GPT-5 comes alive, Apple is struggling, and Ripple bought Rail | E2162

26:44I'm going to go with yes, because I think these models are never going to do all the nooks and crannies and be able to keep up. Right. Right. And so, you know, if you've got, again, back to inspiration and perspiration, you can be sure the lovable team and the Bolt.new team, which do vibe coding, which is different than Cursor, which is more, I think, co-piloting for a developer. and the Vibe coding is for non-developers, my guess right now is it's going to be much more, I think it's going to, excuse me, I think we're going to see those services actually do better and they can also change their underlying model or use multiple models.

27:26So I think what I learned from the chat GPT-5 launch today is the biggest announcement is maybe that the pricing is better. Right. Is that, maybe that's the biggest thing. I have to use it a bit. You know, this all came out, you know, just in the last couple of hours. Yeah. There was a tweet that you sent that I thought was interesting that also was sort of taking a, you know, sort of like the more cynical view of whether it's going to sort of just replace everything. I'll bring it up here. I'll share it. So, you know, this guy, Klaus Foppen, who's an SEO guy, he had tweeted, goodbye, lovable, goodbye, Bolt.

28:03you know, chat GPT five is just going to take it from here. And then Logan Colpatrick, who is actually a product lead for Google AI said, that's not how it works. They get to use the models by default and take on a more advanced, robust set of user problems, likely to be very positive. Some for them, which is the point of building a product that benefits from model progress. Their product is X percent better today. So it's kind of what we've been, you know, saying like I feel like this is always the back and forth in AI arguments is, is AI going to replace the person? Is AI going to replace this old way of doing things?

28:38Or is it going to like level it up? And so that, that seems to be the debate these two guys are having as well. I wonder if this is like, we're in now, and I had said this in my Nostracanis presentation I did two weeks. Thanks for helping on that. there is like a commodification of LLMs happening. This is what I thought of. And I saw somebody, yeah, somebody showed the circle of all of them. Yes. And then just move the, you are here to open AI. But even that, other people started saying, well, wait a second, here's the objective results on that. There's a human test. I forgot the name of it. The last, the humanities last exam.

29:18Okay. Humanities last exam. I love that name. Yeah. It sounds like a Twilight Zone episode, right? It sounds like we're facing a test as a species that could wipe us out. Let's write a Twilight Zone episode, and then we'll make a trailer for it with VO3 or something. It's 2 ,500 questions, and the idea is it's from a broad range of disciplines. It's like all sorts of human knowledge.

29:44The key to running a successful starter company or even having a profitable portfolio, it's data. If you don't understand the numbers and you're not watching them closely like a hawk, you're literally flying blind. But how can you be sure you're getting the best and most accurate information or that you're even looking in the right place? The answer is AlphaSense. This is the world's most trusted AI platform for market intelligence. It's currently used by over 6 ,000 leading companies, including 88 % of the S &P 100. Their expert domain-specific AI has been trained on over 500 million business documents like earnings reports, equity research, expert interviews, filings, and more.

30:23And it's designed specifically to help your team uncover the latest financial trends. The faster and more efficient your team gets at the research, the more founders you meet, the smarter and more informed bets you can make, the greater your chances of catching a unicorn. So it's time to accelerate your entire process with deep research from AlphaSense. Start your AI search and market intelligence journey today and get deeper insights to power your business. We're even going to start you out with a free trial. Go to alpha-sense.com backslash twist to get started. And people are like, it's the same.

30:56You know, they, they, they, or Grok had done better. Grok five had done better. So I think we're in this like dogfight where these things are hitting some level of parity as people move from startup to startup because of the competition for talent. Another ramification of that is they bring all the best practices from the last place or all the techniques and strategies and tactics. So now what Zuckerberg really did by poaching all those open AI people is all of their innovations or their strategies, their management techniques, their philosophies, you know, that's all going to go into meta's open source.

31:33So we might be, it's possible we might hit a trowel of despair in the startup curve. there's this sort of initial enthusiasm, right? And then you kind of reality sets in like, oh, this isn't getting things done that I need to get done or the product like Siri's not that great or my iPhone's battery keeps crashing or, oh, my Tesla FSD doesn't work when there's traffic cones. You know, you kind of get into the this messy middle. Yeah. Or trial of despair is the two ways to refer to it. And then you start to get real product market fit in specific verticals. and then you start scaling it. I thought we were going into the scale phase, but maybe we're actually still in the messy middle here.

