AI Content Licensing, Autonomous Robot Secret, Beehiiv's Creator Growth, Sleep Fitness | Aug 19, 2025

0:12Welcome everyone to the information's TITV. My name is Akash Pasricha. It is Tuesday, August 19th, and we have got another tour of the technology sector today, folks. We are talking to Beehive CEO Tyler Dank about what's next for his hit newsletter platform. We're also talking about robotics and a great story that we're publishing today about how much these machines still rely on humans to do their work. We've then got some new funding news about Eight Sleep, and we've got the CEO of that company coming on the show to talk about it. But I want to start with a great story that we published this morning.

0:47Many media companies have jumped at the opportunity to strike deals with AI giants like OpenAI and Perplexity. But now, some media firms are starting to rethink how they want to get paid for their content. I want to bring on Kathy Perloff, who published a story today on an exact topic. And I also want to bring on Paul Bannister, the chief strategy officer at advertising company Raptive, to give some more of his color on the topic. Welcome to you both. It's great to have you here. Kathy, let's start with you. And what did you find? Yeah, so, you know, a lot of the existing deals that publishers have with AI firms are paid on a flat fee, which is something that I learned from talking to folks in the industry.

1:31And publishers are starting to think about a different model for how they might get money from AI firms, especially as like the whole AI boom sort of existentially threatens websites. You know, if no one goes to websites because they're in chatbots, no one's going to see websites ads, how are they going to make money? So this idea comes of usage. So every time a AI chatbot might crawl a website or use any of its content, a publisher could get paid for how much of the content the AI firm uses. And that's sort of a more recurring revenue model than just a flat fee model. And, you know, I think part of what has inspired publishers to start thinking about this is first, the crawling of their websites has really increased in the past year from AI firms.

2:20It shows that AI firms still need publishers' content, even if their chatbots are up and running. These firms are making a ton of money charging developers on a usage basis. So that means that anybody could pay publishers on a usage basis. And then finally, the IB Tech Lab, which is a kind of advertising trade organization, has put out some proposals for how this might happen. So has Cloudflare, which is like a web infrastructure company. Those proposals have come out both this summer. And then we have learned that Perplexity is launching a new model for out-of-pay publishers. They already are paying publishers based on some ad revenue that they generate.

2:59But some of the new model would take into account some other factors that could be usage-related. And that is supposed to come out in a couple weeks or so. Got it. So there's a lot there. So the way I understand it is, again, publishers have striked these deals with these AI companies. They were a flat-fee model traditionally. The publishers now want to switch that to a usage-based model, presumably because of the fact that they are finding that their stuff is getting linked to more and more often. The center character of this whole story is the Interactive Advertising Bureau. Paul, can you give us a little bit of context on the IAB?

3:35What role have they traditionally played in advertising and why are they important here? Yeah, yeah. Full disclosure, I'm on the board of the IAB Tech Lab. Okay. But I think the IAB and the Tech Lab part of it play two roles. One is a traditional trade group. How do we grow the market? How do we bring more money in? And obviously, we've been quite successful about it in general over the last couple of decades in terms of really growing the amount of advertising money that is spent on the internet, which is great. The tech lab specifically sets the technical standards for how advertising works in terms of programmatic advertising, in terms of the different ad formats and things like that.

4:09And so they've been very involved in, you know, kind of that part of the market. And I think with this kind of new, these new steps around how LLM and generative AI monetization may work, the tech lab is kind of moving a little bit even beyond advertising into like, what are new monetization models for publishers in general, and what does it all look like in the future? And so usage-based pricing, I guess, is one model there. Do we have any data as to how much these publishers have seen their traffic decline because of these chatbots? Well, from our side, what we've seen is very vertical in terms of how things are going.

4:43Certain verticals have been hit very hard. Other verticals have been doing totally fine. And also very - Like which ones are hurting more than others? A vertical that has been down a lot is like personal finance. So that kind of content has seen a pretty dramatic nosedive, whereas like food and cooking content has been doing pretty fine. And I think part of it is what is the type of information people are looking for in those areas? And where is AI a better replacement for that versus AI being kind of more an augmentation to that? Kathy, how big were these deals in the first place, the flat fee models?

