He Built an AI Podcast and It Became the #1 Show.

All right, before we start this episode, I need to play something for you. You are listening to the Epstein Files. The world's first AI-native investigation into the case that traditional journalism simply could not handle. Welcome to the Epstein Files. Last time, we traced the Dubin family's post-conviction ties to Epstein, a 2010 social invitation, subpoenaed financial records, and victim testimony naming their household. Today, we are looking at official White House visitor logs, showing Epstein entered the Clinton White House at least 17 times between 1993 and 2000, while his abuse. So which is as heard is not hosted by humans. There's no studio, there's no mics, and there's no real host to this. This podcast was written, voiced, and published by AI automatically based on whatever the news was that day. And it did it yesterday, the day before, and it'll do it again tomorrow. The show is called the Epstein Files, and in its first week, it had over 100,000 downloads. And last week, only a few months later, the guy who built it woke up to a notification. His show had become the number one podcast in the UK. He tweeted the screenshot of the chart with three words, just getting started. Only a few months after launching the Epstein Files, he launched a second show, called War Desk, a daily AI-generated breakdown about global conflict. We covered March 15, 2026, day 16 of Operation Epic Fury. Tonight, we're looking at what changed on March 16, 2026, as part of our ongoing investigation. As always, every document and source we reference is available at War Desk. It's been live for about two weeks, and it's already crossed 33,000 downloads. Now, just to give you a sense of how crazy that is, the average new podcast gets 141 downloads per episode in its first 30 days. War Desk is averaging over a thousand. That's top 5% on a two-week-old show. So the creator of the Epstein Files and War Desk is a guy named Adam. And we mentioned his show on a podcast a few weeks ago. So he reached out to us. I've been so fascinated by this concept of an AI-generated podcast, and even more so, by an autonomous AI-generated podcast. So I had no choice but to reach out to Adam and have a conversation with him. Now, just as a reminder, as you get into this episode, I've been a career creator. I've been making stuff for 16 years. I went to film school prior to this. My intention has always been to be a career storyteller. And things are changing. I think it's incredibly important for us to understand where the media landscape is going. And talking to Adam has really opened my mind and showed me where I think the future is going. So while this conversation is about podcasting, I think it has implications for all media, all types of creators, and everybody who's working in entertainment. So here's my conversation with Adam Levy, the creator of the AI podcast, the Epstein Files. Adam, welcome to the show. Thank you. Thanks for having me, Samir. I would love to hear the story of how this came together, how you thought of it, what it took to make it, everything. Yeah. I think it's important to preface that I was not expecting it to do what it did. And I really just started, I put my best foot forward of trying to just see if it was possible to do it. And lo and behold, it kind of just like took off the way it did. So completely utter shock to me. And it's completely changed my approach of what I think is next, moving forward for synthetic media, AI media. And yeah, I guess I could start with how I got the idea, maybe that's a good place to start. I like to say that there's life before Opus, 4.5 and 4.6, and there's life after Opus, 4.5 and 4.6. What is Opus for the audition? Yeah, Anthropics Claude Model. Got it. During the holidays is when I really picked it up. I think a lot of us, like we saw the entire Twitter timeline shift from our traditional timelines to become more AI-centric. I don't know if you saw that as well, some of you felt that, okay? And I built this one app for myself called Distill. And Distill at the time, it lives on my phone. It's not public anywhere. It's just what I built for myself. It basically takes a bunch of sources. I could take a bunch of URLs from TikTok and Instagram and any video platform. Put them in this one aggregate text box. And it basically extracts the transcripts and distills information for me, facts check, facts checks it, and allows me to create an audio version of the takeaway for me to learn and to study. So I learn best through audio. I get very complacent when it comes to reading long-form content. I like to listen on 2x and think about it, reflect on it, and understand what the takeaways are. So I'm a very like auditory learner and a visual learner. So I built this app for myself to be able to do that across any type of content, any topic for what it's worth. And as the AppScene files were coming along, I was like, oh, this could be a really interesting problem to solve. Let me just try to take as much of the data, of the 3.5 million files that were published. Try to distill it and contextualize the data and create this true crime-like story around all the files, starting from who he was, who Jeffrey Epstein was, and how he got started to the business model behind what he was doing. And yeah, all the public figures that were named, I mean, like everything in between, it still publishes an episode to two every single day. I'm now at about episode, I think, 150, or something like that. You're on an episode 122. 