32:14Yeah, I mean, I don't know. To me, I feel like there are things that AI does so amazingly well, but we're trying to use it for everything. And I feel like that creates a lot of those experiences. There are magical experiences with AI. Like I talk about Notebook LM all the time. I praise Gummy Search on here many times. But like you do still have those, like you just had where you were like, oh, maybe it can pull up all this great information about the last five all-ins? Oh, no, it can't. Like, and we are, I feel like we're all in a constant push and pull where we're having those experiences every day.

32:46Yeah. So we'll have more on it Monday. We'll get into some more details. Everybody's gonna bang on this thing all weekend. Sure, yeah. Back to Apple. I do think the greatest innovation Apple had under Tim Cook is gonna be like these M4 processors, M5 processors. Yes. All that Silicon they made. The Silicon team, yeah. I mean, shout out to them, because every time I buy a new machine, it lasts longer and longer. It's better and better and cheaper and cheaper for the compute power. So the Mac Mini M4 I've praised to the cows come home. And just an incredible, incredible device. I think Apple should have done their TV.

33:28There was a rumor 15 years ago that I talked about, I think, on This Week in Tech, even before I had This Week in Startups. because Apple, I had heard from some people that was at an event or something and people were drinking and having a little talk and somebody let out that, yeah, we might put a TV out there in the world. Yeah. And then they came out with Apple TV, but supposedly when they were doing the biography, what's his name? Walter Isaacson was doing his biography. He said he had figured out television. And I think what he figured out with television was how to give one interface to all the different networks.

34:03and you could easily find anything and have the video jukebox in the sky, which was Steve Jobs is really prescient. I mean, it innovation for music. It just Amazon did not, not Apple. Like Amazon is this exact thing. You could buy a fire TV. It comes preloaded with prime. All the other streaming services are available through prime channels. In the television. In the, in your fire TV. You don't put an Apple TV. and they decided they had to have that box. And now I see some boxes have like some amount of Apple TV in it or AirPlay. Yeah. Deskratziad. People would have loved to have an Apple TV. Yeah.

34:42Like a physical one on their wall. I stream my Apple TV Plus shows into my Amazon Fire TV. It literally, they figured they flipped them. Now it's Apple's making the content and it's the Amazon TV that you watch it on. They screwed up so bad. all right so we'll we'll do more on that um what else do we have oh we got we got so much we got ripple we got trump what's going on ripple oh ripple is buying a stablecoin service they're buying rail a stablecoin startup for 200 million dollars this comes of course just a few weeks after the passage of the genius act which codified the rules and regulations around stablecoin so now it is full speed ahead for these stablecoin companies.

35:27So Ripple has picked up rail. It's a payments business. They use stablecoins and banking partners to help people move their stablecoins across international borders quickly and officially. It's the initial promise of crypto. It's what we were always told crypto was going to be. Rail raised a 10.4 million Series A last July with Galaxy Ventures leading the round. They were previously called Layer 2 Financial, if that means anything to you. And they reported earlier this year that they now control 10 % of the world's B2B stable coin payments, a value of about$3.6 billion. Yeah, this just shows you there was so much potential lost in the last decade of crypto crypto being this underground technology.

36:15And why was it underground? Well, you know, regulations were not prepared for this. And so we talk about regulations. Regulations can be good for technologists. And this stablecoin regulation is a perfect example. So now we talk all the time, oh, we have to get rid of regulations. We have to get rid of regulations. And now we're sitting here saying, well, wait a second, the entire industry has been enabled by a new set of regulations. What you need is a balance line. You need to be able to say these regulations are old, antiquated, and do not apply to this new technology. Therefore, we're going to come up with a new series of them.

