the question for you, Nilay, is how did you decide to approach that interview? And did you think of it as you having to kind of play a part on behalf of the people whose likenesses superhuman had appropriated? Or were you just, you know, was your strategy just, oh, I'm going to just nail him on this one part around, like, how much he owes me and then we're going to go from there? It's very rare that the story is actually about me. It's not a thing that occurs very often on Decoder or on The Verge as a whole. And so this was one of the rare times where I was just in the story. Just straightforwardly, there was an AI clone of me in their product. And that felt like I could make the story more human just from the beginning. I didn't have to explain how it would affect regular people. It was just very obvious how it was affecting me. My feeling was that by just letting the story naturally be about me, which I don't like doing and which I think no journalist likes doing, but by letting it naturally be about me, I could make the stakes of it plain. Right. And I think a lot of people who felt themselves reflected in that story, a lot of artists want to go up to a CEO and say, how much are you going to pay me? And very few of us will ever get that opportunity. And this was just one of those opportunities. So I took it. I think the anger piece is really interesting. And I do think that is because it was me in first person talking about myself. I didn't feel angry during that interview. I certainly have a temper. It rarely comes out. Um, but I, I didn't feel anger. What I felt was intensity. And I think those things are a little different. The weirdest thing about doing an interview show is that the episodes are only good if the other person is good. I can't make Sarah Personette understand her business more than she does. I tried, man. I don't think she understands it at all. Not even a little bit. And the questions I was asking her, I don't think were particularly adversarial. We got off that recording, and I think it was Kevin McShane, our editorial director, who said, I don't think Sarah realized she's on the same side as you because she was in outer space. I'm not gonna back off on, do you understand the basics of your business? That seems like totally fair game to me, and I don't think she does. I also feel like, you know, Worcestershire is a good example. I knew that he was prepared. I knew he had the experience and the history. He could do it. If you're the guy who runs product at YouTube, people have asked you a lot harder questions. You could face a lot hotter fire than I can provide to you in a one-hour conversation. So there's a spectrum here. And I'm just gonna flat out say it. I thought Sarah blew it. I thought that was one of the worst performances on a show we've ever had, and I think you could tell about halfway through that episode that I was just like, do you know anything? Maybe she does. Maybe she just didn't know what show she was on, and she wanted to give her TED Talk, and she got derailed, and that's that. On the other hand, I feel like if I do think you know what you're saying, if I do think you have the depth of understanding and you're ready for it, then the pressure should only escalate. And so this is, maybe it all feels the same in the end, but to me, just sitting in a room, they feel like very different vibes. And that's what I just want to do. Like I said, to make an interview show, the other person has to, A, they have to want to show up. You know, we always say that Decoder is a game you can win. They have to want to be here and participate honestly and openly. They have to think that they're going to come out the other end, and they won't feel completely attacked, because otherwise we won't get guests. Like they have to, they can just hang up. They can just click the button and go. So the show has to be an environment that reflects and respects the participation. At the same time, if it's a game you can win, it's also a game you can lose. And I think we're just seeing that dynamic. I think everyone is very used to very puffy influencer interviews. There's a lot of that going around lately, and maybe everyone should just be one more turn more prepared. There's a real like hunger from the audience for what you might call accountability journalism, especially in podcast form and particularly with the current tech and AI industries. The joke that you've said before is that the audience wants you to end every episode by arresting a CEO. And we've even had some commenters now referencing that as an editorial strategy. Some people are saying, you know, they want you to be tougher even. But this is running headlong into the idea that companies don't necessarily want to do these kinds of interviews all the time or even often. And that people don't like being put into unpredictable situations where they don't know the questions, they don't know what you're going to ask. And then audiences themselves, are not really even all that interested sometimes in that kind of product, like the end result of that. Like Diary of a CEO is not hard hitting journalism, even though it's very engaging. Acquired TVPN is certainly not journalism. And whatever monitoring the situation from Andreessen Horowitz is not journalism