122, there you go. I'm losing track. Yeah. So just to clarify, the AppScene files is a self-generating podcast that uploads basically every single day. A new episode based on what is happening in the news about the AppScene files. Is that correct? That's correct. Yes. It's either in the news or there's millions of data points that the news has yet to still open and uncover. And I've reached conclusions faster than these mainstream media outlets have reached, because my speed of execution and my cost of producing a piece of content is significantly less than these big production houses. So it's funny, because some of the headlines that I end up creating episodes around, they only came to market maybe two weeks afterwards as well. So I have this elaborate content outline that produces itself, as well as the product is malleable enough for a hidden ingest breaking news. New developments, maybe something that I missed or something that I saw on Twitter, for example, that's a really interesting insight. And then I fact check it and make sure that it's true for what it's worth and then it goes and produces an episode autonomously around it. So how much are you manually doing at this point versus how much is it autonomously doing? Is it pulling those sources itself right now? Or are you inputting the sources for it? I can either spin out a particular direction and then it will know what to do from there. So if I see something on the timeline, for example, and I want to cover that particular moment in time, I could either take a screenshot of that or paste URL and it will know how to enhance it and enhance the all the sources that it basically comprises to then get to the final product, which is the audio itself. Or it just constantly just produces new content based off the outline that it has in front of it. It runs itself. It's a thing of it like this. It's like 95%, it's like 95% autonomous, 5% at this point. And is any of the 5% on the publishing side for you? Do you listen to the episode before it goes out or you let it publish and you listen to it at the same time as everybody else? As part of the bowling guardrails, sometimes it does require me coming in and listening in case there's a flag of sorts. But most of the time, there's QA checks that are programmatically done with a particular mathematical score that it needs to hit for it to then pass a series of checks before it can then publish what's next or publish on Spotify, Apple, etc. Yeah. How many people are listening to this right now? It means it just crossed, last week, across 2 million downloads, every episode gets a minimum of 10,000 downloads within a few hours. It's pretty crazy. It's picked up a following, like a cult following people have built. I look at the retention data. People listen to so many episodes before they drop off, before they cut their session. And then they revisit as well. There's a lot of really interesting data that Spotify provides as you know. You guys have in your channel. And I'm still constantly shocked that this is the case because it's all synthetic content. You know, it's all AI content. But it's done, I guess it's done better than a human could do it for what it's worth, given the amount of data that is required to understand what's happening in this context. I mean, this is such a unique use case for AI in a way that humans could never compete with. If you look at the reviews as well, it's gotten over 1,000 reviews at this point between Spotify and Apple. People either love that it's AI or people hate that it's AI. Sure. And the reviews themselves are really fascinating to read because they like that there's information without the spin. They don't like talking heads. They like facts, not the bullshit of commentary and conspiracy. I try to avoid conspiracy at all costs. I like to document facts. And there's so much data to then draw conclusions from the facts that are presented from the emails, the photos, the videos, the PDFs, you know, the court documents. Like all these things that that within itself is the story. You don't need to have a conspiracy beyond anything what's already enlisted and shown to us and given to us. So I try to curate with that intention, you know, and maybe that's also why people have found the liking to it. Maybe there's a lot of conspiracy content. It's really hard to tell what exactly is the real, you know, like, oh, it's because of that. It's to why people like it. But yeah, yeah. I mean, look, I think for a guy who's been in content creation for 16 years, you know, as we've seen historically, a daily habit is also one of the stickiest products in media, right? Like it's why The New York Times has the daily. It's why Morning Brew's new show Morning Brew Daily is picking up steam and rising on the charts. It's why subway takes on Instagram has become an incredibly popular show. It's daily. Something that you do every day as a habit or a ritual is an exciting thing for people. And it grounds you and it roots you. And I think your ability to turn around an episode every day, like you said, only few human creators can do that. There was a time where Casey Neistat did it every day for 800 days in a row. The most culturally relevant creator on YouTube when he did that. And I think we can't overlook that daily programming and daily habits are a major part of human connection. But I want to talk a bit about the commercial side of what you're doing as I bring that up of like two million downloads. You've done a hundred and twenty one episodes. And we've been