← The Artificial Intelligence Show
The Artificial Intelligence Show
#191: Ads in ChatGPT, Claude Cowork, Elon Musk vs. OpenAI, Personal Intelligence from Gemini & Real-World AI Use Cases
01:24:32
Loading summary
Transcript73 lines
- [00:00]
A
But if the AI can do that now, all you really need is a director level person with experience and instinct and intuition who can just like oversee and review all this work and maybe go in and do some edits and things like that and kind of get it to the finish line. But it starts to change the way work is done. Welcome to the Artificial Intelligence show, the podcast that helps your business grow smarter by making AI approachable and actionable. My name is Paul Raitzer. I'm the founder and CEO of SmartRx and marketing AI institute and I'm your host. Each week I'm joined by my co host and Smarter X Chief Content Officer Mike Kaput as we break down all the AI news that matters and give you insights and perspectives that you can use to advance your company and your career. Join us as we accelerate AI literacy for all. Welcome to episode 191 of the Artificial Intelligence Show. I'm your host Paul Raitzer along with my co host Mike Put. We we are recording on Monday, January 19, 8:40am A little earlier than normal this week. Got a birthday party I got to go to this morning. This episode is brought to us by AI for Agency Summit. We've been talking a little bit about this. This is coming up on February 12th from noon to 5:00pm Eastern Time. This is our third annual AI for Agency Summit. This one is presented by Screen Dragon. It is a virtual event. So again this is Thursday, February 12th. It's designed for marketing agency practitioners and leaders ready to reinvent what's possible in their businesses and embrace smarter technologies to accelerate transformation and value creation for themselves and their clients. During the event you'll join other forward thinking agency professionals to see how agency leaders are using AI to drive innovation and efficiency. Learn AI workflows you can launch immediately, including research, onboarding, automation. What's your talk Mike? You're doing?
- [01:52]
B
Yeah, I'm doing one in NotebookLM specifically for agencies. Yeah.
- [01:55]
A
Cool. And my my opening KeyNote is about AI and innovation within agencies. Understand what AI agents actually look like inside agencies today. Assess legal, IP and risk issues agencies need to address and evaluate how AI is reshaping brand agency relationships and what clients expect next. Is all presented by a group of world class agency leaders and industry experts. You can go to aiforagencies.com that is F O R A I4 agencies.com There are free pass options available so you can go and check that out and see the full agenda. So that is live. Again that is February 12th virtual event, half day. There's I think six sessions total ending with Ask me Anything session that has you me. I think Will Reynolds is in that with us. So yeah, yeah, definitely check that out. And then I think I mentioned this last week, but we have another free event coming up. This is the 2026 Marketing Talent AI Impact Report webinar, so we'll be launching that report that day. This is part of our Industry Council, so AI is reshaping marketing roles, teams and the future of work much faster than most organizations are ready for. Join us January 27th at noon Eastern time for a live free webinar to break down the most important findings from our inaugural report. We'll cover how AI is changing hiring, training, workflows, governance, and what skills will matter most in the next one to two years. If you're a marketing leader, people manager or practitioner looking to future proof your career or team, we would love to see you there. And if you're not in those roles, we ask you to pass it along to team members who are. So you can go to SmartRx AI forward slash webinars to register for the January 27th event. It will be recorded and available on demand as well. So if you get registered, you will get a notification when that is ready to go. Plus, all registrants will receive a copy of the 2026 Marketing AI Talent Report. It's a really good report. We've been working on this one for months with research. We have about 30 or so AI council members and we actually interviewed them on these key topics. It's really good stuff, really cool insights. So Mike and the team did a great job putting that together. And then one quick reminder, Macon 2026, which is coming up October 13th to the 15th. We have an open call for speakers right now. Can go to Macon AI that is M A I C O N AI again, that is happening in Cleveland October 13th to the 15th. It's our seventh annual conference at the Huntington Convention center of Cleveland. If you are doing incredible things yourself, if your team is doing cool things, if your organization is going through AI transformation, you've got a story to share. We would love to have you submit to speak at Macon 2026. So again, go to Macon AI. Okay, AI Pulse. So if you're new to the show, we start each episode off the weekly episodes with an informal poll that we do with our listeners so you can take part in this poll. I'll give you that address in a second, but last week we asked two questions. How likely are you to connect your medical records to OpenAI's new ChatGPT health interface. We had 85 responses to this question. So again, informal poll, just sentiment from our audience. Mike. We got 48% said, somewhat likely, but I have privacy concerns. So that was the number one answer. Number two, not likely. I prefer to keep my medical Data separate from AI. That's good. About 19% said, very likely. I want personalized health insights so that those are kind of the big ones. And then the other one was I already use other tools to track my health. I think I mentioned this last week, but like my personal thing is if it's in Apple, you can have it. Like I already, everything I already have is in Apple. I don't know if I'm connecting stuff to ChatGPT. Yeah, would you do it? Like, are you. Where are you?
- [05:48]
B
I'm weird. I'm probably a little more on the like, loose side.
- [05:52]
A
Just have it.
- [05:53]
B
For me personally, obviously I totally respect and understand the nuances and the trepidation. I would probably do it. I mean I already have done it informally by dropping stuff in there.
- [06:04]
A
It's true. I tell the story. Last week I put all my heart data in there, like, right. So I guess I've already given it to it. Okay, then the second question last week was, have you personally used Claude code yet? And this became much more relevant. We'll talk more about cloud code today. So 48% said, no, I haven't used it. Which isn't surprising because it's not easy to use. It's not really user friendly yet. So I would think the people who are probably more advanced. 34% said, yes, I'm using it for coding or technical work. That's, that's interesting. And then about 12% said yes, I'm using it for non technical workflows like research or file management. So we've got some cloud code users.
- [06:41]
B
Yep.
- [06:42]
A
Okay. So this week again you can go to SmartRx AI forward slash pulse to participate in the survey. This week we have two questions. Again, these will make more sense after we get through today's topics. The first is, how does the news of ads coming to ChatGPT affect your view of the platform? We'll talk about this as the first main topic. ChatGPT is going to start testing ads this week. And then the second question, are you comfortable allowing an AI agent like Claude Cowork to. To read and edit files directly on your computer? Hold off answering that question until we get through the Claude cowork topic because Claude Anthropic themselves provide some pretty stark warnings about using Claude Cowork So you definitely want to listen in before you answer that question. So again, go to SmartRx AI forward slash pulse and you can check out past surveys and participate in this week's poll. Okay, we, we have some drama in the AI world. I I we're going to start off with talking about ads and co work and stuff, but we are gonna move into Elon Musk's battle with OpenAI thinking machines Exodus of their found it's, it's wild. Like I every week that goes by just someone has to be writing this as a show like the next version of Silicon Valley, like the next seasons of it or something like it's truly starting to just get beyond reality. You would think this is the kind of stuff they would write in Hollywood, but okay, let's, let's kick off though with OpenAI and their ambitions in the ad space. Mike all right, Paul.
- [08:19]
B
Yeah, the honeymoon period's over. We're back to like all full bore AI news and drama all the time. So first up, OpenAI has formally announced its plans to begin testing ads within ChatGPT. And you know, ChatGPT has historically relied on paid subscriptions and investor funding to generate revenue, but now they are moving towards a diverse revenue model to support the free and low cost tiers of the tool. So these initial tests will roll out in the US in the coming weeks. To start, ads will be displayed at the bottom of the interface. They'll be clearly labeled and separated from the chatbot's organic answers. However, in the announcement Post about Ads, OpenAI CEO of Applications Fiji Simu Simo has hinted that in the future ads could go beyond static placements and links. So these sponsored results these ads will appear only right now for users on the free version of ChatGPT as well as on the newly launched eight dollar per month what they call their go tier. Now according to OpenAI, ads will be matched to the topic of the current conversation, but they will not influence the substance of the Chatbots responses. The company stated it will not sell your data to advertisers or show ads related to sensitive topics such as politics or healthcare. Additionally, as of right now, users on Plus Pro and Enterprise plans will continue to have an ad free experience. So Paul, we knew this was coming but you know you had highlighted in our research it doesn't seem like this was always the plan. Sam Altman has previously expressed distaste for ads. He's even called them a quote last resort. He has expressed worry they would reduce user trust in ChatGPT. So I'm curious as to what changed here. That ads are now the play.