5:16Yeah, you know, they're confidential, but we have some kind of insight into them, which is, you know, at the larger end, they're like maybe tens of millions of dollars a year. Like the Wall Street Journals reported that their parent company, News Corp's deal with OpenAI was like$250 million over five years. Similarly, it was reported that Axel Springer's deal with OpenAI, tens of millions of euros a year. But, you know, also a lot of them, and I've heard this from, you know, people in the publishing industry too, only single-digit millions. And, you know, the information actually reported last year that some of the deals OpenAI was offering publishers were like one to five million.

5:55So, you know, I think they're... From a company that really, I mean, OpenAI is making billions of dollars in revenue right now. They're making a billion a month. So, like, you know, maybe they have more to give. That's what publishers would say. I don't know. They have a lot of other costs, like infrastructure and et cetera. So, Paul, you know, we heard some of the numbers here. Single-digit millions, maybe 25 million even in some cases. How much upside do you think there is if the publishers would switch to, you know, if the AI companies would agree to these usage-based models, could these deals get in the hundreds of millions?

6:29How big could they get? Yeah, I mean, I think, I mean, how big can they get exactly is a hard question. But I do think that the model of paying on usage is the right way to think about things. Let's all work together to create something that is used and great in the future. and if our interests are all aligned and we're all kind of working towards the same goal, that will make things better. I think as publishers think about, should I implement MCP to make it easier for AI companies to scrape my content and give access to that in more structured and useful ways, there's an investment in that. And if publishers can say, hey, I know I'm going to be compensated in the future for this in a way that really aligns my interests, I'm more likely to do those sort of things because they're going to help the AI companies out.

7:08So I think it's like alignment of interest is the most critical thing here. And other than, And you talked about the different monetization models. Usage-based pricing is one way to go. What are the other sort of ways you could see perhaps publishers getting creative insofar as how they're compensated by these AI companies? In my opinion, it's more about within usage-based pricing, what's the right unit? There are paper crawl models. There are paper query models. There are other things being developed right now. What is the right unit that the AI companies are paying for? And what's the way that aligns interest in the right way?

7:48For me, I think we're biased towards the paper query. I think that is the most aligning. But there's a lot of testing going on right now around paper crawl just to see, is that a good model of ChemNetwork? And is there any data out there right now around how many crawls there are per day or per customer or anything like that? Cloudflare has a lot of data around this that's pretty public. Like I think Kathy has shared some of this in her article that was really good. I think the biggest thing is almost like the absolute numbers are quite large, but there's also this point of the ratio of the number of crawls an AI company does to your site and the amount of traffic it sends you.

8:22Google is the example and paraphrasing the numbers as I remember them. But a couple of years ago, for every one time Google crawled a site, it would send two visitors. Now that number has inverted into something like one to 10. So for every 10 crawls, you only get one visitor. And the other AI companies are even more extreme than that in terms of the ratio perspective. And I think that's what makes it so clear to publishers that AI companies need this content. This is critical to be able to feed their models and to be able to do the grounding in real time to really answer the queries in the right way.

8:50And so, okay, that crawl number has gone up a lot. Is that the right way to price things? Remains to be seen, but that crawl ratio, I think is the most important metric. Kathy, we got to go, but say more about that data. You seem to have known that data pretty intimately. uh no actually um no i okay i thought paul suggested maybe but i have heard it paul knows it better but i think it's a really interesting point okay well i thought maybe you guys had talked about this on your phone call last night or something like that um okay last question for you kathy before we let you go why wasn't the the pricing model a usage-based model in the first play.

9:29They designed these deals. They went for the flat fee. Why didn't they see this coming? You know, it's a great question. And I think the issue is that publishers didn't want to get, like, scammed. And I think there was an idea that there wasn't enough data out there to really come up with the right criteria to figure out, like, what a usage-based model would look like or what What criteria at all should be publishers should bring to the table when negotiating with AI companies like OpenAI? A lot of publishers I talked to just, like, didn't know what would be relevant because traditionally the way publishers have been paid is via advertising.