a couple months like this is a breakout pod. Truly like a new podcaster trying to do that would be very challenged to try and replicate something like that. How much would you say it costs you to make this podcast? And has it made any money? Are you a part of the Spotify ads program? Are you selling sponsorships? Are sponsors interested? How would you approach that? Tell me everything about the monetization of this. The cost to produce an episode is significantly cheap relative to what it would cost. What would it cost to produce a traditional episode? Your main cost is the tokens used to create the episode. And regarding monetization, I've gotten a bunch of inbound from different types of companies that are interested to either figure out what long tail sponsorships could look like. Data integrations, you know, just like ad lives and running, you know, traditional pre-rolls and mid-rolls. And I've intentionally avoided monetizing the podcast just yet. I've intentionally avoided monetizing it and I'm using the traffic to grow other IP right now. I also think there is something to be said around ads diluting the quality of your listening experience. And there is something that I'm testing in particular to see around synthetic media. Even in the context of what I'm creating, I want to see what retention looks like for stuff like this. And there is more to understand before jumping into monetizing directly. And I know that I'm so long-term focused. I know that if I get this right, you know, and I do it correctly, then there's a whole ocean available to go after. So let's explain some of what you just said. Like number one, you're using it to promote other IPs. That is the example of that your new pod, the War Room about. The War Desk, yes. The War Desk, sorry. The War Desk, which is about the War in the Middle East. Correct. Correct. Yes. Because I could get really weird as well. That could be very political. The data could be, you know, it's like you want the nonpartisan, no bias, no spin approach, data first, approach to documentation, something that is super heated. So trying to leverage that traffic. I'm working on a couple other series right now outside of War Desk and the Epstein files. So my goal is to build a net online of all this IP and yeah, build a next great media company. That's where my vision is at doing that. So the potential future is a podcast network with a bunch of these podcasts that are self-generating autonomous posting nearly daily, if not daily. The very minimum, yes. At the very minimum. And the monetization of that in the future is advertising or actually pushing traffic to another place where people could transact. I think once you own the attention and you build a dedicated audience, you could do anything with that attention. The very minimum, there's advertising opportunities to, you know, you know, you know, obfuscate the burn that you have producing this content running the business. So at the very minimum, it's as I think there's also a narrative to be said around gated content. I'm really intrigued by the Netflix model of building a lot of IP, building a subscription base around it, creating a really intimate experience around consuming the content. Today is going to be the worst in which these models perform. They're only going to get better from here on out. So yeah, I think there's many ways to slice and dice is I love building product as well. So I'm a very creative person. I love creating my imagination and it's never been easier to create your imagination. So I think that's the only way to create your imagination with these tools. So who knows what all come down across the line or products that can build leverage that traffic? To yeah, grow out the pie for what it's worth. So when you say the cost is tokens, like could you estimate on a monthly basis, what you think that is based on the tokens you're using to produce? I would need to, I don't know the exact cost per episode. Okay. Everything and everything and be a super user of all the AI tools, understand them and be the subject knowledge expert and how to leverage them, reverse engineer them, get what you want from them and with that comes a premium overspending on them as well. And so I'm only a month into this, like literally a month into this, you know. So in any traditional sense, like you measure the output of a business, you look at it's monthly burn, it's hard to tell exactly what the cost per episode is so early on in doing something like this. But I know for a fact that once the model, you know, progresses itself and you have the exact template cookie cutter approach, then you can get to the unit economics more clearly. I would have to assume though it's probably safe to say it's in the hundreds of dollars at maximum in the thousand dollar a month range. Sure. Sure. I mean, that's insane. Yeah. To make a good podcast with two hosts, there's no possible way a media company would spend a thousand dollars a month to do that. It would cost exponentially more. Yeah. And when you think about like even market CPMs for podcasts, even at like the most simple programmatic setting, you'd be profitable immediately. You know, at a $25 CPM or anything like that. Sure. So yeah, that's it. It's exceptional to me to hear the two million download number. And I want to know what you've learned from the retention data, what lessons you've taken from the fact that there's real listenership to this. And then also if the models are reading the data and adjusting the next episode based on the data