- [10:11]
A
Yeah, let's start off with the why ads question. So they're, they're burning cash. I mean, if you a regular listener to the podcast we talk about all the time, obviously the cash burn for OpenAI and their need to sort of insatiable need to keep raising funding not only to sustain the current business model, but to build out the infrastructure for what they see as an intelligence explosion in the, in the years ahead. So that, that is an issue. They have to show revenue model. So in December, we had the Wall Street Journal report that I'm pretty sure we would have talked about at the time, but it said OpenAI is aiming to raise as much as 100 billion as it seeks to pay for ambitious growth plans in a market that has cooled recently. On the AI boom, they said the fundraising round, which is in the early stages, could value the company at as much as $830 billion. So that was again December 18th. Just a. We know they're pursuing an IPO. It might not be in 2026, but we know they're setting the stage for that. And an ad revenue model obviously would go well within the investor world. The interesting thing, like I found myself asking, okay, so they're only going to do this for free? And the $8 a month, like, well, how much of the user base is that? So I went back and found the information article from November 2025, and they had some leaked internal data that sort of put, I believe this is the best snapshot we have to date of that distribution of who pays, who doesn't. So I'll just read a few excerpts from the Information article and we'll put this link in the show notes. So again, this is from November of 2025. It said as of July, about 35 million people paid for plus or Pro subscriptions, which typically cost $20 and $200 a month. According to a person with knowledge of the figures at the time, that amounted to 5% of its weekly active users. In five years time, OpenAI has projected that about 8.5% of 2.6 billion weekly active users, or 220 million people will pay. So again, Snapshot, in July 2025, about 5% of their user base was paying for the subscriptions. Everybody else was using the free version. And they look to the future and say, by what is this? By 2030, basically, they think they'll go from 800 million active users to about 2.6 billion weekly active users, and they think about 8.5% of those people would be paying for a plus plan. So then it said if ChatGPT can attract that many subscribers, that would put it among the biggest subscription businesses in the world. Netflix and Spotify each have about 300 million paying subscribers. To put it in context, and then they also gave, maybe a better analogy would be Microsoft Office 365 productivity app suite has about 4,450 million paying subscribers. So they said planning to capitalize on a three year old chatbot's early lead to drive nearly 200 billion in annual revenue by the end of the decade. From around their number was 13 billion. We know this to be actually closer to 20 billion now based on article we'll talk about in a minute that their CFO published. So OpenAI increased revenue expectations for ChatGPT subscriptions will generate 270 billion in revenue through 2030. That includes 87 billion in 2030 alone, up from 10 billion. And then they talked about business subscribers to break that out even further, Business subscribers make up a small portion of Chat GPT subscriber base. Today. Earlier in the month, which would have been November, OpenAI said 7 million users at companies access ChatGPT subscriptions through its business plan. So 7 million out of 800 million is less than 1%, right? Yeah, I'm doing really quick math in my head. These include Chat GPT business for smaller companies. Okay, so that gives you the context of like what we're talking about here. So if we have, if we assume they have about 830 billion users or whatever 800 billion users will say, Is that right? Am I saying 800 million?
- [14:06]
B
No.
- [14:07]
A
Yeah, 800 million users right now. So the vast majority of those 800 million are not paying for the subscription.
- [14:14]
B
So it's like the world's biggest possible. Yes.
- [14:17]
A
Yeah. So what, what they're saying is, and so leading into this conversation about why ads and stuff, the vast majority of people using Chat GPT don't pay them for it. And so it's very expensive to serve this up and they, they imagine it becoming more expensive. Okay, so now Mike, to go back to your point about Altman not liking ads. So I went and pulled the transcript to get context around what he said. And so this is actually from a fireside chat at Harvard University in May of 2024. So not, not that long ago, we're talking about a year and a half ago. So the quote that kind of got shared a lot in the last few days on Twitter, it was like the video was just being clipped and shown everywhere was Ads plus AI is sort of uniquely unsettling to me. I kind of think it adds as a last resort. So that was the quotable thing like, okay, so let's go back and get the context. So here's where that quote comes from. During this fireside chat, an audience member asked this question. He said, in the last two decades or so we observed how the Internet landscape has evolved and how businesses like Google revolutionized search with ad based monetization, click through rates, et cetera, which spurred a wave of innovation in businesses and provided numerous free services. I personally am very bullish on GPT technology and I think it will provide a platform to start many new businesses that we cannot think of now. However, the subscription model that GPT is based right now on, personally I think although fair, it could be a barrier for early stage entrepreneurs or startups or even small businesses. Given this context, do you envision OpenAI exploring alternative monetization strategy that could include free API access which they don't currently offer? I believe perhaps supported by advertising or other methods to foster innovation in the future? So then, this is Sam's verbatim reply. I will disclose, just as a personal bias, that I hate ads. I think ads were important to give the early Internet a business model, but I think they do sort of somewhat fundamentally misalign a user's incentives with the company providing the service. I am not totally against them, I'm not saying OpenAI would never consider ads, but I don't think them in. I don't like them in general. And I think that ads +AI is sort of uniquely unsettling to me. When I think of ChatGPT writing me a response if I had to go figure out exactly how much was who paying here to influence what I'm being shown. I don't think I'd like that. And as things go on I think I like that even less. So there's something I really like about the simplicity of our model, which is we make great AI and you pay us for it and we're trying to do the best we can for you. And then given that that has some inherent lack of access and inequality, meaning other people can't afford it. We commit as a company to use a lot of what basically the rich people pay to give free access to the poor people that was not greatly worded or the poor people, it got worse. You see us do that today with the ChatGPT free tier, which we just discussed is the vast majority of their users. You'll see us do a lot more to make the free tier much better over time. And I'm interested in figuring out how we bring the equivalent concept to the API. Again, the API is other people using accessing ChatGPT models to build other technology on top of it. But I kind of think of ads as a last resort for us for a business model. I would do it if it meant that was the only way to get everybody in the world access to great services. But if we can find something that doesn't do that, I'd prefer. Okay, so that then sets the stage for, as they started ex, I guess, admitting publicly, yes, we are going to do ads. Like they, they were kind of like hedging around this for a while. They published a post on our approach to advertising expanding access to ChatGPT. So they're sort of now playing the PR game here of, okay, we got to get out ahead of this. We haven't been, you know, very positive about the idea of ads before. Like we have to address how we're going to do this in an elegant way. So they lead this post, which we'll put the link to in the show notes, with the story about access. So it comes back to this idea that this is all about enabling a quality of use of intelligence, that everybody gets access to intelligence. So they said AI is reaching a point where everyone can have a personal super assistant that helps them learn and do almost anything. Who gets access to that level of intelligence will shape whether AI expands opportunity or reinforces the same divides. They then talk about this go. Which I don't even remember talking about on the podcast, Mike. So this.
- [18:48]
B
Yeah, I don't know if it was like we, it got buried in some other.
- [18:50]
A
Yeah, they just like slid it out there or whatever this $8 a month. But apparently it's available in 170 countries and they're bringing it to more countries. So then right after they talked about the access part, they get into the trust thing. So then there's a paragraph again. This is right on OpenAI's own blog. It says, People trust ChatGPT for many important and personal tasks. So as we introduce ads, it's crucial we preserve what makes ChatGPT valuable in the first place. That means you need to trust that ChatGPT's responses are driven by what's objectively useful, never by advertising. You need to know that your data and conversations are protected and never sold to advertisers. And we need to keep a high bar and give you control over your experiences so you truly, you see truly relevant, high quality ads. And can turn off personalization if you want. And then they highlighted these principles with Fiji had a separate personal blog post I think about these principles. So this is basically their own internal justification for why an ads model can work despite the fact that Sam personally hates ads. So it calls out a few elements. Mission alignment Our mission is to ensure AGI benefits all humanity. Our pursuit of advertising is always in support of that mission and making AI more accessible so we have to make money so we can make this available for the poor people to use their own words. Answer Independence Ads do not influence the answers ChatGPT gives you. Answers are optimized based on what's most helpful to you. Ads are always separate and clearly labeled Conversation Privacy we keep Your conversations with ChatGPT private from advertisers and we will never sell your data. Choice and control. You control how your data is used. You can turn off personalization. That's the same way Instagram works and you know, Facebook does theirs. And you can clear the data used for ads at any time. We'll always offer a way to not see ads in ChatGPT including a paid tier that's ad free for now. Long term value we do not optimize for time spent in chat gbt. We prioritize user trust and user experience or revenue. That's actually a really important if they hold to it because that is not what Facebook does right. So they're, they're taking a different approach supposedly how it'll work. You touched on a little bit of this mike. They're not launching ads yet. They plan to start testing in the coming weeks for logged in adults in the US on free and go tiers. You'll be able to learn about why you're seeing that ad or dismiss any ad and tell us why during the test. They will not show ads and accounts where the user tells us or we predict they are under the age of 18 and they ads are not eligible to appear near sensitive or regulated topics like health, mental health or politics. And then they said the best ads are useful, entertaining and help people discover new products and services. Given what OpenAI can or what AI can do. We're excited to develop new experiences over time that people find more helpful and relevant than any other ads. Conversational interfaces create possibilities for people to go beyond static messages and links. For example, soon you might see an ad and be able to directly ask questions you need to make a purchasing decision. So obviously this stuff is already in internal testing. They have other visions for what's going to happen. Sam Altman Then tweeted, we are starting to test ads and chatgpt free and go the $8 a month option tiers. Here are our principles. Most importantly, we will not accept money to influence the answers ChatGPT gives you. And we keep your conversations private from advertisers. It is clear to us that a lot of people want to use a lot of AI and don't want to pay. So we are hopeful a business model like this can work. And then he put in parentheses, an example of ads I like are on Instagram where I found stuff I like that I otherwise, otherwise never would have. We will try to make ads ever more useful to users. And then the final note, Mike, here is just like the other side of this is like, hey, like if you handle media buying or if you're a small business and you want access to a whole new market, yeah, here's 800 million users and take out the 95%. You know, basically 95% of them aren't paying. So is a massive audience, like, like it's called 700 million for, you know, just to round up. So ads can be like transformative. Like, this is like a whole new distribution channel. And so as a brand, you start like, well, should we get in there? Should we start running some ads and testing some ads where you can get low cost per thousands? And that's something to start debating is like, you know, should you be gearing up as a marketer, an advertiser, be thinking about like, let's get in there and let's start testing some stuff. Whether it be backlash to your brand for being in there, people are going to get annoyed. I don't know. But like, you definitely got to be thinking about that if you're a brand. And then the other thing, Mike, I found myself wondering is like, okay, what's Google's play here? Right? They're, you know, they're not going to be far behind. I'm sure they're already testing stuff as well. So, yeah, just a really fascinating thing. See how this all plays out, given the size of their market and their ambitions and could be a whole new advertising channel. And then I guess the trickle down would be like, what impact does that have on Google and Meta where like now you have more competition, a massive new competitor for fraud dollars, which are limited. Like, there's a finite pie of ad dollars and like, how are you going to shift those around?