10:06That doesn't really exist in the chatbot world. So you can't say, oh, well, you've brought me this much ad revenue, this many visitors are connected to this much ad revenue. So it's a little bit the Wild West. And I think that, you know, usage is just like, we're just starting to be mature enough here where maybe we can talk about some actual math. Right. You know, one publisher I spoke to actually said, we tried to think about like with ad revenue or how much content we had and come up with a number to give OpenAI. Right. Like, we realized it would be more than they would pay. So we just were like, let's get the highest number we can get.

10:41So I think that like they're right. There just wasn't enough data and maybe we're starting to get there, but we'll see. Got it. Well, look, Kathy and Paul, thank you so much for coming on the show to talk about it. Again, Kathy, it was a great story. We will link it in the show notes. That is Kathy from The Information and Paul from Raptive. Okay, for our next segment, robotics have been front and center this year with many stunning advances in what these machines can do. But apparently many of these robots aren't as good and as autonomous as we might think. I want to bring on the information's Rocket Drew to talk about a story he wrote today on that topic.

11:17And I also want to bring on John Miller-Schwartz, the CEO of Ultra Robotics, to share his perspective of his own company. Both of you, welcome to the show. It's great to have you here. Thanks so much for having us, Akash. Likewise, thank you. John, now you said you were going to bring a robot to the stream. If we could get a full stream, is that a robot behind you? Do I see it? Exactly. This is one of our robots. This is in our lab. We use this robot for development and testing of our AI policies. Okay. All right. Well, I'll tell you what, we'll talk about your robot in a second. Rocket, you wrote this great story today about the humans behind robots.

11:52Tell us about what you found. Yeah, I'd love to. So the idea of robots and autonomous vehicles is, of course, that they should operate autonomously. That's kind of the goal. It's right there in the name. But if you've been seeing some of the flashier demos, videos of robots working over the past year or so, and you've been listening to some statements, public statements from the more booster type CEOs, you might have the impression that robots can basically do anything autonomously already. and they're about to march into your home and do all your chores for you tomorrow. So you might be surprised to learn about the large role that humans have behind the scenes in supporting robots of many different stripes, robots on wheels, robots on legs, robots in industrial settings, robots in homes.

12:39So our article is a look at some of the ways that humans are supporting these robots behind the scenes, from providing guidance and judgment about decisions, all the way down to remotely operating, sort of puppeteering the robots, what roboticists might call teleoperation. So that's the premise of the story. I think it's not some big scandal, but it might come as a surprise if you've been following robotics from a distance. I think people in the industry would tell you this is just an important step. It's a key ingredient to bringing robots out of development into the labs and into the real world.

13:12And what are these humans doing in the background for these robots? It depends on the case. In the case of some autonomous vehicles, maybe your car gets confused. It encounters a situation it hasn't seen before. Maybe a strange looking construction site. Maybe a swarm of bees fills the road and it just doesn't know what to do. It could sort of call for help, figuratively speaking. And a human behind the scenes could say, yep, that's a construction site. You've seen these before. You know how to proceed. Other times it involves a person wearing a specialized glove or wearing a VR headset and controlling the robot, actually using your remote control to teach the robot how to manipulate certain objects in the world.

13:56So, John, are your robots, are they semi-autonomous? Do they have humans controlling them in the background? So our robots today are semi-autonomous. And what that means is the operation is a combination of both an AI policy, which is literally controlling the movement of the robot. But we also have a human in the loop, we like to say, that is watching the live stream from the robot. And in certain cases can intervene if the robot is off course, not doing the right thing, hits some issue and gets stuck. And that's where there's a lot of value, basically, in rolling out these AI policies, is it allows us to actually get them into the real world, have the robot create real value for our customers, but then also have people there to make sure that the robots can operate reliably over time.

14:39And the goal, eventually, is to make it fully autonomous without any humans in the loop. Is that the idea? Yeah, eventually we'll get there. You know, it may be a number of years away. Ultimately, the goal is just how do we make it more autonomous over time? There are really two benefits to that. One is from a margins perspective, we would obviously love our robots to be more autonomous, which makes them lower cost to operate. But it also, in our view, improves the product experience for our customers. Part of the dream of having a robot is having basically always available productivity and labor.