from the previous episode. To your latter question, yes, they do. The benchmarks in terms of retention, listen links, they're either at the industry benchmarks or higher. And the episode consumption per sitting. I need to get I need to understand. I don't remember exactly the particular detail, but it's significant in that I remember doing you know, like comparative research to industry benchmarks relative and it was higher than what people typically consume traditional podcast series. And yeah, that's why it's important for me to not get to caught up in the Epstein files because it's such an inanomally of a topic in world event. And for me, it's important to prove that I didn't get lucky and that there's actually something here. And that's the ambition about doing War Desk. War Desk at the end of this month is going to cross 100,000 downloads. And for a channel that has never had any content, any audience, it's a brand new Naked channel. You know, it's already performing above industry benchmarks for a brand new channel. So my focus is like nothing is as good or as bad as it seems. And I'm using that energy, you know, to just see things for what they are with the Epstein files and translate them into other IP. So if I was to sit down right now and try and vibe code a podcast, is my initial prompt just I want to make a self generating podcast about X. No. Help me. No, it's not. What is what was your initial prompt for the Epstein files or how do you how do you start that conversation? Great. So when I was able to contextualize, you know, the emails, images, videos, PDFs, all the court files, whatever and get like a similar format across all the data, you then kind of you have to, you can't feed all that. data to the LLM and will break the LLM. It has a context window, a token limit window, you know, up until certain time, there's 200,000 tokens and now it's a million tokens, whatever, not important. What is important is that you kind of have to analyze the data and chunks and then from a content perspective, kind of kick it off with like, what's the storyline behind all of this data, right? And then from there, you go down a rabbit hole of understanding, okay, this is a storyline, these are the locations, I'm going to have an apartment that was purchased, less Wexner. You have the island, you have his upbringing, like you have all these different touch points. In the AI, you could do a really good job of linking the data together, you know, so then you kind of, from there, you propagate what's the storyline and even that is not enough to really get to the final product as to what I have right now. You know, there's an entire engine that was built now autonomously, but the seed in which that, you know, planted the initial path for creating episode one, for example, started from that, from that first, first, first story, yeah, first approach. Do you think a guy like me who's proficient in like storytelling and media creation, but literally knows nothing about code and software development is capable of doing what you just did or do you think that you have to have like you have a background and software development. So you have some ideas of how this model is going to interpret what you're doing because what you just said, I do not understand. So I understand it human language if I'm like, Hey, I want to make a podcast, but it sounds like that would be too simple and I would have to have a deeper understanding. You would have to have a deeper understanding. If you want to replicate the Epson files, you can't just prompt chat GPT, Geminier, Claude, like, Hey, I mean, it can. Gemini will be able to scrape the internet for different sources and it can give you a topic outline, but okay, great. You've produced an episode. How do you build the channel around that? You know, what's the workflow for productizing the entire system? How do you ensure that every episode gets better with the, with another generation? Right. You know, there's a bit more than just create me this episode. Making no mistakes. How many days and did it take you, you know, how much time did it take you to get that first episode up and then, you know, but what do you think like in the time was from idea to publish? It took me two days, 14 to 16 hours each day to get 64 episodes out. And all those launched at once, 64 episodes launched at once. Correct. And then it starts self-generating after that. Correct. Yeah. And how quickly did the listenership come? Within seven days, 100,000 downloads. Yeah, it kind of blew up. I mean, you find it. You think people have searching the Epson files? I think people are just searching Epson. Yeah. People are just searching Epson, EP, Jeffery Epson. Probably Trump, you know, was probably in that mix like any keyword that was tangential to Jeffrey Epstein and the people that kept going viral tangential to Epstein on the timeline. Yeah. When do you think Spotify or do does Spotify get involved by making originals like this? Based on the data they have of what people are searching, based on the data they have of what people are listening to, feels like they would be in a better position to know what's going to hit. Do you think they get involved in this type of? Yeah. I think they could create their version of Amazon Basics. I think that's a good analogy. Like Amazon has all of its merchants. You know, they don't want to cannibalize, you know, all the sales, but they'll produce their own product lines based off what, you know, where they can compete on margin and where they can