- [24:01]
B
Yeah. And one final thing I would add, I will say just straight up, I am someone who shares Sam Altman's view that I hate ads. I don't get a lot of value out of them. They muck up a lot of the Internet. For me personally, however, there are some very real use cases for ads in Silicon Valley. Like I think people forget. Like I've seen some really good posts on X about this that like first, the more compute you can buy, the more low cost intelligence OpenAI can provide you. Therefore, ads fund compute. Therefore a lot of people are willing to make the trade off. I want access to great AI. I'm willing to look at ads to do it. But also like for decades ads have funded R and D. Like some people have made the argument, which I actually get, find myself agreeing with, that if Google wasn't selling ads for the last 15 years, they would not have had the funds and the talent to create the transformer. So I think we're underestimating how much of an incredible model it is in terms of revenue for these companies to then do other R and D. Now you can absolutely make the argument, and I'd agree with it that ads are ruining a lot of Internet experiences. But there are a lot of plus sides to the technologies we have gotten from Silicon Valley that have been basically funded by ads that wouldn't have been possible otherwise.
- [25:14]
A
Yeah, I think the big strategic, like the, I guess the really interesting strategic play here is. I agree with all that. But if you're Google or Anthropic and you don't necessarily need the ad revenue, do you actually just like, is that your play? It's like, fine, let them do ads and maybe the experience becomes less and let's not use do ads yet. Yeah, because we have the funding. Or in Anthropic's case, you're making all your money through the API. You don't need ads in Claude necessarily. Then does it start changing user behavior? It's like, well, I just use the free thing anyway and all I'm doing is like, you know, these basic uses. I'll just jump over to Gemini that doesn't have ads. It's a better user experience like that. I don't know, it'll be fascinating to see how that plays out.
- [25:58]
B
Could be a differentiator for sure.
- [25:59]
A
Yeah, for sure. I mean that's. I would. If you're not going to do it, it's like, hey, come to the ad free. You just go after them with that.
- [26:08]
B
All right, our next big topic. This week Anthropic has announced the launch of something called Claude Cowork. This is a new agentic tool designed to automate general Purpose Office tasks by interacting directly with a computer's local files. So Coworker is a feature right now within the Claude desktop application, importantly only on macOS at the moment, that allows the AI to perform complex workflows with all more autonomy than your standard Claude chat. So this is basically Claude code, but for non technical people. So Claude code is built very similarly, but it's specifically, at least on the surface, meant to be for software development, whereas Cowork is intended for non technical users to manage and automate routine administrative work. So this tool entered a research preview on January 12th and basically it functions by asking users to designate specific folders it can access, and once it's permitted to cowork can then go read, edit or create files within that directory to complete multi step objectives. So people are using it for things like organizing files on their computer, generating spreadsheets from images, drafting reports. Basically you chat with it just like you would Claude, but it takes a lot more autonomous steps and can actually create and use files and tools on your desktop to maintain security. The system operates within a sandboxed folder partition and requires explicit user approval before taking significant actions, such as things like deleting files. So this is currently available only for Claude Max subscribers. The Max plan is like 100, I think it starts at 100amonth and then scales up from there. And it's only on Mac os. They have plans to expand to Windows and other subscription tiers in the future. So Paul, in my very, very early test of Cowork, I think this is pretty awesome. I think it's going to open up the eyes of a lot of non developers and people who have not experimented with Claude code. But as you alluded to at the top of our episode, Anthropic themselves have warned around some of the security issues. And my gosh, even just using it, I'm like, I'm using on my personal computer and I'm like not doing anything I think is problematic, but I'm like, wow, this could go really wrong.
- [28:21]
A
Yeah, and I thought it was funny, like episode 190, just last week you were sharing some of what you were doing with Claude code and I think we talked about like, wow, what an opportunity for some to build a user interface that makes this easy for non coders. In like 48 hours after that episode drops, here's Claude Cowork. So sort of fortuitous. So there was a tweet thread by Boris Cherny, who's the creator and head of Claude code, and he actually talked a little bit about the backstory, we'll put the, the link in the show notes. But he said like a couple months after he'd created Claude code, many engineers and Research Entropy were using Claude daily and he started noticing non coders were starting to use it to specifically, specifically talked about like finance uses and things like that. So that's when they started realizing like, oh, okay, maybe we should make like a simpler version for this. So they did publish a blog post that we'll link to as well that just talks about this research Preview. And they said when we released Claude code we expected developers to use it for coding. They did and then quickly began using it for almost everything else. This prompted us to build Cowork, which they did tweet that Claude code created Claude Cowork, that it basically wrote all the code for the user interface. So go into this whole like, you know, that what Dario Amade said over six months ago or whatever, nine months ago, that they envisioned a near future where like Claude would be writing like 90% of their code and they're kind of eating their own dog food in this case, I guess. So it says this prompted us to build Cowork, a simple way for anyone, not just developers, to work with Claude in the very same way. Cowork is available today as Research Preview for Claude Mac subscribers on the Mac OS app. And they'll improve it rapidly. It said, you know, Claude makes a plan, steadily completes it. Once you give it access to these files, it can kind of like work with those files on tasks. It talks about initial set of skills that improve Claude's ability to create documents, presentations, other files. And you can pair it with Claude and pair Cowork with Claude and Chrome and then it gives it the ability of the browser too, which is where you got to then start being really careful about what you're giving it access to. An interesting thing they highlighted that's going to become a really important topic for people is this idea of Claude working on multiple things in parallel. So one of the lead people on Claude code shared the story of how he'll have like five terminals running and like basically just. And so this is the idea like when you start thinking about the impact on organizations. And again, this is a very early form of AI agents being integrated into workflows. But imagine like, okay, I'm going to do a product launch and I have the project management components, I've got the marketing tune there thing, I've got the go to market overall where I'm thinking about the sales and customer success. And right now you would have people on each of these teams doing a thing, what they're proposing here. And the not too distant future for people, sort of early adopters of this is, hey, we're going to launch this product in March. Let's spin up an AI agent to develop the plan for all the go to market. So the marketing plan, the sales plan, the customer success plan, all the documentation that needs to be created, all the email workflows that need to be built, the websites that need to be created, the media plan that does the media. But like let's go ahead and build that plan and then maybe the human approves, edits and approves that plan and then imagine just turning it loose. And now in parallel, the marketing agents, the sales agents, the customer success agents, they're just doing the stuff and you give it access to a file system with all the past campaign data. Like all these things. Yeah. And then like it comes together. And so now maybe the CMO is overseeing the team of agents that are working in parallel over here and the sales leaders do. And this may sound. What's the minority report? Is that the Tom, this may sound very minority report to you. This is not sci fi. Like this is the kind of stuff that people like anthropic are doing internally. Like they're trying to test this and figure out can we build these agents that work in parallel and do all of this campaign stuff. And so to the point you and I, Mike talked about recently about, like what do entry level people do? Like what, what are the roles for entry level people in, in these really AI forward organizations that are like figuring all this out fast? Because all those things I just described is, is what the entry level people were doing. Like they were doing the tactical parts of the campaign. But if the AI can do that now, all you really need is a director level person with experience and instinct and intuition who can just like oversee and review all this work and maybe go in and do some edits and things like that and kind of get it to the finish line. But it starts to change the way work is done. And cowork's a really good name for, for this. So this idea of agents working in parallel is going to be really important concept for people to get used to thinking about and hearing. They then talk about being able to connect to the folders and other Claude connectors, which starts getting you into this like really dangerous range. Yeah, so it said that. So this is right from their thing that said there are still things to be aware of before you give Claude control. By default, the main thing to know is that Claude can potentially destruct take destructive actions such as deleting local files if it's instructed to. Since there's always some chance that Claude might misinterpret your instructions, you should give Claude very clear guidance around things like this. And then they also address the idea of prompt injections. You have to be aware of them. These are attempts by attackers to alter Claude's plans through content it might encounter on the Internet. We've built sophisticated defenses against prompt injections, but agent safety that is the task of securing Claude's real world actions is still an active area development in the industry. So this is like use at your own risk for sure. I obviously have not tested this, but there was one that got a ton of views of 1.3 million views. This is a tweet from Clairvaux, who is the founder of Chat PRD and the host of How I AI podcast. And she had a really good synopsis. It's only maybe like 800 to 1000 words. You could read it pretty quick on Twitter or X, but she said what is broken at the end? And it said honestly, a lot. Connectors didn't connect. Terminal commands failed with scary error messages. Local files like X didn't load when I clicked View file, but did load in the inline tool message. It showed too much context, too many scripts exposed too much. It asked me to approve opening files way too much. It connected mcps. I didn't ask for too long. Don't read. Research Preview was previewing. The challenge with this sort of thin wrapper on Claude code UX is that it's not quite optimized for the non technical and too kneecapped for the technical people. Cowork right now sits in fuzzy middle and the team is going to need to optimize for one or the other to win over a new audience. My final note here is update your generative AI policies. Like if. If you're. If your employees are unclear if they're allowed to try this on their systems, you should be clear about that. So that is one takeaway. And then again, I think this is all just a preview of what's to come, you know, in the coming months. Gemini Chat GPT Everybody's gonna be doing this. So yeah, we gotta start being prepared for this and, and know how to handle it within enterprises.
