Siri needs an App, OpenAI Acquires health startup Torch, Claude Cowork reactions | Diet TBPN

0:00Everyone wants to know did John get a haircut? Yes. Yeah, I did yesterday. Cleaned it up a little bit. It's getting wild. Some length still. We'll see. I think Siri needs an app. I think that that's under discussed. The news is that Apple and Google have a deal. We talked to Ben Thompson about it yesterday. Billion dollars going from Apple to Google. Now, my real hot take is that I think it's going to flip at some point. And I think Google or someone else is going to pay Apple to route LLM queries to them. because with the universal commerce protocol, agentic commerce protocol, with ads and LLM responses, there's going to be a whole bunch of moments where an LLM query is actually profitable.

0:43And on average, I think they will be profitable. I think the price of inference will continue to decline and the value of each query, the monetization of each query will increase until there's a flippening. and then all of a sudden every LLM query that's generated is on net across the entire category generating profit instead of generating losses, which is what's happening right now. Of course, just like with Google, when you search for, you know, how old is Leonardo DiCaprio or something, they're not making a lot of money on that. They're not running a lot of ads on that. But when you go to search for insurance or something like that, they charge a pretty penny for those ads.

1:20And I think the same thing will be true for LLMs broadly. It is kind of interesting they don't, I just looked it up, they don't run ads on how old is Leonardo DiCaprio. You'd think that they could figure out some type of ad to serve. Yeah, you'd think Brian Johnson would be buying an ad on that. Like some peptide. Because you're like, oh, he looks great. What's he doing? Oh, and then they look like Leo. He's 51. Leo was born in the 70s, which is crazy when you say it like that. Well, he's looking great. He was at the Golden Globes having fun, celebrating his movie. Goofing around. Yeah, yeah, he was goofing around.

1:50He was caught on like some mic looking at somebody. I don't know. I saw some random clips. The real thing I wanted to debate with Tyler was I was arguing that Siri needs an app. It's crazy. Siri came out in 2010, hasn't had an app for its entire life. I mean, it actually started as an app back in the day. Siri is from the Stanford Institute of Research and Intelligence or something like that. Yeah, what was it? It was like a$200 million acquisition? Yeah, yeah. So it was an app that you needed to install on your iPhone. There were a few of these speech recognition apps that you could go and you'd open up the app, click a button.

2:27Siri, because I think the fact that they were on the West Coast, was able to get the deepest integration in Apple. Sort of win the home button over time, win the Siri button, and eventually they teamed up and actually, oh, I have triggered Siri on my Mac. There you go. Eventually, Siri gets baked into the operating system level and the rest is history. The first couple of years are pretty good. People are excited. You know, you can, it was somewhat magical to be able to just press a button, ask for the weather, ask for a stock chart, ask for, you know, what's on your calendar, dictate a text message.

3:03A lot of that was pretty great. Estimates are around 500 million active Siri users globally, which is huge, but there's 1.5 billion iPhone users. So having only a third of your user base use your AI feature seems a little bit low, if that's the case. Obviously, juicing it up with an LLM like Gemini makes a ton of sense. But this deal between Google and Apple makes a ton of sense for a few reasons. Google's actually profitable, has the money. They don't need to take a huge amount of cash from Apple. They can give them sort of a deal on it because it's a billion dollars. Ben kept throwing around big numbers yesterday.

3:44Like, let's be honest, we're in tax. It's not a lot of money. It's actually true. It ramped up because at first it was the$30 ,000 DaVinci Resolve, the Blackmagic URSA immersive camera that's$30 ,000. He's like, that's nothing. You just put that over there. That's true for a tech company. And then he was like, Apple's paying Google a billion dollars. That's nothing. It's like they're both nothing. Yeah, the whole thing with the - Alphabet's a$4 trillion company now. The infrastructure, the hardware to do a live Apple Vision Pro broadcast being$30 ,000 or something in that range is like the fact that they could copy and paste that around every stadium and suddenly have like a probably a I imagine they could scale that to hundreds of millions of dollars of like you think so pay-per-view and stuff yeah yeah not not even pay-per-view but just a subscription yeah it's like you can just sell headsets I mean the headset's$3 ,500 yeah you only need to sell 10 headsets to offset the price of the of one immersive camera my take is that I think over time this will flip and I think over time Google will be paying Apple for all of the LLM routing that happens because Google will be monetizing those queries.