compete on, you know, product for what it's worth. So I think Spotify will have its version of Amazon Basics. Sure. Same with Netflix. Yeah. I mean, it's a different model. Yeah. Like Netflix buys IP, you know, at a very, from what I understand, at a very aggressive model and kind of screws over the actual producers and directors that bring these stories to life. But that's just because they have the distribution and the power to do so. So maybe with the new, I think there was like a north of $600 million, it just spent on it. Was it Ben Affleck's company? If I'm not mistaken, to who know, it's really unclear what his product, what his product does. There isn't much documented on what the value prop was, but I assume it's a video model of sorts or a process on how to produce films using AI. Yeah. And they're for sure also looking at their inputs and their outputs and identifying bottle next in their system and trying to see where they could optimize it. So, yeah, I think there's only a matter of time until all these production houses and streaming sites, you know, they produce their own stuff. Now, I think they're going to need the right teams and the right technology to do so, which makes a great venture outcome for building in this category as well as an entrepreneur. And I think that's a really interesting opportunity. Yeah. So it's only a matter of time to do this content. It's interesting because the creator economy disrupted, you know, the legacy entertainment industry based on an incredibly lower cost and the same concept you're saying. We're able to iterate faster. And it feels like we are now getting disrupted in a way by technology and the ability to generate entertainment news content through AI because we can never compete with that. A team of creative people can never compete with the cost and speed of what you're doing. But the team of creative people could use these tools to then build and compete in any other media. Sure. You know, it's like that's the key. It's like if you have the ideas in your creative enough to not be, I guess like stronghold by these tools where you can still think for yourself and they're not thinking for you, you know, you're still the orchestrator behind this entire thing. Then it just goes back to what I said earlier. It's like a 10 X's, all the areas that you otherwise were previously strong at. Now you could just do it 10 times better, 100 times better. So I still think the creatives, they still have the opportunity over here to be at the forefront of this. They just need to use their creativity to execute and use these tools to help enhance everything that they do. That's just what they're designed to do. Yeah. I may be jaded with my with my perspective on that, but I really believe that. Yeah. Now I think that's fair. I think it's not going to look anything like it looked like before the same way that there was a reduced team going from making a movie to making a YouTube video. There will now be a reduced team or options for an incredibly reduced team. I do think one of the most fun parts of being a creative part that I personally enjoy is collaboration and working with other people. I think it's very fun to get excited about something with other people. But I am finding that like if you look down into the future, there's going to be a lot more single person media companies like yourself. The first you build future for you, this podcast network that you build could be doing 10 million downloads a month and it could just be you. And that can be by choice, but truly by just necessity, it may not need to be more than you. Totally. So yeah, I think that is going to be something that we're all up against. Like I think overhead is a major challenge in creative companies because it can be at odds with creativity, meaning my financial burden and my necessity to create something can be at odds with my own creativity. So that will be a consideration. But I'm curious if you have any advice for new media creators for us, for the crop that is making YouTube videos, making podcasts right now, making stuff on Instagram. Like what is the advice to us for the future? Sure. When I was publishing 40 to 60 pieces of content a week, it was Twitter, TikTok, Instagram, it was the newsletter, the website, long form audio, short form audio. At some point when you're doing contentful time, you build an assembly line around what you're doing. You build a process and operation a system. And within every system there you have your inputs and your outputs. And you need to understand what part of your system contains the bottlenecks that lags you the most. And AI is like the perfect opportunity to optimize those systems. I think that's the meme that you see online with Palantir. It's like they go into these US government, they figure out the problems, they optimize them using AI. I'm obviously saying it very lightly, you know, like stripping a lot of the details. But that's like the analogy and that's how I think about it. So I'm a person of systems and operations. So I think about my bottlenecks and how I can improve them. That's number one. Number two, be the very best at these tools. Subscribe to anything and everything. Download the open source models, get yourself a better computer and better equipment. If you can afford it, download these models locally, experiment and try to take them to the very edge of what they're capable of doing and compare it to the other models. So if you're particularly focused in video, be the very best at all the video models, be