- [35:33]
B
Yeah, your job as a knowledge worker is going to involve a heck of a lot more orchestration and management. Whether it's of P, it's of AI agents, not people. So if you're uncomfortable with that, I don't know what to tell you, but that's going to be like a skill for sure.
- [35:47]
A
Yeah. And that's, you know, somebody asked me, I think it was maybe on the intro to AI class we did last week, somebody asked me about, like, what are some of the titles? And you just hit on the one. AI Orchestrator, AI Orchestration, AI Ops. Orchestration. Like that idea of orchestration of all these agents. It might be something that you actually have a title for. It's probably like a. Because it's not even an ops thing, honestly. It's a strategist. It's probably a director VP level who is orchestrating. I guess you could have lower level project managers who become AI orchestrators and they're focusing more on did it follow the processes and the workflows? Did it stay within the confines of what it was supposed to? Did it only access the files it was supposed to. Like, you're going to need that level of project management ops. And then you're going to need, okay, is the strategy sound? Is it making sense? Like, what? Is this the kind of thing I would build? Did it have new ideas? Like, yeah, so orchestration is a really interesting one. I can see a web of roles that have to emerge here real fast.
- [36:48]
B
I wonder too, if it ends up, who knows, but if it ends up being particular to departments. Because I'm even thinking of your marketing examples.
- [36:55]
A
Right?
- [36:55]
B
Like, I sure as heck don't want a random VP who doesn't understand anything in HubSpot or a general Ops person.
- [37:01]
A
Who doesn't know marketing. Right, Right. It is. Yeah. It's interesting. I don't know, I, I haven't really thought about that before.
- [37:09]
B
All right, our third big topic. Let's get to some juicy drama here because a federal judge has ruled that Elon Musk's lawsuit against OpenAI and Microsoft will proceed to a jury trial this spring. This legal battle centers on Musk's allegation that OpenAI abandoned its original nonprofit mission by transitioning into a commercial entity. So Musk, who helped launch OpenAI in 2015, had given it 38 million in seed money, and he claims he was defrauded after the startup took billions in funding from Microsoft and its affiliates. Now we'll talk about this. But OpenAI disputes this version of events. They've released communications that seem to show Musk is, has actually been in favor of a for profit solution in the past. And now we have thousands of pages of unsealed evidence that's recently been released in this case that offers some new insight into a bunch of internal friction at the company. So, deposition transcripts revealed that prior to Altman's brief firing in late 2023 board members cited six key concerns about him, including misrepresentations regarding safety board reviews and the release of a test model in India without oversight. The records also show that former chief scientist Ilya Suskavok held an estimated $4 billion invested equity at the time of the leadership conflict. So, in this new filing, Musk's legal team is seeking between $79 billion and $134 billion in damages based on OpenAI, a $500 billion valuation. OpenAI and Microsoft have dismissed this demand as, quote, unserious. The jury trial is scheduled to begin in Northern California on April 27th. Now, Paul, lots, lots more to unpack here. Can you maybe walk us through this? Because Elon Musk got involved in a very vocal way in the last week.
- [38:55]
A
This is oddly public. Like, I. I can't. I mean, maybe this is their attorney strategies to allow them to talk about these things publicly, and they need to do it. But. But I can't remember a case that's going to trial where the defendants were this public about, like, all this information. So it is. It's crazy to follow, but I do think, like, the inflection point here was the release of the page, like, the thousands of pages of evidence. And then everybody obviously starts jumping in and analyzing this stuff. But it does have depositions from Altman, Ilya, Greg Brockman, Mir Maratti, Thinking Machine, Satya Nadella. Like, all this stuff's in there. And past board members like Helen Toner, Tasha McCauley, who played key roles in Altman's firing. Like, all this is in there. But the. The thing that started triggering the tweet war basically, was Elon retweeting and commenting on things out of Greg Brockman's personal journal or diary where he was, like, you know, just, I guess, contemporaneous. Like, here's his notes, his thoughts, all these things. So on January 16th, there's basically, there's a. There's a few accounts online. I mean, it's probably lots of them, but there's a few that get surfaced all the time for me that are very pro. Elon may Burner accounts relaunch. I don't know that he often will retweet. So there's Guy. I think it's Sam Merits, the one guy, the one that. That kind of started the firestorm last week. His Whole Mars catalog. It's like. And they always have information that, like, no one else seems to have, which is why they're like. I assume they're, like, pretty close to Elon. So Elon retweeted on January 16, and he said they open. So this is Elon's tweet was they openly discussed their comp. Conspiracy to commit fraud and steal the cherry. That was Elon's exact words. He was retweeting a whole Mars catalog post that showed a screenshot from the files that were released. And it says Brockman's personal files. And so this is what, like, the. The evidence says it says discovery in this case likewise revealed Brockman's private thoughts about OpenAI's structure, which Brockman documented in his personal files. Musk refers to these entries as a diary. Brockman wrote in September 2017. So, again, just for fun, context, this is about four months after the invention of the transformer. So we are. We're in the very formative moments of, like, generative AI. Google themselves haven't even, like, figured out yet the significance of their own creation. And GPTs hadn't been built yet. But we're just kind of in that point where we're starting to realize this is all about to take off. So just for context, so here's what his diary says. This is the only chance we have to get out from Elon Musk. Is he the, quote, unquote, glorious leader that I would pick? We truly have a chance to make this happen financially. What will take me to $1 billion? So, again, this is Greg Brockman's notes to himself. Basically accepting Elon's terms nukes two things. Our ability to choose, though maybe we could overrule him. And the economics. The evidence then says, after Brockman and Altman reaffirmed to Musk their commitment to OpenAI's nonprofit structure, Brockman discussed OpenAI's future with Sutskova. On November 6, 2017, Brockman wrote after the meeting that the conclusion is we truly want the B Corp. Honestly, we also want to get back to work, but it's not super clear how we get there. And then it actually quotes, he continues, cannot say we are committed to the nonprofit. Don't want to say that we're committed. If three months later we're going, we're doing B Corp, than it was a lie then it says in evidence. Brockman conceded that he was not feeling so great about all of this. The true answer is we want Musk out, he wrote. Can't see us turning this into a for profit without a very nasty fight. I'm just thinking about the office and we're in the office and his story will be correctly, more correctly be that we weren't honest with him in the end about still wanting to do for profit just without him. After the November 6 meeting with Musk and after Brockman and Altman informed us, they remained opening eyes. And then it kind of keeps going. So that was the thing that Elon Musk tweeted that they were like just basically trying to commit fraud. So later that day, Sam Altman now gets involved and he tweets the face melting emoji and then he links to OpenAI's post. So then OpenAI shares a public post about this and Sam Altman says Elon is cherry picking things to make Greg look bad. But the full story is that Elon was pushed searching for a new structure and Greg and Ilya spent a lot of time trying to figure out if they could meet his demands. And then he followed that up, said, I remembered a lot of this, but here is a part I had forgotten. Elon said he wanted to accumulate 80 billion through OpenAI for a self sustaining city on Mars and that he needed and deserved majority equity. He said that he needed full control since he's been burned by not having it in the past. And when we discussed secession, he surprised us by talking about his children controlling AGI. He said, I appreciate people saying what they think and think it enables people to resolve things or not. But Elon saying he wants the above is important context for Greg trying to figure out what he wants. Then Greg Brockman tweeted, so again, I'm watching, like, how are the attorneys allowing these people to be doing this? I'm not an attorney, but this seems really unusual. So then Greg tweets, I have great respect for Elon, but the way he cherry picked from my personal journal is beyond dishonest. Elon and we had agreed a for profit was the next step for OpenAI's mission. The context shows these snippets