at all. But I think we saw a version of this kind of play out recently with NVIDIA CEO Jensen Huang and dorkish Patel, right? Where he's not a traditional journalist, but he asked some challenging questions and it kind of broke people's brains because they weren't used to seeing somebody kind of check somebody like Jensen, a rich tech guy, but also like head of the most profitable, head of the most valuable company in the world. And people kind of think he's unfallible. And this was a rare moment where people were kind of divided on what's the purpose of this interview. So Nilay, the question for you is, how are you thinking about that tension between the functions of journalism, what the audience wants, and then what the audience actually responds to when it comes to tough questions? And also, you know, why do you think people are coming on this show when they're, you know, we're not paying them, we're not telling them what we're going to ask beforehand, even though they know what to expect, they are going kind of, they are, they are going kind of blind. My favorite is when people show up and they're not ready to be asked what the structure of their company is or how they make decisions. I feel like those are gimmies at this point. And every now and again, it's just like, oh, you didn't know. You can always tell how things are going to go when those questions seem like a surprise. I think journalism is critically important. Obviously we make journalism here. All of us who are making the show right now are journalists. We're steeped in it. Maybe we're just high on our own supply and the platforms are going to kill us all in the end, but I think it's important. And our audience, when we, even you can see it now that we like make more clips and put them on social platforms. The audiences who've never encountered us before because the algorithms are just taking the videos to wherever they go. They're like, oh, I love this. Like, finally, someone's asking the questions. And that is remarkable. It's refreshing. I don't know what it is, but there's not a lot of that that people see. So that's the product we make. It seems to have found some audience. I hope we continue to find more audience and we can all keep doing this because I like making journalism. I know why people come on the show. It has finally clicked for me. I've had a lot of conversations about it. You know, Nick and Kate, our producers will tell you, we don't do a ton of outbound booking anymore. We have an incoming list. It's a mile long. People want to be on the show. And the number one reason that I hear is that all of these executives know that their own teams aren't going to listen to the audience. Like their own teams aren't going to read the emails. And it is good for them as leaders to go get the external validation, not their own comms stuff, not their own branded content, not their own fake TED talks. Some of them do fake TED talks, which is wild. We've moved on to fake podcasts. We've gotten pitches for me to do a fake podcast that will then be clipped into like fake podcast clips. And I'm like, I don't need to, I have a real podcast, but this is the market that we live in. And everyone can see through it. So if you can come on this show and explain your company well, explain how you make decisions as a leader, explain how your company is structured, take a little heat, be asked some challenging questions, do a good job. It is actually good for those folks out in the world with our audience, our actual audience, which is big and growing, but it is also good for them inside of their companies. And like I said, you can win that game and you can lose that game. It's the fact that they are not in control that makes the thing valuable. And that external validation is so important. I look at TVPN and congratulations for selling a podcast to 70,000 YouTube subscribers for $200 million. Like, that's great. It is very engaging. I've watched a lot of it. They were inside the industry. They're unapologetic boosters of the industry, and now they're inside a company in the industry. They have no ability to provide external validation. They can't, they've lost the thing that might provide conflict, and conflict is what drives all great stories. Andreessen Horowitz has started and failed 10 million media brands. They had a tech blog called Future that was just about how great everything was, and it failed because no one wants to read it because conflict and emotion is what drives stories. And you can't get that if you're inside. If you are working at a place where you are not allowed to criticize the people who work at your own company, you are never gonna write a good story about that company. You can write great press releases. You can write engaging, like, here's how the factory works videos. But then we're gonna come along and we're gonna say, hey, your robot Apple that you said can recycle all the iPhones, there's only one of them and it's not recycling any iPhones. Like, you can see this pattern happen over and over again. So I know what role we play in the ecosystem at The Verge and on Decoder, and it is to be outside. You have to show up here on our terms and do a good job. And we have a big audience, and if you do a good job, I think the audience will be excited