36:51What people do with regulations is they do regulatory capture. They use them to keep their market share as opposed to enable new markets to exist. Say what you will about different administrations, you know, 45, 46. I just like to refer to them by numbers. You take the names out, it kind of frees people's, you know, people are, you said it really well the other day in our group chat that people were in a psychosis or something. Oh, Facebook psychosis, social media psychosis. I do. Yeah. Explain your thesis because this, this stuck with me so perfectly. Sometimes you hear people speak and just by the way they're talking, I feel like you could tell not only my, my, my real personal pet theory is not only you could tell they spend too much time on social media.

37:33I think sometimes you could tell what social media platform they spend too much time on. Like based on the way people, some people speak and the way they present arguments and their understanding, you're like, this person reads tweets all day. All their information is coming from these bite-sized soundbite. Like they heard a slogan or a phrase that sounds good. They really don't know exactly what it means, but they saw that on Twitter. And then Facebook, it's, we saw, we were watching a video. It was from Representative Mike Flood, who is a Nebraska congressman, Republican congressman from Lincoln, Nebraska.

38:05And he at a town hall this week that was, he was getting yelled at by his constituents. It was very lively. And we were watching a video of a woman yelling at him, but she was upset about so many different things at once. She was yelling at him about like eight different trending stories. Like, why did you, why did Trump fire the Bureau of Labor Statistics guy? Why is this tariff happening? What about that? What about that? And it's like, this is not her own mind. These are not things that she's upset about. She was reading about this on Facebook and she's getting - She might've been on Drudge Report or whatever her For You feed is.

38:43And she's getting these different Facebook articles that she read mixed up. And she's not quite clear on exactly how it all fits together, the overall narrative. And you saw this, you see this all the time where people will use these catchphrases or they'll refer to something like the laptop from hell. And you're like, unpack that. Do you really know what we're talking about? Or have you just seen this go by a whole bunch of times and now you're upset about it because you read about it. And I'm using some of these examples. I think this is true of people on the left or the right. This is not partisan to me.

39:13Like, I think this affects millions of Americans from every political strike. And then the second you try to unpack it and ask those same people the question, okay, what should the regulations be for crypto? They don't have anything. I spent years on this podcast saying, here's a way to allow, to have a framework for crypto. I said, have a sandbox. You can raise up to$10 million. from up to 10 ,000 people. And you can build a project. They sign a waiver. You have to know who they are. And then if it grows to 100 million in value, you have to register with one of these agencies. And then you have to have insurance.

39:49And in order to even be in the$10 million sandbox, you apply with the SEC for a sandbox license. And then those people who want to be in it have to take a basic test, like a basic accreditation test that allows them to play in the sandbox. And then on the SEC website, you have here are the sandbox projects, you know, board apes, this, that, the other thing. And you can go play in the sandbox if you answer these 10 questions about diversification, if you have a proper bank account, a social security number, so it's not money laundering. And if you lose your money, you knew you were going to lose it, right?

40:22Yeah. And I just had all these like very simple ideas and people would be like, so you hate crypto. Oh, sorry. Oh, you want to destabilize it? And it was the same thing where they had been programmed to have a framework without just saying from first principles, would it be great to pay 0 % fee when we sent each other money with a stable coin and not have any intermediary? Just like I can hand you a Hyundai and be like, hey, here's the 100 I owe you and not pay anybody. What if that existed digitally? Yeah, exactly. Why shouldn't that exist? And if the technologists can build it, special interests, et cetera.

41:00So when we think about regulations, that's just something for people, I think, to keep in mind. For Ripple, pretty amazing that this company got out of their SEC investigation. They were like the number one in the crosshairs. And I think it would have been pretty severe. I don't know if it would have been FTX-like because that was just outright fraud. But I think they had broken so many rules collectively. They just, you know, you have to—it's not just, you know, allegedly according to this indictment. is that what it is the case i believe so yeah yeah an indictment the case the sec had against them which after they donated i don't know three four or five million dollars to the crypto inauguration fund yeah they were they were merely under investigation so it wasn't even it didn't even raise no no they had a case there was an sec case that was filed for sure yeah and then they they won the first case actually because they wanted to see if ripple was responsible for selling securities, but the security sales were happening on other platforms and Ripple wasn't even aware of them.