were actually about whether to accept Elon's draconian terms. So then the post that OpenAI put on their own website was, quote, the truth Elon left out. That's the title of the post. And so it says, the truth is that we and Elon agreed in 2017 that A for profit structure would be the next phase for OpenAI. Negotiations ended when we refused to give him full control. We rejected his offer to merge OpenAI into Tesla. We tried to find another path to achieve the mission together. And then he quit OpenAI, encouraging us to find our own path to raising billions of dollars, without which he gave us a zero percent chance of success. We did find a path to advance the mission with OpenAI now structured as two main entities, a Public Benefit Corporation or the B Corp, and a controlling nonprofit which owns equity in the PBC. The Public Benefit Corporation, currently valued at approximately 130 billion. So that's the nonprofit is now since they restructured in the fall. Elon's latest variant of his list lawsuit is his fourth attempt at these particular claims and part of a broader strategy of harassment aimed at slowing us down and advancing his own AI company, xai. He is now grossly misrepresenting the written record to further his harassment. This post shows just a few examples of the truth that Elon isn't telling and then it goes through and actually shows these notes specifically with Greg Elon and Ilya in a conversation they had that kind of reiterates this stuff. And shortly after that meeting in November 2017, Elon actually created OpenAI PBC A B Corp. So he went to the step of doing this and then it gets into the self sustaining thing. So then again, just fascinating. This is all happening in public. OpenAI's message to investors so this is part of my consideration is they're trying to raise all this money like this uncertainty is not good for that. And so they're being very proactive about trying to like clarify for investors what's going on. So then a guy, Alex Heath of Source News, got access to a message that was sent to investors about the lawsuit and here's what this read an excerpt of that. It said this is again OpenAI to their investors. We have strong defenses and feel confident about our chances of winning the case. Regardless, based on the record so far we believe this case is worth no more than the 38 million that Elon donated. Though that is not a guarantee, we nevertheless expect Elon to make deliberately outlandish, attention grabbing claims aimed solely at generating headlines. To be clear, such claims would not be grounded in reality and would instead be typical of the harassment tactics he's previously deployed. See his fake bid to buy OpenAI in 2025. Elon's lawsuit remains baseless and without merit and our team is focused on ensuring the jury sees these claims for what they are. So I don't know, like again not a legal expert here, but if this does actually go to trial it it will be crazy. I was just trying to think of like all the people they could call like Dario Amade, who doesn't seem to be the biggest fan of OpenAI was obviously right in the midst of all of this, but back in those days, so you have competitors like Ilya Mir Maratti who just had people poached from there, Dario Amade, people who might have like a, an axe to grind with Sam and OpenAI who could absolutely be pulled in to tell part of the story, who also, by the way, are probably equity holders in OpenAI. Like we mentioned, the thing about Ilya, like Ilya owns, it sounds like probably somewhere between 8 and 10% of OpenAI equity. So if they're valued at a trillion or trillion and a half, 2 trillion when they go public and Ilya owns 10% of that, like, does he want that company to go under? Probably not. Like, that's a big. So you have these people who could be called as witnesses who could maybe harm OpenAI's case, who still own equity in Open AI.
- [48:29]
B
Yep.
- [48:29]
A
It's just wild. Satya Nadella, obviously from Microsoft. There's all kinds of Microsoft employees that could get pulled into this. Mustafa Solomon, past board members who aren't big fans of Sam Altman. So this creates all kinds of uncertainty for OpenAI. And I do think that that is the whole point. Does Elon need, even at the high end, $134 billion? Hell no. Like, he doesn't give a shit about $134 billion. He's worth a trillion, basically.
- [48:54]
B
Right.
- [48:54]
A
So, like, nothing he would get from OpenAI is going to change his lifestyle at all. So this absolutely just seems like let's just create the most chaos possible. Let's slow him down. Let's, let's, you know, use it as an advantage for xai and let's just make their lives miserable. Because I can and I've got the money to do it, but it's going to be absolutely wild.
- [49:18]
B
I also just think, God help your average juror who has to figure out all the nuance here. Seriously, that's got to be so crazy. I mean, I couldn't even imagine we're having trouble pulling on all the threats. Having your average person who's just kind of thrown into this figure it out is, is wild.
- [49:35]
A
Well, like I said at the beginning, like, you, you could. I don't even think a movie about the last like eight years of AI would be sufficient. This has to be like an ongoing series because every episode could be something crazy.
- [49:49]
B
All right, let's dive into some rapid fire topics. This week. We've got some More XAI and OpenAI news up front here. So first, XAI has announced a series of technical updates to its GROK AI model following the global backlash that we covered last episode over its generation of non consensual sexual de fake. So the model was initially designed with fewer guardrails than its competitors and as a result it recently drew intense scrutiny after users started to use its image editing features to digitally undress real people online. So the platform has now implemented geo blocking measures to prevent users in certain jurisdictions from generating images of real people in revealing attire. Additionally, the ability to edit images using GROK will remain restricted to paid subscribers, which X claims will increase accountability for those who violate its policies. So this update follows legal and political pressure. There's an ongoing investigation by the UK regulator Ofcom and a probe by California's top prosecutor into the spread of sexualized deepfakes. X has not detailed how it will distinguish between real and imaginary subjects or how it plans to enforce these rules against users using VPNs to bypass regional blocks. So Paul, since we talked about this last week, we're at least seeing some walk back here. I can't personally help but think this just isn't enough. I'm curious how non consensually undressing other people publicly using AI is a free speech issue like Musk has claimed, but I guess at least they're doing something about this.
- [51:25]
A
Yeah, I mean I think the jurisdiction word you use is the key here is like they're basically just pulling back in areas where, you know, people are giving them a hard time. It seems like they're just going to do whatever is the least they can possibly to adhere to whatever laws are, you know, people are going to push on but I, I don't feel like they have any personal desire to actually resolve this and make this so it's better for society. I think they're just going to keep doing the minimum viable thing to, I don't know, just keep pushing this forward and, and make it a non issue for them. As from a business perspective this is like I always like in the early days we were talking about like open source versus proprietary models and things like that. This was always my biggest concern with open source models it's like yeah, you know you can stop this on, on Instagram and Facebook if you want if you had motivation to which you know, Facebook doesn't care. Like they're, I mean they can pretend like they do, but they don't. It creates more user time and things like that that so you know, these Proprietary models and distribution channels can make these choices and, and they can control these things like they can monitor who's actually generating these and shut accounts down, things like that. But the same technology lives in open source models. So yeah, you can suppress it all you want, but it's not like it's going to stop people who actually want to do this from doing it. And so this really in the end, I mean they should. So like actually I should do more, they, everybody should do more. But it becomes a social media distribution problem more than anything. And that's where X comes in, is they're allowing this to happen, like to be distributed. So GROK being able to do this is one thing. Yeah, and them allowing users to do it with the GROK technology is one thing. Allowing them to then distribute it on X is the major problem here because again, you can go do this with open source models. GROK could shut this capability off tomorrow. You can still do it. And people who have a lack of morals who are doing this already are going to do it regardless. To me is probably a larger social media distribution problem. And that's where it's on these companies to shut this down. No matter what model is used to do it, they have an obligation to shut it down. And I just don't see that a sense of urgency around that right now. That's a major problem.