15:11And so the more autonomous it is, the more easy it is for our customers to just decide to turn it on at any given point in time and have it operate and do the job. Rocket, it feels like this is actually more common than we might think. I mean, I remember hearing about the Alexa voice devices, for example, you know, these articles about people are actually listening to anonymized voice cues in the background or stuff. I mean, this is pretty common, right? Yeah, I think this is great to clarify. I mean, there are a lot of robots in the world and robotics isn't new. So robots have been performing repetitive tasks in industrial settings in factories for decades, you know, say, welding parts of cars together to assemble them.

15:50So there are plenty of robots that do work autonomously, but they tend to be working in more limited, more narrow ways, say, bolting the same part together every minute after minute, day after day, without a lot of variability. The kind of robots that we're really zooming into in this story is the frontier of robotics, the place where AI is sort of pushing the capabilities of robots, allowing robots to perform tasks they haven't been able to do in the past in environments where they haven't been able to work in the past. So I think Ultra is a great example of this, actually, right on the frontier of what AI is making possible in robotics right now in warehouse settings, in e-commerce logistics, for example.

16:29So, John, I want to ask you, there's been a fear, I think, that has come with robotics that, oh, robots are going to replace humans, humans are going to run out of jobs, this is going to be a big problem. And here we have a story about the reality that there are still humans behind the robots. And so do you think that fear is overblown? I mean, it seems like we're still maybe decades away from a robot replacing a human entirely because there's still a human in the loop. What's your thought here? Exactly, Akash. I think that fear is overblown to a degree. It's the scary sci-fi future that we all want to avoid.

17:05The truth is, when we walk into our customers' warehouses in the United States, there is a real labor shortage. And they have a very challenging time finding the people to fill the roles that they need to operate every single day. You've read a lot about this in the newspapers, but this is the experience that they live continuously. So I think that there's a much greater upside in the near term for actually helping to fill that labor shortage and provide them with affordable, scalable productivity in the form of autonomous robots. And that's what we're trying to do at Ultra. Robert, what do you think of that?

17:38Yeah, I think these kinds of jobs offer an interesting counterpoint to narratives about AI-enabled job displacement. The role of being a teleoperator for robots behind the scene is an interesting example of an AI job, a job that you wouldn't have seen, say, five, 10 years ago. It's like the new prompt engineer for robotics. Exactly, prompt engineering, data labeling is maybe another example of that. Now, as companies, you know, their goal is to be more autonomous over time. We'll have to see what happens to these, this class of jobs. But for the time being, it seems like there's something that humans have to offer still, according to their sort of comparative advantage, right?

18:16The humans are offering really, you know, dexterous movements that the robots haven't quite mastered yet, or judgment about how to understand different scenes and what actions you're supposed to take in those scenes. So humans still have something to offer robots, at least for now. Also, just to add, Rocket, that on the other side, we've seen that within our customers' warehouses, there's an exciting new opportunity for people on the warehouse floor to be working alongside with these robots and helping to maintain them, make sure that they're up and running. And it's really a great opportunity to upskill yourself and get to work with new cutting-edge technology.

18:48So there are lots of different names for that role, but it's an exciting kind of direction that we're seeing things head in. That's a great point also, because in the labor question, there's always this sort of secondary equity question. It's about, well, maybe jobs are created, but is it the same people whose jobs were taken that now have a new job, right? If you have an autonomous vehicle riding through the streets, but it's being puppeted by someone who works in Mexico, maybe your taxi driver in New York City is still out of a job. But there are a few instances where people who, you know, they went from operating some piece of equipment to now teleoperating a robot to do the same sort of task.

19:25So there's complexity there. John, last question for you. How big are these teams of humans behind these robots right now at some of these companies? Yeah, I mean, I can't speak for other companies, but here at Ultra, we have on the order of, I'd say, a dozen people who we've trained up on how to teleoperate our robots. And these people are core to our day-to-day operations. They're the ones who are controlling the robots in real time when the AI policy isn't. And then they're also helping us just with the ops because they're actually doing the job. And so there needs to be a lot of tight coordination between us and them to go and provide value for our customers.