4:48You will go to Siri and you will say, what's the weather? And Gemini under the hood will tell you the weather. And they probably won't monetize that very well. But then every once in a while, you'll say, hey, Gemini, order me a new TV. And it'll say, what size do you want? And it'll just order it. Order me rain from Rainmaker. Yeah, change the weather. And they will get either an affiliate fee or a transaction fee or there will be an ad that's placed in the stream of content that comes back and I would imagine that right now Apple saying don't do any of those ads don't monetize these queries but over time I imagine that they will but the big thing that I wanted to discuss with Tyler he was fighting me on this he says that he says that Siri does not need an app I think that Siri will need an app I think that the the LLM chat interface is so dominant at this point that everyone has the experience of going back and forth asynchronously.

5:43Like it's chatting to a helpful assistant, like you're texting with a friend, you want to be able to scroll up and see. Yeah, if you think about it, it'd be very annoying if you had a real-life assistant and they were like, you can only call me. Well, Tyler, explain. Explain your position. So it's not that I... Oh, no. You flash banged him. I hit him with a flash bang. Sorry, he's trying to help you out, John. Opinion denied, Tyler. Opinion denied. We have a flash bang now on the stream. What good timing. So, Tyler, once the flash bang wears off, please tell us. Okay, so it's not that I disagree.

6:19I'm not, like, anti-app. It's just, like, the fifth most important thing that they need to, like, do. Okay. Like, I don't care. like okay would you rather have the same siri but there's also an app or a new siri there's no app yes yes yes it's a better model like obviously you're picking the latter yeah you're picking the better model but what i'm saying is that as soon as you integrate a better model that has more jordan's gonna flashbang again i'm too powerful i'm watching this uh as soon as you have deeper responses and more back and forth and more uh knowledge retrieval siri has they had wolf from Alpha integration.

6:54I think they had a Wikipedia integration for a little bit, but it was more just like pulling up a Wikipedia snippet. If you want to go deeper, you would open that web page in Safari and look at the actual Wikipedia. Now there is new context and new content that's being created on the fly by these models because you can ask questions that don't exist in a single Wikipedia page. Gemini under the hood in Siri will instantiate that for you, give you those paragraphs and you might not be able to read through all of that in one screen time session. You might get a text message, you need to answer it, you might get a phone call, you might have to put your phone down to keep doing whatever you're doing, and you want to come back to it.

7:30And so, yes, you could just say your prompt again and reinstantiate it all, but that takes time. These models are slow, especially if you're firing off a deep research report. I mean, like, yes. I agree that these are important things, but these are, like, far from the most important thing. I could just say, okay, if I'm asking about chips yesterday, we have this long conversation, and then I say, okay, yesterday we were talking about chips. Let's continue that conversation. That whole thing solves now. Yes. So that would be amazing. But the current chat apps don't even have that functionality.

8:02If you go to a new empty chat box. Yeah, but you could very easily implement that. Okay, on Claude, on Chat Between and Gemini, you can search through your past chats. Do you know how bad the search is? The search is not LLM-powered. It's not that good. It's not that good. It's not actually an LLM query. It's not just you're talking about something like you're talking with a friend. You have to use keywords again. You're back in like traditional search world. You don't just go to the same empty box and say, hey, remember when we were talking about the fall of the Roman Empire? Pull all that up.

8:36It doesn't do that. Okay, sure. But I don't think that's that hard to implement. You just put them in betting models. But we're talking about Siri here. It can barely pull up the weather. We're going from like a D tier product to now we're baking in a new model, Gemini, which is going to take it to like B tier. The implementation of the model is way more important than if the app works, if I can like search through it correctly. So, okay, if you're doing a deep research, are you even sure that that is, you want to be using Siri for that? Like maybe it's just better to do it. Whenever I do deep research, I never do it on the app.

9:04I'm always on like my laptop. Nerd, nerd alert. Because it's just like way more information dense. there's a bunch of links that I want to click on to do that. People definitely fire off deep research reports on their phone. Maybe they go read them later on their laptop. But we live in a mobile first world and Apple, and that's the Apple customer base. Part of what's cool about these apps is that if I pull out the sidebar, I can scroll through and I can remember, oh, wow, I was looking at, you know, the NBA's history and, you know, their attendance over different, I don't even remember firing this off.

9:38But like I did and now I can go back and enjoy the fruits of its labor. What would you rather have? Would you rather have Gemini 2.5 and an app or Gemini 3 and no app? It's not a trade-off. They're getting the best Gemini, obviously. And then also the app is coming. That's what I'm saying. Apple is bad at implementing AI. Yes. Do you agree with that? They have been to date. Yeah, no one debates that Apple intelligence was sort of botched. So I think there is definitely a case to be made that they need to prioritize certain things. and the implementation of the actual model, like them using the best model that's going to be fast.