the expert in the room to the point where you can host a three hour seminar on all these models, right? So that you know with the skills that you have now, how that could be shape shifted and applied into what you're doing today. And then challenge yourself to create a new form of art through your content. If you're a great storyteller, try to use these tools to prompt out a 15-second story of video. an audio, whatever it may be. And just like always stay curious, just like be at the very, very edge. Because if you're able to stay at the very forefront of what you're doing, then I believe you'll never be replaced. And you'll always be ahead of the game. And that's what I practice, at least. You can only speak from my perspective. Yeah. I like that. Yeah. I think the only thing I would add is like the acceptance of the moment that this is happening. And I think that is hard. It's hard to accept because we are inherently, I think, our desire is always to preserve as humans, like preserve what is now or what is the past. But the train is moving and it's moving fast. And like I said, I think it opens the door for a lot more honest creation, which I love, and a lot more exploration of what this type of technology is capable of. And I did not touch vibe coding for a long time. I didn't really have that much of an interest in it. And once I started, I recognized that it is satiating a lot of creativity that I didn't know was unsatiated. And so I've really enjoyed my time exploring what's possible. We had a conversation with Mark Zuckerberg about a year ago. And he was talking about technology. And he said something that's always stuck with me that even though we have iPhones, we all still have TVs on our wall. Meaning the new doesn't necessarily replace the old always. There's typically actually more room than you think. We always get nervous that like this new technology will replace the old technology. But oftentimes what happens is actually it's just in addition to when I think about what's happening here, I don't know that we've ever experienced the exponential nature of this type of technology. Would you agree with that? Or do you think this is comparable to previous advancements in technology? No, I think it's completely outperformed anything that we've ever seen. And I think there is a genuine withdrawal feeling that you get from not using cloud or not using chat to be tea, because the feeling of productivity is so addicting. And the fact that you can be more productive with the end of your 24 hours a day and achieve more is like playing a video game. In a way, I feel like it triggers the same dopamine and hormones that you would get doing anything other that's relatively addicting on the internet. At least that's how I personally feel. That's what I've learned about my interactions. It's gotten to the point where this past weekend, I was at a bachelor party in Vegas, we were at the sphere. My mind is racing with ideas. I'm one in a very creative setting, and I'm like, should I need to be clotting right now? I need to be able to bring these ideas. So I literally, I got up in the middle. I went to the side and was like, you can access your CLI from your phone, whatever, there's all these integrations. And I just need to get these ideas down before they run out. A new idea for a concept for an IP tagline, whatever it may be. And it's never been easier to create that way. And that feeling is so addicting. So for me, I feel like that's going to lead to like, down sides, it's definitely a lot of benefits that come with it. I don't know, I mean, you use these tools. How do you feel using them? What is your reaction to that? Yeah, I mean, to me, I would say the only thing I can compare it to is the first time I edited video on Windows Movie Maker where I felt this rush of unparalleled capability. Like I felt like I was capable. You know, at a young age, that's a very powerful feeling to feel. And I think this is that multiplied by a thousand, if not a million, where I can use human language to describe what I want to get done. You know, recently I, this weekend, I was trying to prioritize my week and I had Gemini take over my Google calendar. And I said, here's what I want to do. Here's when it needs to get done by just organize my calendar to give me the tasks, you know, so that I know when to sit down and do these things to get them done when they need to be done. And my calendar was filled out in a way that made sense to me. And I was like, okay, cool, I got my tasks. Great. And that's a very simple version of it. And I think on the most expansive version, like the fact that you can use human language is what's insane to me, that I can just describe something that I want to exist and it can exist. And I think that is, like you just mentioned, like a video game, and I think video games have always informed us about human behavior. Like the addictive nature of a video game actually tells you a lot about who we are and what we want. And even when you look at the rise of the creator economy and the rise of any internet platform, whether it be Twitch or YouTube, they're rooted in the experience of the early adopters were video game users. And the early viewers were people who watched video games. And I think if you keep going down that path and you even explore like what's happening with prediction markets and what's happening with essentially the gamification of everything, everything traces back to the product design of a video game and why we can't stop playing them. So I think