- [53:43]
B
All right, next up, OpenAI's chief financial officer Sarah Fryer has released a new update on the company's business model and growth trajectory. So according to this release, OpenAI's annual recurring revenue reached 20 billion in 2025. That's up from 2 billion in 2023. The tenfold increase in revenue followed a similar expansion in compute. The company reported its compute usage grew from 0.2 gigawatts to nearly 2 gigawatts over the same two year period. So Fryer stated that the company's ability to serve customers currently tracks directly with the available hardware. To sustain this growth, OpenAI is diversifying its infrastructure providers and moving beyond its initial reliance on a single partner. The company's financial strategy now includes a multi tier revenue system. This consists of consumer and workplace subscriptions, usage based fees for developers, and of course the free tier supported by advertising and commerce. Looking ahead to 2026, the company is prioritizing practical adoption in sectors such as healthcare and science, alongside the development of autonomous agents that can execute tasks across different software tools. So Paul, some interesting numbers here, some official numbers from OpenAI. I mean that growth is explosive. It was also interesting, she Also hinted in this statement that OpenAI plans to introduce some type of outcome based pricing in some of the more complex domains they're doing business in. So not for I think your average user, but more in areas where they're trying to kind of share in some of the value that AI is creating. Like pharma or R and D. Yeah.
- [55:19]
A
It was, it was interesting and a bit refreshing to see actual data come out from them rather than getting leaked through information or Bloomberg or Reuters, because that's usually how we have to get at this data. I'm not sure about the timing. Like anytime I see something like this that's a anomaly with how they would normally communicate. It's always like why, like why are you doing this? Why are you sharing this? Now, I assume it all ties to need for funding and things like that, you know, to try and drive confidence in investors as they're trying to raise this 100 billion and, and more. So I, I would guess that, and they're probably starting to set the stage for the ipo. I would imagine you start kind of putting that sort of things in place. But the, the 20 billion number was interesting because I think the, the guidance was like around 13 billion was what everybody was kind of assuming this year. And then Sam in that one interview in December, he got kind of pissed at the interview. I forget which one that was.
- [56:09]
B
Yeah, that was on BG2 with yeah, yeah, Brad Gerstner. Right.
- [56:13]
A
I think, yeah, was that. And he got like really pissed and told him to sell his shares and stuff. That's right. Because he's like what about maybe like closer to 20 billion and things like that. So Sam had alluded to this was the run rate they were on, but now they've publicly said it. So yeah, I, I, that was the most interesting part of all this to me was the fact that they communicated this. And why would they communicate this right now? But I don't know, it seems like there's just sort of a lot of things right now is like laying the groundwork for the next five years.
- [56:43]
B
Next up, Google has launched a beta feature for Gemini called Personal Intelligence. This is designed to integrate user specific data from across the Google EcoSystem into its AI assistant. This update allows Gemini to securely access private content from Google applications like Gmail, Google Photos and YouTube. This is currently rolling out to Google AI Pro and AI Ultra subscribers in the U.S. it basically enables Gemini to actually reason across your data. For example, it could now maybe suggest books you'd like based on interests found in your emails or in One of their examples, they show it recommending new tires for your car based on your driving history in your photos and vehicle specifications that may be in your docs or emails. Google states that while the system references this private content to answer specific requests, it does not use the raw data from a user's inbox or photo library, for instance, to train the underlying models. Now, access is off by default. It requires users to manually choose which apps to link. The company has also implemented guardrails to prevent the AI from making proactive assumptions about sensitive topics like personal health. Now, while the feature is presently restricted to premium personal accounts on web and mobile, Google intends to expand access to its free tier in the future. Now, Paul, this really seems like one pathway to that type of, of truly personalized AI assistant, which is a concept we've talked about a bunch of times in the past. And I guess for me at least Google certainly has some of the most useful data to work with here. What was your experience with this?
- [58:21]
A
Yeah, so I did turn this on. I was, I was really curious about this. And now the difference for me here with like would I do these things with chat GPT vs not my. My theory is like Google's got it all already. Like I already trust them with Gmail and Calendar and all these other things. I don't use Google Photos. That one's kind of weird to me because it it like there was the example about the license plate where yeah, I can't remember my license plate and it has photos of your license plate. Like I the photos one was weird to me but so I turn it on. I go into Gemini app in my personal account and it pops up. It says supercharged Gemini with personal intelligence. Make Gemini uniquely helpful for you by connecting more of your Google apps, all while safeguarding your info. This is a beta feature. You can turn it off at any time. So I click get started then takes me to a screen that says connect these Google apps to Gemini and it gives you options. Gmail, calendar, drive photos, YouTube and search services. YouTube is your viewing history and then search services. Anything you ever search for basically on Google and it's retroactive like it goes back in time through their data. So you have to choose those things and then it gives you like a things to know data share between Gemini Connected X includes things like your saved data from search, YouTube, emails, files, events, photos, videos, location information that Google Maps as well as insights drawn from them. Data shared may relate to topics you find sensitive. So they give you all these things and then you got to approve that and then you go through and you choose. You choose apps don't connect or connect. All are the options. And then once you go through that, it's like connecting your apps to Gemini. It does the thing and then it says, okay, your apps are connected to Gemini. And so that was. That was that. And then I go in this morning and I was, I. Because I wanted to prep for this and I chose a. It says like, for you. So I'm in Gemini and it says for you. I was like, oh, let's see what this is all about. So then it pops up and says, hello, it's great to officially meet you. So this is now Gemini talking to me. I am ready for this new chapter. Now that I'm connected to personal intelligence, having this context allows me a much more effective partner in your work and life. Based on your current activities and interests, here are three themes that highlight what makes your perspective so unique. Unique. And then it goes through. And I've got the human centric AI visionary. And it talks about, like, the work you do in AI and the book you're thinking about writing. And it's like referencing some specific things. The dedicated coach Persona talks about the fact that I coach my kids and I research things related to K pop groups for my daughter. And you have this generation. So I mean, it's, it's getting into some stuff. And then it says, the modern Renaissance hobbyist. Your downtime is fascinating mix of precision and relaxation. You balance the high stakes world of AI with the tactical, patient world of LEGO builds. And specifically those intricate, intricate botanical and Mario sets and a strategic focus on golf. And then it talks about these other things, like home projects and things like that. And so I was like, all right, I'll play along. So I said, I do enjoy Legos, but the botanical and Mario sets were white. For my wife, who loves Legos, my daughter does as well, it's like, oh, that's a great clarification. Like, what kind of Legos do you like? And so you have this conversation. So then I didn't explore this, but I, I thought, I'll keep playing along, like after the fact, but I said, why don't you just interview me so you can learn what you need to know? When I have time, I'll answer different questions. And it said, that's a fantastic idea. A slow burn interview is a perfect way for me to fill in the gaps. I'll keep these questions tucked away and you can jump in whenever you have a spare minute between keynotes or Lego builds. We can Call this the Context project. And then it starts with like the first two questions. So that's been my experience so far. I don't know, I'll be, I'll be really interested to see because I could imagine as I start having more of these conversations and using as like a financial advisor personally or business plan or like new businesses in my life, things like that, that it knows all this and it's referencing emails. So I don't know, like I'm, I'm optimistic about this. I'm actually kind of excited about the personal intelligence thing.
- [62:10]
B
I'm really bullish too. It's kind of exciting. You know, I did similar experiments and just very briefly, one useful thing I found to get started is, you know, I have a pretty extensive YouTube history, which is pretty helpful. So I just was like, hey, you know, based on what I'm interested in, I'm always on the lookout for new podcasts. Can you recommend some? And you know, I have a pretty good sense of the popular podcast landscape, but it recommended a few shows I actually hadn't heard of. I was like, oh, I'm going to check that out. So that's cool. It's pretty cool.
- [62:36]
A
Yeah. Yeah.
- [62:38]
B
All right, so we have talked about Google personal intelligence, but we've also got some more Google news. Read. Google and Apple. So this past week, Apple and Google announced a multi year partnership to integrate Google Gemini into the Apple ecosystem. So under the agreement, Google's technology will serve as the primary foundation for Apple's next generation AI models. Most notably, this infrastructure will power an overhauled, more personalized version of Siri, scheduled for release later this year. Apple leaders confirmed that Gemini was selected following a careful evaluation determining that it provided the most capable foundation for Siri to understand personal context and execute in app actions. Despite this collaboration, Apple did state it will maintain its privacy architecture by processing requests either on device or through its proprietary private cloud compute system. The announcement follows multiple development delays that we've discussed extensively that pushed the Siri upgrade into 2026. Financial terms were not disclosed, but some reports indicate Apple may pay approximately $1 billion to Alphabet as part of the deal. So, Paul, we've been circling this story for a while. It seems like Apple has finally made a decision on who to partner with to make Siri more intelligent. How big a deal is it that this is Google?