10:10Maybe on Siri, people are just having fairly short responses. You don't want paragraphs and paragraphs unless you ask for it. Stuff like this is, I think, much more important than having a specific app. Yes, and I wouldn't be surprised. I mean, truthfully, developing the type of app that I'm discussing about Siri, it should be able to be one shot by cloud code in a day. It's super simple. OK, let me say one more thing. So I'm just saying like, OK. I'm about to flashbang you. OK, if you're writing this, who's the audience? like apple right you want to make a change yeah maybe yeah okay so it's like ben thompson comes on yesterday all i'm saying is that is that is that i'm excited for i'm excited to be able to access gemini with a button i feel like that's what this deal is giving me yeah that's nothing to do with an app also after i access gemini with this button that triggers trigger flashbang then i want to be able to see my list

11:10Pull up this clip from Gary Tan testifying in the Senate. Okay, okay, let's see. Who here actually uses Siri? I personally do not because I know that it does not have the cutting-edge technology that... I don't get along well with her. Yeah, exactly. That Anthropic or OpenAI or many other American labs could provide. and imagine if Apple opened it up so that similar to when you open Windows, you have to choose a browser. What if you could choose your AI agent? Then there are a billion consumers in the world who would suddenly have access to not just one self-preferenced Siri, but a variety of American labs.

11:53And that would open up investment. That would open up prosperity. That's a good pitch. I like this. When is that from? Is that recent? Ian is on an absolute tear. Let's see. Thank you, Ian. He says Apple's currently hiring 300 Siri-focused roles. Oh, they're going big. They're going to do everything. You don't think one of those is an app developer? I'm ready to move the goalposts. Let's do it. Ready to move the goalposts. We have the goalposts. So my new definition for Apple AGI, Apple intelligence, Apple super intelligence? Apple super intelligence. There's just a functioning search feature in iMessage.

12:31That would be very good. That's AGI for Apple. That's ASI. No, my definition is I want to be able to go to the new Siri. And so I got an iPad, and the iPad just accidentally installed, like, every app that I've ever installed on my phone on the iPad, which wouldn't be that bad because on my phone, I have one home screen, and then I have a second screen, and then the third screen is just the app library. and it just has a search box. And then the app library actually very intelligently organizes things into productivity, utilities, entertainment. It does all that for me. I don't need to organize it.

13:06I don't choose where things go. It knows if it's a creativity app or a travel app or a news app. It puts it in its correct category. I don't need to manage that. So it's great. It's a good test of agentic AI to be able to, with a single prompt, one shot, hey, clean up the whole desktop because we're seeing that today with Claude Cowork. As trivial as that example is, I think it's a good example of what an agentic system should be able to do if it has the proper hooks into the OS layer. Claude Co-Work can do it on desktop. What's the iOS equivalent of that? It's got to be Apple Intelligence. They have their walled garden.

13:40The walls are staying up. They're not letting Claude Co-Work go around your iOS installation and hook into all your different local APIs. That's the domain of Apple Intelligence. That's the domain of Siri. Siri is now powered by Gemini. Let's see if that's what they launch. Apple ran a blind test of frontier models and picked Gemini. Interesting. And there's this old photo of the Tim Cook meeting with Sundar Pichai from Google in a dimly lit cafe or restaurant. Tim Cook very clearly mewing here. I think they're both mewing at each other. They're both mewing at each other. They're just having a mew off.

14:17Yes, that's how they really decided. They're not even talking. They want to do business together. The New York Times has new profile. John Ternus, a low-profile but influential executive at Apple, could be next in line to replace the company's longtime chief executive, Tim Cook, if he steps aside. They say around 2018, Apple considered adding a tiny laser to its iPhones. The part would allow consumers to take better photos, more accurately map their surroundings, and use new augmented reality features. But it would also cost Apple about$40 per device, cutting into the company's profits. John Ternus, Apple's head of hardware engineering, suggested adding the component to only the more expensive Pro models.

14:52And we have to take a minute to talk about the nominative determinism of turn us. Turn us around. Mr. Turn Us Around. Our AI strategy is failing. We need somebody to turn us around. Mr. Turnus could be the man for the job. He joined Apple in 2001. He is now considered by some company insiders to be the front runner to replace Tim Cook. Apple last year began accelerating its planning for Mr. Cook's succession. Like Mr. Cook, Mr. Ternus is known for his attention to detail and his knowledge of Apple's vast supply network. Both men are also considered even-tempered collaborators. If Meta owns WhatsApp and includes that in the family of apps, and messaging is obviously a huge driver of social networking activity these days, people sharing content through there, Snapchat is a messenger, Instagram is, you know, you look and you see there's more shares on this reel than likes because people are sending it to each other.