what you just said is very accurate. And I think there is no, there is no stopping that feeling. And the output being what used to be reserved for the few, like the artists and the creators, I think being an artist is incredibly aspirational being a creator has become incredibly aspirational. And the fact that using just a description, if you have the taste or the vision that you can become that is fascinating. - I've been creating content for five years. I started a podcast called Mint. I grew my email and invited us on that pot. - Yeah, that was our first interaction. - Yes, you invited us. - We should have come on. - You should have come on. - Yeah. - I grew my email list over 100,000 subscribers. Like I built the human side of content. There was a point where I was publishing like 40 to 60 pieces of content a week. And I know the grit of publishing. I know the grit of showing up every single day. I also have a background in building product and building software, I have a background in big data as well. So I think the way this world converged is unique in a sense to me in my background of being able to create content, build product, understand software and data. And that's a really interesting intersection. But more broadly speaking, reflecting in what you just said, I could be in a place where shit, these AIs are going to replace me in my voice, and they're gonna be able to out execute me or let me just try to use these tools to better my stack and better perform. And I think these tools are just another tool in the shed. Just like we use Riverside to record. We use a sub-stack or VI for our newsletters. These are tools that were built for us to enhance what we do. And these synthetic tools are no different than that. I think there's always going to be a market for human taste and human creation. And I think these tools help expand that frontier. I think also if you look at the traditional arts world, there's so many people who are anti-AI. And even when it comes to the movie industry, they don't even want to use cloud or chat GPT and the script creation because it defeats the whole purpose of creating art. And I think those people, by default, carve out a market for folks to do things in a more human nature because they'll just value the human presence significantly more than those who may be find the advantage of consuming synthetic content. I think there's going to be a market for both. I think I would like also like to argue at some points, Samira, I don't know if you guys are already doing this, you guys will create clones of your voices, you and Colin. And you'll end up creating adlibs of sorts, right? Using your voices because your production team could produce that in a more efficient manner. If you said something or if I said something in inconsistent way that didn't fit the narrative of the story, maybe there's a better way to introduce it. These tools are going to be at our disposal. They're only going to enhance everything that we do. So I don't know. Are you fearful of that? Like you-- No. I am actually not fearful. And I'm also not anti-AI. I am a pragmatist. I look at the world with an immense amount of curiosity and intrigue. And I think the outcome that I've found for myself as a creative person who originally got on the internet because I wanted to be a filmmaker is I have been pushed to think about what I honestly want to say more. Meaning I have been a strategic creator for a long time. I'm running a business. I think what you did, what you made with the Epstein files, as mentioned on the show, that would be the right strategic move at that point. It's a very smart move. It's a very creative application of technology and it's a very strategic thing to do. Now, I don't know if there are two humans who are passionate enough to spend every day of their life making that podcast. I don't know. But what I think we are all getting pushed to be as creators is I just want to watch what you honestly want to do because the strategic thing will already exist. Right? And so I don't know if that makes sense to you of honesty versus strategy. But I can feel it. as a creator when I have something that I honestly want to say versus something that I know will strategically do well. And those two things are getting separated in my mind. And the strategic thing is getting less interesting because it will just be created no matter what. Whether I created or not, it will just get made. And so it pushes me to go like, what, who am I as a creative that is in pot? Like it wouldn't, it could never be an output. And I think that's a really interesting prompt. I think it's going to change how I create. I think it's going to change how a lot of people create. And I hope that it actually, over the next five years, what people see out of both calling it myself is a radically honest version of us. And I think that'll happen across a lot of creators because there's going to be really successful content that emerges that we could never make and we probably don't really want to make. But we would have made before because we're running business. So that's my honest interpretation. And I think it will push people to recognize if they are honest, if they honestly like making stuff or if they just want the outcome of making stuff. Does that make sense? The distinction between like, I enjoy this conversation you and I are having. If we didn't publish this, I'm totally fine. I think this is a fascinating conversation. So this to me is fun. And for all the people who find the video making