for you. If you do a bad job, I think the audience is gonna let you know it. And that is, that's hard to get. And, you know, we're also precious about all of the rest of it. We won't do brand deals and integrated sponsorships and all this stuff that compromises that core promise that we make as journalists. I talk about that stuff a lot. I don't need to overdo it now, but that, to me, is, that's why everybody shows up. It's hard to find the thing that we make anymore. The producers and I will not give anyone the questions that they're going to face on Decoder in discussions like whether the industry is a bubble or whether there's mainstream appeal or product market fit for this technology. How is that thinking evolving? I have really mixed feelings on how to cover AI, and it is really related to all of the polling we're constantly talking about, where regular people are encountering it more and more and they're hating it more. And they're not being shy about it. And I really take to heart that Decoder is the business show that sits on top of a big consumer tech website. So The Verge as a publication is very much for consumers. That's what we cover here. We don't do a lot of enterprise tech coverage on The Verge. We focus relentlessly on technology and how it makes regular people feel. Decoder is a business show, right? I'm asking CEOs what their org charts look like. That is very far from anything any consumer cares about. And I think understanding how the companies and the people think really helps you understand the products, so that when we do the product coverage, we get a really interesting feedback loop where I understand the businesses that built the products. And I think that's reflected in the products. And then I can come back around, and you can redo it on Decoder all the time and say, also, we run a giant reviews program. We use your products. And I think your products are bad. And that is—it's hard to find that dynamic anywhere else. So I think that's honestly what makes The Verge unique and what makes the relationship between Decoder and The Verge unique. Specifically as applied to AI, I think for a long time, we were using the products, and they just couldn't do the things the company said they could do. You can use free ChatGPT all day and all night, and if you have an ounce of self-reflection, you will say to yourself, this is not alive. It's just prompting me to ask it another question at the end of every response. And I don't see how you get from here to this thing can run an entire business to this thing will attain sentience to this thing will be AGI. You can just look at the product and see it doesn't work. David Pierce recently just reviewed the Starbucks integration in ChatGPT, and the thing is a miserable failure. Okay, we can just look at the products and see what they are and see the promises these companies are making and ask very directly, are those promises being kept? And I think on the consumer side, the answer is manifestly no. They cannot do the things they've promised consumers they can do. I think that is very much why consumers are turning on AI. They're not getting the value, but they're getting all the demands. The thing that has changed, and I think this is the reason the feedback is getting mixed, is on Decoder particularly, we have a business audience. It's a business show. It has a business audience. And there is real product market fit for AI in the enterprise. You can see what Anthropic's revenues look like. You can see OpenAI basically sloughing off every consumer thing it was doing, including Sora, and trying to focus heavily on codex and enterprise use of AI. And there's a lot to be said for that. I think a lot of business processes should be automated. I think having agents run around and do things inside your business so that real people can do actual tasks of higher value is—that's great. I think the cutting edge of marketing is automation in some way. I think it's going to be really weird for a lot of people, but it's happening. And you can't deny that it's happening. You can't deny that AI has found uses here, and some of them will fall flat, and some of them will succeed. And that will be really interesting to cover. So that's where I think the mixed opinions come from, is that if you're looking at one part of the market, you say, oh, AI has a lot of value to offer here. But then you kind of take the jump, and I think we've recently heard Jensen Wong say AGI is already here. Jason Kalkanis has said AGI is already here. And what they are describing is, it can write software, it can automate some business process, which means maybe you can run a company all by yourself. AGI is here. That's just pure nonsense to me. And I think the thing that I'm looking at a lot is, where is the product, the AI product, that people love that actually changes their minds? And to me, that product doesn't exist. So I think we're going to hammer on that divide pretty hard in the years to come here. That relates to a comment we got from a reader, Chris. He says he thinks the AI polling is bad, Schick lately on Decoder and Vergecast is underrating how much one, he cannot trust images or video anymore, two, this is really bad right now, three, it's easy to understand that it is genuinely apocalyptic