15:48that's what people do. Yeah, and I think the people have long had the idea to build a family-focused social network, and I don't think they'll ever be successful because group chats on iMessage function do that very well. You just can't keep up with that. Apple's plans for artificial intelligence are also a big question. Is he HGI-pilled? Is he ASI-pilled? What are his timelines? Is he a doomer? We've got to get to the bottom of this. Hopefully the New York Times is asking the hard questions. We will find out. While other giant technology companies have spent tens of billions of dollars developing AI, Apple has largely been on the sidelines and it pushed off making major changes to its products with new AI technology.

16:28It'll be up to Apple's board of directors to decide who will eventually replace Mr. Cook, who also sits on the board. The rest of the company's eight board members did not respond to requests for comments and Apple declined to make Mr. Ternus available for an interview. They couldn't get the New York Times in the room with Is he a nice guy? Yes, says Mr. Rogers. He is a nice guy. He's someone you want to hang out with. Everyone loves him because he's great. And here's the quote that lives in infamy. Has he made any hard decisions? No. Are there hard problems he's solved in hardware? No. Cameron Rogers, hater of the year right here.

17:03We should have Cameron on the show. Hater in chief, chief hating officer. It's crazy. Certainly you cannot be at Apple for two decades and never have made a single hard decision or solved a hard problem in hardware. Like that seems impossible. It's funny. There's a hardware engineer at Apple today named Cameron Rogers, but it is different person. Different. Okay. Very fun. Well, good luck to him. I don't know. I think you can figure it out. I would be shocked if he's really never solved a hard problem. Seems like he has some beef with this Cameron fellow. OpenAI. They've acquired Torch. It's a healthcare startup that unifies lab results, medications, and visit recordings.

17:41Bringing this together with ChatGPT Health opens up a new way to understand and manage your health, says OpenAI. Okay, founders of Torch. Torch was a four-person company. Oh, small. The co-founder of Torch, Ilya, I believe one of the other founders, previously built Forward. Yeah. Ilya was also an early Uber GM. So trained at the Travis Kalanick School of Mogging. Healthcare startup founded in 2016, they wanted to rebuild the healthcare system from the ground up, sounds cool. The general belief was that healthcare was too reactive instead of proactive, right? You go to the doctor when you're sick, not to just kind of generally be healthy.

18:28They had an app, they wanted the sort of physical spaces to feel like Apple stores, right? They wanted this Apple-like experience for health. Subscription-only model. They didn't accept insurance initially. A bunch of folks, Founders Fund, Kostla, Eric Schmidt, a bunch of legends were in the round. 2018 to 2022, they expanded to dozens of locations. They had body scanners, real-time blood work, a bunch of different software that they had built out. 2021 reached a valuation of over$1 billion, raised a quarter billion-ish Series D round. At this point, from that point on, they sort of pivoted to focusing on this thing called CarePods.

19:15And these were like trying to be autonomous, like AI-driven medical kiosks. So honestly, probably right idea, maybe a little bit too early. I think the right way to read into this, OpenAI just launched OpenAI Health. They're now acquiring Torch. I would assume that they want all of your health data. They really want to help you with your health. They want your blood. They're out for blood. They want your blood. They want your blood work. Maybe they could say, like, hey, this month, instead of paying, you could be a blood boy for us. You give us blood. Our researchers are working really hard. You need your blood.

19:52I'm ready to upload. Upload me. Just let me know. Let me know. I'm going to read the message that Torch wrote. Torch says, everyone deserves the best answers modern medicine can give them, and yet we've all had people close to our hearts who didn't get the answers they needed about their health. We have more data about our bodies than ever before, but it's never been harder to bring it all together. Patients see only a fraction of their own records, while clinicians have too little time to parse the growing stream of data patients bringing in from wearables and consumer health companies, you know, function, etc.

20:21etc. AI is the most important new tool we've had in decades for turning the chaos into clarity, but AI can't help you if your health data is scattered across four hospitals, two labs, seven apps, and three web portals. We started Torch to build a medical memory for AI, unifying scattered records into a context engine that helps you see the full picture, connect the dots, and make sure nothing important gets lost in the noise again. Every time I see one of these new features drop from Claude, Co-Work or OpenAI, ChatGPT, Health. It just feels like the Siri thing is like bigger and bigger back to what Tyler was saying about implementation of the model because Apple has Apple Health.