or the content creation process, fun, I think they will survive. Yeah, I think to your point, before I do anything or I spend my time doing something, I need to be able to hit three C's. I need to feel creative. I need to feel challenged. I need to feel curious. That's how I spend my time. It needs to hit those three things. And the thing that keeps me going with what I'm doing is the sheer curiosity and challenge. And of course, the creativity that comes with producing media with code. Treating media as a product and building media with software. I think is a thing that's like scratching the itch across those three C's for me. And to your point, it's like you enjoy this conversation. I enjoy the art of building product in the form of media. Like that's so fun for me. Yeah, you know? So to see how far I could take that across any vertical is the thing that I wake up very fired up to do. 100%. And even for me, like I'm playing around with Gemini has a coding side of it called Gemini Canvas. And I built this really interesting music software with it, where I could extract stems from different songs. And I grew up loving music. And that was something I was wanted to do. I used to do it in Garage Band. But it was really hard and take me forever. And I haven't done something like that since I was 15 years old. And it was so fun. And it felt like creativity unlocked in this way of like, if you can just think of something, you can make it. Whether it's a video, whether it's whatever it is, if you just think of it, you can make it. And I find myself actually exploring creativity in a non-commercial sense now more than I ever have before. And that is very fun for me. And that is why I approach it all with curiosity. How personalized does the internet get moving forward? Meaning like, when I think into the future, I know right now you are producing an autonomous podcast, or multiple autonomous podcasts that can satiate my desire to be in the know or learn about something on a day-to-day basis that's happening in current events. But with notebook LM, my original use case of it was just putting, because I'm the same as you, I learned better in audio and in podcast forums. So I've done things like I'm interested in like a quarterly S1 document from a company. So Spotify is a great example. I saw their S1 drop. I took the PDF. I put it into notebook LM and listened to it as a podcast. That is a personalized podcast just for me. As we look to the future, do you imagine a world where I'm just prompting my own podcasts? Where I'm like, OK, I have a seven minute drive. So I actually just want to know about what happened in the NFL. I want to know about what's going on with global politics. And I also want to know about what's happening in Hollywood. I got seven minutes just make me a pod that fits that. Yeah, totally. I think that's definitely an outcome. I think we're maybe three to six months away from that actually being feasible within the existing UIs. I think we recently saw Twitter release this audio feature in all their articles. And every LLM has the voice option to read something. But I still think there's an art in being able to prompt engineer properly and to curate entertainment and content for people who otherwise don't have nor care to do that level of diligence and spend the time doing that for what it's worth. I do think the internet is going to somehow consolidate into a chatbot or at least a lot of the tools that we otherwise were used to using that needed interfaces no longer need at the interface. I don't think there's an argument for actually most of the products that we use anymore. Even one of the products that I was building that we built an entire front ended around it. It was for whatever data analytics, not important. But that's just now a single prompt in chat to be tea or cloud. So a lot of these tools that we would otherwise use are no longer relevant. And I think the internet is going to consolidate into a chatbot. And people are going to be building a lot of infrastructure to interact. They're going to be building agent first infrastructures so that the interfaces that do have the attention will be optimized to leverage whatever it is that consumers want to do to interact with other means outside of the chatbot. Because that's really what it is. A chatbot has its general intelligence and people build connectors, kind of legos, into them to enhance their features. And that's only going to get easier with time. And I think there's already an interesting behavior that you are doing. And that I am doing, we're building our own personal software and solving our own personal problems with a few prompts. And that already within itself, like the example that you just gave me, some year that you were doing, that's an app that I pay for 30 bucks a month. Because I'm a musician. I like to separate the tracks and the drums and the vocals and the piano. Guess what I'm going to do after this? I'm going to go try to build that myself. What's the point? And-- but we still all have 24 hours in a day. And we all will spend our time in a particular direction that we think will bring the most value to us. And I still think there's an opportunity to curate the services or the content or the experience for other people. And people will be willing to pay for it, regardless. Well, thanks Adam. I appreciate it. And we'll definitely see you soon here in the studio. Because I might collaborate on something. Let's do it. I'm here for it. Thank you so much. [MUSIC PLAYING] (upbeat music)