in the near future and apocalyptic seems large in American imagination, and four, those bad outcomes are the fault of gestures broadly toward AI. So he's saying not only is there no good consumer AI product, but that the consumer AI products that do exist are a threat to the social contract in real and immediately obvious ways. Obviously, you mentioned the AI polling around Gen Z. It's manifesting in some very dark ways. There have been attacks on politicians, attacks on Sam Altman's home, a lot of pressure mounting against data centers pushing back on AI executives claiming that they're going to create more jobs, not destroy them, and then some AI executives, of course, just plainly saying, we're going to just destroy all jobs. How is this affecting how you think about talking to people about AI on Decoder, particularly tech leaders and people who are working on this technology? One, I think I want to make sure I keep asking them if the technology as it's constituted today can actually do all of the things they say it's going to do. I don't think that answer is clear at all. You can listen to Jan LeCun, who used to be the head of AI at Meta, who got pushed out of Meta for saying he didn't think LLMs could get to AGI. He's still out there saying it. The latest argument that I've heard him make is you can't have an agentic system that's taking action for itself when it can't know or predict the consequences of its actions. And that's just sort of the nature of the LLM. It's going to do stuff and see what happens. But like true intelligence is going to take repeated actions in a way that is predictable. It's just like you and I would take actions and know what's going to happen next. The LLMs are sort of reacting on first impression all the time. That's a big conversation you can go have. And maybe you can build some affordances to get around that sort of inherent fact of an LLM. But I think there's a bigger debate in this field that anyone wants to acknowledge because the market opportunity for the tools we have now is huge. So you have to say it's going to do the next thing and the next thing and the next thing. I want to keep pushing on that. I don't think that is settled at all. And I think making people say out loud what they actually think the technology can do and what its limits are is important. The second thing I want to make sure we keep doing is talking about the polling, talking about the fact that this industry is demanding so much from everyone. All of the power, all of the land, every stick of RAM in history. For what? It really cannot be we've automated marketing. It just can't. It has to be something better than that. And I keep saying it, and I know people argue with me in a million ways about this, but ChatGPT has what, 900 million weekly users? Gemini is everywhere. If you just like blink at a Google product. Claude is famous now for a lot of people because it is now also a political story. Everybody has seen slop on their Facebook feeds. People are aware of this technology. They have made up their minds. You cannot market your way out of this problem. You cannot advertise people out of their honest reactions to what you're putting in front of them. And unless you have a product that can overcome it, I don't think you're going to change the hearts and minds. And there is not a product that regular people are using every day that they feel love for that overcomes this. I can give a lot of examples here. Uber. You can list all of the policy criticisms people have had with Uber for years and years. There are labor concerns with Uber. There are safety concerns with Uber. At one point, Uber was getting banned in various cities. People really liked the product. They were able to overcome it because the product was compelling. And drivers liked the product, as Uber will tell you over and over and over again. Some drivers don't want to be full-time employees. They like the flexibility. To the point when Uber had regulatory problems, they were putting ads in the app, asking people to lobby their local politicians. This is a product that was compelling enough to make people take political action in a way that AI is a product that is anti-compelling enough to make people take anti-political action. And there's a long list of products like this. You can overcome the policy objections, the societal objections, if your product is compelling. And I just do not think there is a consumer AI product that people feel good about at the level that rises to the kinds of demands this industry is making. And that you can't be like, this is great for business. I don't think that's going to do it. Okay, we've got a few questions left, Nilay. We'll do these in more of like a lightning round style about the current structure of the show and what to expect in the future. One question here from Joe Rodericks is that he really enjoys the occasional episode where Nilay is really fired up. He says, I would love for you to consider wherever you get your podcasts. Decoder is a production of The Verge and is part of the Vox Media Podcast Network. The show is produced by Kay Cox and myself, Nick Stat, and it's edited by Ursa Wright. Our editorial director is Kevin McShane. The Decoder music is by Breakmaster Cylinder. See you next time.