21:02There's a whole promise there of what Apple Health as an app can do. You want that to be AI enabled. So they need to update that as well. There's so many different, okay, shopping, agentic commerce, like there's Apple Pay. Is Apple Pay integrated with Siri and the new AI function? So they really do need to go around and update all over the place if they want to stay. They have some dominant positions carved out, but every AI company is trying to eat off Apple's plate as much as they can. But there's a really, really strong lock-in to most of the Apple ecosystem. So it's not over if they don't move right now and get the app out and get updates out.

21:42But it's very clear that if Apple or ChatGPT Health is rolling out, it's going to take time for people to ramp up, start integrating things, start using it. But if that behavior develops and it's another two, three years until Apple responds, like, yeah, they are going to be a laggard in that category. Claude Code for the rest of your work, American middle class desk jockeys watching the asteroid hit the dinosaurs. We will see. We'll see how much this moves productivity, how much this moves GDP. That's the big question for this year is how much will people actually be able to use this to do the work they do every day.

22:21I've been in jobs where my job... Organizing desktop files was actually a$1 trillion industry. Yeah. And... I mean, I have been... I was an intern once, and my job was basically to open up an Excel template every day, copy-paste some numbers, sort of make sure all the formulas held. And over the internship, over the couple months, I wrote more and more visual basic so that my job went from eight hours on the first day to four hours to eventually it was like 15 minutes. We might see a bull market and water cooler talk. Ooh, yeah, that's possible. As people automate more and more of their work and just decide to yap.

23:03I'm AI pilled after seeing what the new grad philosophy major I hired to fly drones from a mountaintop has automated at Rainmaker with Claude. Yeah, we keep asking Tyler, hey, can you automate this? and he says, sorry, sir, it's not possible yet. It's not possible yet. I don't believe it. I never said that. I'm kidding. He basically said everything. Okay, Claude, take this unstructured data and turn it into dashboards that will give executives deeper insights into critical business functions. Make no mistakes. Cheer tickets is all reliable because Sophie posts some version of this. Wait, really?

23:41Once a day. Once a day? Basically. No way. He's gone viral with the same concept. Atlas says Claude here's a picture of my crush Claude here is her phone number make her my GF make no mistake That's also his been playing it Everyone has their role to play in the post singularity hangout session. Juwan says there's no point in learning or doing anything anymore Just learn English and acquire a good mental model of things and just type bro. Just type We'll see I mean you have to actually be inspired to come up with something that people want talk to users or something like that. We got to go over to the OpenAI device.

24:18There's new leaks, alleged leaks. We'll see. It's a new audio wearable meant to replace AirPods. This aligns with what the information has been leaking. The code name, Sweet Pea. Interesting. It looks like a metal egg stone with two little capsules behind the ear aiming for a two nanometer chip, maybe a custom chip for phone-like actions. Big ambition, 50 million units in year one. That's a lot of devices. Yeah, so Foxconn has been told to prepare for five total devices by Q4 of 2028. Whoa, this is interesting. All not known, but a home-style device and a pen are still considered. I wonder how serious going to Foxconn, if you're open AI, you go to Foxconn and you say, prepare for five devices.

25:04What does that really mean? Does it mean, okay, help me prototype, do some demo runs, build me one or two, and maybe we'll do one of the five, maybe we'll do two of the five. Is Foxconn really reorienting everything around this? Are they totally prepared to make five? How serious is that? I don't know. All I know is that it's exciting. I like hardware. It's fun. We're seeing it. We're seeing glimpse of this with the kids, with the board, and with the, what was it called? Camera box? No. uh sticker box yeah sticker box yeah i think i mean opening it sounds like uh johnny's team yeah are the ones that are kind of pushing uh for this form factor according to obviously this is a rumor but picking a form factor similar to airpods which have insane product market fit which last time i checked are like a 20 billion dollar revenue line for apple yeah very it could be a standalone business by itself.

26:01And again, if you look back to the infamous interview with Gerstner, Sam was saying, like, we have a device coming. Yeah, we have a device coming. Don't worry. Don't worry about it. That's a lot of revenue, potentially. I mean, I don't know if they sell them. What do you think the price point for a for a open AI device would be$300,$1 ,000,$999, $3 ,500? I feel like they have to stay in the range of the AirPods. And AirPods are between$100 and$300, basically. So low, single-digit hundreds. Well, Andrew Curran has a projection for what it'll look like. If they ship this, I think they've got an absolute blockbuster on their hands.