- [63:59]
A
I think it's a big deal for OpenAI. I mean, they already have ChatGPT integrated. So last year there was a partnership and that still exists between OpenAI and Apple, where you can call on ChatGPT if like Siri doesn't have the answer for something something so I assume that goes away. I would guess this is a relationship Apple or OpenAI would have liked to have gotten. My guess is a few things are at play here. One it definitely and we'll talk about this in the in the next topic OpenAI seems to be more directly going to compete with Apple both on products and services. So products might be the big one here is they don't want to you know, partner up with somebody who's going to come at like the iPhone or AirPods, things like that that I, I, I got to think some element of open eyes uncertainty plays into here. Like think about what we just talked about with the musk lawsuit and what happens to OpenAI moving forward. Like Apple's not the kind that like loves uncertainty. Like they tend to be a little bit more conservative in their partnerships and things and Google's a known entity. They've worked together on search on maps like they they've done big partnerships before even though they also compete. So I think there's maybe a comfort there and a history that's more stable TABLE so I don't know and I think all this just buys time until Apple figures out their on device play and fine tunes their own models. And again historically Google's been okay with that. Like they paid Google for Google Maps for years until they build Apple Maps and now they coexist and users choose and yeah, so I don't know, I feel like it's just following a precedent that has worked for both companies and they find ways to do deals when it benefits both of them.
- [65:32]
B
All right, so let's get to that next news bit here about OpenAI because OpenAI and former Apple design chief Johnny I've are reportedly finalizing their first consumer hardware device which is an AI wearable codenamed Sweetpea. This project marks the first tangible output from OpenAI's multi billion dollar acquisition of I've's hardware startup which is called I.O. and a manufacturing partnership with Foxconn. So according to some recent reporting from Gizmodo, a supply chain leaker online indicated that Sweet Pea is an audio centric device that deviates from traditional earbud designs. So rather than sitting inside ear canal, the hardware consists of two pill shaped modules designed to be worn behind the ear. These components are stored in a metal egg shaped charging case when not in use technically the device is expected to feature a 2 nanometer smartphone grade processor and custom silicon developed to facilitate Voice driven interactions. These leaked details suggest the interface is built to execute iPhone actions by commanding Siri, effectively serving serving as a hands free gateway to ChatGPT. While the device is intended to reduce reliance on screens, it is positioned as a companion to the smartphone rather than a standalone replacement. Manufacturing is currently underway for a Projected release in September 2026. Internal targets suggest an annual or initial production run of 40 to 50 million units as part of a broader roadmap of five upcoming AI devices.
- [67:05]
A
Devices.
- [67:06]
B
So Paul, if these leaks end up being true, this is a pretty interesting play where OpenAI may just be coming for AirPods. What, what do you think of the details here?
- [67:17]
A
I'm trying to picture like there's a, there's a drawing in the. What is this, the Gizmodo? Yeah, one I'm trying to picture how that would work and if I would wear one of these things. I don't know. Again, like this is all rumors like we talked a few episodes ago about like a pen maybe? Yeah, yeah, I'll be interested to see. I do think a device that works with the iPhone is the best play and gives them the best chance if it's complimentary too. So yeah, I guess we'll wait and see. But it's interesting to more of these rumors, more that are being sourced from the supply chain, which is where you're going to probably learn about this stuff first. So it's good to keep an eye on.
- [67:51]
B
And for anyone who doesn't realize AirPods are an absolutely gargantuan line of business for Apple, they don't have have total specific numbers, but analysts estimate that In, I think 20, 23 are the most recent numbers I could find. They did $18 billion in revenue alone for AirPods, which make them like 5 to 7% of Apple's total revenue. Analysts estimate more than a hundred million active users of AirPods. These are enormous. These alone, AirPods can be their own company.
- [68:20]
A
Yeah. And I, I mean I have three pairs of AirPods. I mean I have them in strategically placed. It's like, because I always misplace them. It's like one's in the gym bag, one's in the work bag, one on my desk. I thought, I mean I, like this is fine but like what's saying that Apple doesn't just put cameras into the airpods and improve the batteries and like my bet would be on Apple if it comes to the innovation from that perspective and the supply chain and all that stuff. So yeah, maybe this is where they go, but it doesn't mean AirPods next generation don't, you know, get out ahead of them and make a better product. And they already have the distribution, so we'll see.
- [68:54]
B
All right, next up in the last couple weeks, Paul, we have been sharing some ways we have personally been using AI in our work and lives and the audience has really responded. They've reached out multiple times saying they like hearing these use cases. They're valuable and helpful. So we wanted to do a segment this week under maybe we'll see if it becomes kind of a recurring weekly or periodic segment that we're nicknaming right now. You know, our AI Native Chronicles and we just wanted to share what we're kind of up to to with AI in the past week. So Paul, first, why don't you maybe share anything you've been doing with AI this past week, any challenges or opportunities you're running into. I have a quick thing to share as well.
- [69:32]
A
Yeah, there's two quick ones I'll I'll share that. I put in the Exec AI newsletter and then I, I did put share that editorial on on LinkedIn on Sunday, but one was I was traveling to Chicago for a talk Friday morning with Good Karma Brands, which by the way is doing some awesome stuff. I know they're listeners to the podcast podcast, but they're doing some really cool stuff on the AI transformation front. I got a chance to meet with their digital team and talk with them. So I was doing an Intro to AI class on Thursday from noon to 1 and then I was going to drive to Chicago because I like the if it's under six hours, I'll just drive like turn the autopilot on the Tesla and can do phone calls and meetings. And you're still paying attention to the road but like nowhere near as much as you normally need to. And so I like long drives. I find them crazy calming. Wednesday night Ashley and our team messaged me like 9:30. She's like, are you sure about that drive? Because like the weather does not look good. So I go into Gemini and I was like, hey, here's the situation. I'm going to take this drive about 1:30pm Eastern Time on Thursday. I would leave. You know, what do you think? The weather's not looking great. And it goes pulls the weather forecast. It comes back and says hey listen, do not drive one. It could be an eight to ten hour drive. Here's why. The heaviest snow swells are going to actually hit this stretch in Indiana that you're going to have to go through through just change the flight. Here's flight options. And it went and pulled flight options for me, which I'd already kind of looked into. But it's like, take this direct flight. It'll be the best there. It'll get you plenty of time. And then take this return flight back after your talk's done on Friday. So it's just like this super practical, but, like, saved me 20, 25 minutes of like, bouncing between things and searches. That was awesome. The other one that I was super geeked about that I reached out to you, Mike, about was org chart app. So I mentioned Lovable, I think on episode 190 that I've been starting to test. And again, I'm not a coder at all. I have zero coding ability. And Lovable allows you to build apps with prompts, like, same way ChatGPT work, but you just give it a prompt and it builds things. So I've for years been trying to find an org chart builder that could help me visualize different structures for the organizational design of the company. And always been disappointed. And like, sometimes you got to pay, like, I don't know, like Chardio or something like that. Like, there's these apps where I'll pay my 20 bucks a month. And it's like, org charts are one of the capabilities. And so you pay the thing and you get in there like, this sucks. This, this is a new half of what I wanted it to do. So I just always give up. And I try and like, use Apple Keynote or Excel or whatever. Like, I come up with these alternatives and they all suck. And so I was like, I wonder if I could do that in Lovable. So I literally said that. I'm like, yeah, I need to build this interactive org chart. Here's what I wanted to do. And then it interviews you. It goes through, like, step by step. Like, what do you want the functionality to be? What do you want these features? And so because I've thought about all this for, and I'm a CEO, like, I've done organizational design. Like, I. I could just prompt it and I could talk to it. And after like 10 minutes, I. I just said, go. And I have this like, minimum viable product. And I was like, oh, like, how about, let's change this? Let's add color coding to that. Can you make this shape look like this? And things that I used to have to, like, email a developer, say, hey, here's five changes I'd like to make. And then like 24, 48 hours later, I'd get the comp sets back from the developer. They are not what I wanted. Like, it's this like super iterative, very long, drawn out process. Now it's like, well, what about this? Like, let's take a look at this. Okay, great. Ten seconds later, boom, here you go. In production. Like, it was amazing. And it. I started using it to build an initial model and then I messaged you, Mike. I was like, dude, you got to see this.
- [72:58]
B
Yeah, it was really impressive.
- [73:00]
A
Yeah. So the thing I told Mike when we talked was what I'm starting to feel, and I might have mentioned this on episode 189 or 190 was like, we're three years into Gen AI and there are still a ton of companies that haven't figured out how to use ChatGPT or Google Gemini or whatever. Like there's still just this fundamental basic gen stuff. I feel like we're gonna have the same adoption curve to where like you can just build things. Like you have an idea like a software you need or an app you need and you don't have to deal with all the complexity of actually releasing it to the public and all of that stuff. Like this is just like internal tools I'm trying to build, build. I can just go build them. And that knowledge, that ability to run a company that way, or a team or campaigns, like where you can just build tools, that is going to take years for the rest of the business world to adapt to. But for those of us who know how to do it right now, like you have this massive head start to go do these things. So that was. Those were the two. One I was really excited about. And then Mike will like get to yours. But the one thought I had that I had shared with you, Mike, is like, I'm. I don't know what to call this. I was calling it AI overload, for lack of a better term, because I now have access to build whatever apps I want to have, these strategic conversations I want to have. I often am finding myself like, I'll do these things in these like down moments and then like a month and a half goes by and I'll be like, oh, I should talk to ChatGPT about this thing, or I should go build this app, whatever. And then I'll get in there and I'll start like the thread and it's like, wait a sec, second, I think I've had this conversation already somewhere. And then you're going into Gemini and ChatGPT. Where is that? And this, to me, the biggest issue is I said in the early days of all this like, I just want to go into Google Docs and have Gemini in there. I don't want to have Gemini out here where I'm having these separate conversations that I have to like open and create a doc and go over there. It's like I forget that I have these conversations and so I'm like overloaded by what's possible and it makes it really hard to keep track of all the stuff that's going on on. But yeah, really cool stuff and exciting to see.