20:02Great. Well, look, John, next time you come on the show, what we're going to have to do is we're going to get that robot behind you to give us a bit of a demo, and we're going to see how the whole thing works. And maybe you can introduce us to one of those people in the background that are helping to operate it. Thank you so much for coming on the show, John. And thank you, Rocket, for that great story. That is Rocket from The Information and John from Ultra. Well, newsletters have become the new highway for publishing, and Substack has long been one of the only dominant players. but over the past few years, Beehive has quickly gained traction and is now duking it out with Substack as Beehive becomes increasingly popular.

20:39The information published a great piece in our Weekend Magazine a few weeks ago about Beehive. I will link it in the show notes, but I want to bring on the CEO of Beehive, Tyler Denk, for an inside look as to how he is building the company. Tyler, welcome to TITV. It's great to have you. Good morning. I wish I had a background like John, but it's just my bedroom. Well, I was going to say, I thought you were going to, you told me you were going for a swim this morning. I thought maybe you were going to zoom in from the swimming pool. Yeah, I did get a mile swim in this morning. So I woke up early for you this morning.

21:08Okay, great. Well, the key to a great newsletter is getting a swim in the morning. That's from Tyler Degg, folks. Look, we wrote the story. Some of the financial figures interested me. The story said that you are doing 20, well, you've done$20 million in annual revenue, and you're at a$30 million run rate. Help me square those two numbers. Yeah, exactly. So our core business is the SaaS business and we're generating 20 million ARR right now. We also have an ad network built on top of the SaaS business. So we have advertisers like Nike, AG1, Netflix, Roku that are advertising within the newsletters in our ecosystem.

21:45And that's a$10 million revenue run rate business right now. So that's how you get the 20 million plus the 10 million gets you to 30. The ad network is growing extremely quickly. And so, yeah, on pace for about 100 % year-over-year growth in total this year. So you're planning to double your total revenue this year in 2025? Correct. Got it. And for the ad network, do you find the advertisements for the newsletter authors, or are they bringing their own sponsors? So we are unopinionated. We are pro-creator monetization, right? So if you want to monetize your newsletter and you already have a direct relationship with advertisers, you can do that on the platform.

22:23we don't take a cut of revenue that is yours to entirely earn. What we do is we are sending two and a half billion emails per month. We have over 30 ,000 active publishers on the platform. And so that attracts advertisers like the Nikes and Netflix of the world who want to get in front of these niche audiences via email. And so what we do is we have a ML team building models based on all of the historical data that can identify for this specific advertiser in this campaign, what are the best few thousand newsletters who perform the best for the advertiser, also who is in the best interest for the publisher to promote to their audience.

22:59And so we have all these models running that we've been training for several years now. And the end result is publishers wake up every Sunday. They receive an email with five to 10 different ad opportunities and it's totally opt-in. They can decide to run ads for$3 CPC,$10 CPM, whatever that is, or they could choose to run their own ads or not monetize at all. we are just pro creator monetization and however we can help our users grow and monetize we're in favor of building tools to do that i want to talk about the models you're using in just a second but i just want to go back to that growth figure you're planning double revenue you said this year are you taking writers from sub stack uh i'd like to think we're taking writers from everyone in the ecosystem i think we're because of when we launched and the ecosystem that sub stack had built a few years prior, we're always very closely coupled and compared to Substack.

23:50I think Substack is actually closer to building a social network competing with like an X and a thread more so than an open newsletter platform and website builder that we are building. They're clearly going for, you know, network effects and social. And you can see that in the app and the way that they have this closed ecosystem. We are open. We have APIs. We have We play nicely with other third party platforms. You could be a large publisher or creator who has a website on WordPress and a massive YouTube channel. And we play nicely with both of those platforms. So I understand the comparison to Substack, but we're getting users both from starting newsletters from scratch, from being large content creators on YouTube and other social platforms.

24:33And then also all of your traditional email platforms like MailChimp, ActiveCampaign, CampaignMonitor, etc. cetera. There's a pretty broad ecosystem with email tools. So basically, when people ask you, how do you compete with Substack, your answer is we're not building a social network. Yeah. And we've been seeing that more and more. I think Substack actually launched something yesterday with like Apple subscriptions or whatever, where if you read the comments, it wasn't a very well-received launch. And so a lot of the feedback to us and our users is... What were they saying? It's just like the take rate in the app ecosystem, right?