- [75:04]
B
That's so true. I find myself struggling with the exact same thing. I'll just quickly share just an example of a very small thing. Maybe this is very easy for someone else to do, but honestly it was a bit transformative for me as I'm experimenting more with Claude Code. 1. I actually worked with Claude, not Claude code, just Claude to like brainstorm, like what are actually useful things for me to be able to build in my personal life that are worth the effort. Because you can build anything, but the right thing is a very different thing. So. So long story short, I kind of hit on this idea. I record a lot of voice memos on my phone using the voice memo app. I have probably hundreds of hours from 2025 alone. Super useful. They're all in voice memos already. I often will pull the transcript, which is auto generated and drop it in AI and use it for something.
- [75:46]
A
Right.
- [75:47]
B
So what I thought it'd be nice to go a step further. I was like, ah, it can't be that hard to have these auto transcribed all into the same Google Doc. So it becomes a database of everything I talked about automatically across my computer. Cloud code built this in 10 minutes and there were some, there were some tweaks I wanted to tweak. I was like, oh no. Make the voice memos appear in reverse order, like the most recent ones first. Because my goal is like the next step here will be to layer AI over this. I literally have created a digest of email that emails me interesting articles every day. I'm going to do the same thing for this and say every 24 hours, email me an AI summary of what I talked about, what I said to myself. It could be anything. It could be little stuff, big stuff. There's no downside to just recording every thought I have at this point. Point. So long story short, that's up and working super easy. I get email alerts if it fails. It's just incredible. It planned it all out and it was incredibly easy to tweak and troubleshoot as well. Now obviously, if some big thing breaks. I have no idea how to fix it, but hopefully Claude code does.
- [76:48]
A
Well, that's an example of where these like always on products like OpenAI is working on could be valuable. Is like you're hacking together this idea but it's like that's a great use case if you actually have an always on and you're just talking to the pod all the time instead then it summarizes it and send it to you each day.
- [77:03]
B
Yes.
- [77:03]
A
Yeah.
- [77:04]
B
So it's really cool.
- [77:05]
A
Yeah.
- [77:06]
B
All right, so next up, a couple more topics this week and then we'll wrap up. Thinking Machines Lab, the AI startup launched in 2025 by former OpenAI chief technology officer Mira Murati, has announced that it has parted ways with its co founder and CTO Barrett Zof, who immediately rejoined OpenAI AI alongside fellow co founder Luke Metz and researcher Sam Schoenholtz. This transition has been marked by conflicting accounts of Zof's departure. According to an internal memo from OpenAI's Fiji Simo Zaf informed Marathi on Monday on this past Monday that he was considering leaving and was fired two days later. However, reports from the Wall Street Journal and Wired cite internal Thinking Machines Labs communications that allege Zaf was terminated for unethical conduct, including an undisclosed workplace relationship and the sharing of confidential company info with competitors. OpenAI has stated it does not share these concerns. So Zaf is back at OpenAI, reporting directly to Simo and leading a team that includes Metz and Schoenholtz. Thinking Machines Lab has appointed Sumit Chintala, the co creator of Pytorch, as its new cto. Now, Paul, it wouldn't be AI without a boatload of human drama here. So is Thinking Machines in trouble here given the type of exodus they're still seeing?
- [78:24]
A
Yeah, this got crazier. So they're in the midst of trying to raise it a $50 billion valuation and apparently during all hands where they're explaining all this, three more people quit on a Slack channel. Oh wow. Live while Mira was explaining what was happening. So they lost five total and four of them went back to Open AI. Just crazy. Like I like I said, you, you can't write this stuff in a Hollywood script so it doesn't look great. Like I, I don't, I would imagine if you're an investor in, in Thinking Machines, which didn't have a product or a plan when they raised their first couple billion, I, I, I don't know, I, I, I would imagine you're getting a little hesitant to write that next check at a $50 billion valuation when, when all this is going on. So yeah, lot of again, people don't like uncertainty, investors don't like uncertainty, and this seems ripe with uncertainty at the moment.
- [79:15]
B
All right, our last topic this week. A new AI video capability is going viral. So venture capitalist Justine Moore had reposted this demo of a tool called CLING AI and it has this motion control feature. And this demo was put together by an artist named Eder Xavier. And in the video we'll link to this so you can see it all on camera. Xavier is on camera in one window in the bottom right of the screen. He's doing things like waving, smiling, moving closer and further away from the camera, camera behind him in a larger window. The exact same movements are mimicked by a sequence of photorealistic video avatars. And in this case he chose to depict them as the actors from the mega hit show Stranger Things. So basically imagine when Xavier moves behind him, one of the actors who look a hundred percent real make the exact same movements in real time. Now Moore's post that shows off this demo is currently at tens of millions of views because it's showing this like jaw dropping realism and seamless real time capabilities in AI video. So this tool is currently available for public use through the CLING AI tool. It's also been integrated into the KRIA AI platform where the technology is listed as a motion transfer feature on the site's homepage. So Paul, I mean if you go look at this, this is just jaw dropping. I mean we know what AI video is capable of. But seeing demos like this really hammers home the fact that we are just not going to know what's real real anymore.
- [80:45]
A
I showed this to my kids like right away and I, we had a conversation about AI influencers and you know, things you see online and not believing it and having to go to the verified channel of anybody that you, you know, want to follow. So like, yeah, and I've mentioned this before, like I feel like I, I have an obligation to educate my kids because I, I know other parents aren't going to necessarily know this stuff and I feel like they need to educate their friends and like you gotta pass the knowledge on on. So yeah, this stuff is terrifying. There's so Justine, who's an investor, she shared a, a post that we'll put a link to and said how to guide for viral AI character swaps. And it actually plays out the entire recipe of how to do these things. Yeah, I, man, this, this goes so many different ways I'm not even going to bother getting into at the moment. But yeah, let's just, let's just say that like stuff you, you see online is, is you have to assume is probably not real when it comes to people. And this certainly starts to impact different businesses that are built around, let's just say adult industry type things. Yeah. So I, I don't know. Like this is stuff that just worries me. Like I, I don't see the good of this, honestly. Like I did. It's like the tech exists. It's like for what benefit other than like humor and, and like keeping people on social media sites? Like, what is the benefit of allowing this and like promoting the fact that you can do this and making it simple to do it. Like, I get that it's under the guise of, oh, it's just fun and it's like cool. It's like, yeah, but like 8 out of the 10 use cases for this are nefarious and like why are we not focusing on that? So yes, I, again, I would just stress that like we talked about with the GROK thing last week week, like education awareness you have, if you are in the know on this stuff, you have to educate other people about what is possible. This can lead to fraud and scams and emotional connection. It's just so many bad things can come from this and I, I don't like when AI influencers, like actual human influencers, talk about this only as the technology and don't address the fact that this is very, very bad probably for society.
- [83:05]
B
All right, Paul, another packed week in AI. We're starting off the year with a bang here, so appreciate you unpacking and demystifying everything for us. As a quick reminder to everyone, if you haven't left us a review on your podcast platform of choice, we read a lot of them and really appreciate all the feedback, good or bad. It helps us improve and also helps us kind of stay motivated to keep going. We've had some really nice comments from folks, so please leave us a review. Regardless, it helps us us get the show into more people's AirPods or headphones and helps us make it better. Also, don't forget to take the weekly AI pulse at SmartRx AI forward slash pulse and we will see you next week. Paul, appreciate everything you have broken down for us today.
- [83:47]
A
Thanks Mike and thanks everyone. Have a great week. We'll talk to you again next week. Oh, we do have an AI Answers episode this week, so we'll have a second episode this week from our intro class last week, so on Thursday we'll be dropping and episode 192 will be one of those AI answers special editions. Thanks for listening to the Artificial intelligence show. Visit SmarterX AI to continue on your AI learning journey and join more than 100,000 professionals and business leaders who have subscribed to our weekly newsletters, downloaded AI blueprints, attended virtual and in person events, taken online AI courses, and earned professional certificates from our AI Academy and engaged in the SmartRx Slack community. Until next time, stay curious and explore AI.