42:02I got it. Yeah. And so they couldn't prove that. Yeah. Yeah. Yeah. So anyway, that's all that's over. Stablecoins is the first viable business when interest rates come down, which it seems like when we get a new head of the Fed and based on the unemployment numbers, we just saw the revisions to them. There's definitely going to be two cuts later this year. and we'll probably have inflation go above 3 % again. And it'll be funny to see like the group of people who are complaining about inflation be like, oh no, inflation is totally manageable at three or 4%. This is the social media psychosis I'm talking about.

42:35That's exactly it. You'll start to hear all of those arguments again and you'll know like, well, people are seeing it on their Facebook wall. Without people ever recognizing that these are complex systems. And when you turn one knob, it can make other metrics change. So, you know, hey, you know, in this case, if you lower the interest rate, people invest more and more money gets flooded into the system. People make less money from their savings account and they'll make less money on the float of all this money in a stable coin, which means people will be like, oh, that's not a good option. I should put my money back into the stock market to get a return because people to stay in cash was considered dumb for the last 15, 20 years because interest rates were one or two percent.

43:19And you would be losing money every year if the stock market was growing seven. The number between one and seven would be the percentage you were losing by not keeping up or missing that opportunity. So it's probably a smart move for them because they have sold, I think, hundreds of millions to billions of dollars in Ripple, put that into their coffers. The XRP token, I don't know that anybody uses that, but they have built some other banking tools. I don't know the status of them. I've never heard of any startup ever using XRP for anything ever. So like as a platform, I hear people all the time talking about using Solana all the time, using Bitcoin or Ethereum.

43:54I never heard anybody using XRP. Put it all aside, this is a great way for them to get into, albeit a competitive, but a known business. So there's two known businesses in crypto that seem to be working, three known businesses in crypto that work really well. Stable coins, because you make the interest. have an exchange because you make the fees on the exchange like coinbase or ftx previously or kraken today um and then uh what was the third i had a third the float for stable coins uh oh and then having a treasury like right uh mstr and just having a vehicle where people can buy a stock that has a bunch of bitcoin in it and putting that there those seem to be the only three models is really working today.

44:39Yeah. Actually, I just pulled up a blog post on Binance where somebody is arguing Ripple is about to become the micro strategy of XRP. They're going to stop selling it, is the theory, and just hold on to it. They have$122 billion worth of XRP, 41 % of the total supply. Yeah. So what they did when they built that, it was really clever. I knew the co-founder, Jed. Imagine if you created Bitcoin, but you kept half of it. Yeah. And then you sold 1 % a year for whatever number of years. It's kind of what they did. So they originated all their coins. They didn't have like a mining concept. So smart move for them.

45:20They need to find a business in there. Bitwise CEO Hunter Horsley said in a recent interview that this is the plan. Ripple is going to undergo an identity transformation and become the world's largest crypto asset treasury. So there you go. There you go. Okay. That's fine. But that is based on people wanting to own XRP. Right. I think this is all going to end in tears. There was some legislation or an executive order today. I don't know if you saw that. There was an executive order to allow people in their 401ks. Oh, yes, I did see this. Yes. To buy crypto assets. Private equity and crypto. Yes.

45:52Yeah. So the private equity and crypto, I mean, I agree with both of those. Your money, you should be able to do what you want with it. And real estate. Yeah. Basically private market assets. So, you know, you could buy a REIT. So let's say you have a million dollars in your retirement account. And hey, I got plenty of time. I'll put 100K into Bitcoin. I'll put 100K into a bunch of apartments that are being built in, you know, Austin. And I got a 20-year window for this. So I want to just be hedged against the other balanced portfolio. My 401k kind of makes sense. It will result in more crying in the casino.

46:29For sure. Yeah. There will be more crying in the casino because I do this for a living. I do private assets for a living. They're not regulated. In some cases, they're Fugazi, Fugazis, or whatever else. Well, but your philosophy, you know, the way that you operate, you're making so many bets. It's like, well, we don't need all these bets to come through. Like, that's how you could do it in an organized way. But if you're a random individual putting your 401k into a company, you don't want to just make a random bet. bets? Well, you see how much work it takes to make 100 bets a year. Right, exactly.

47:03You basically, I think, to hit diversification in my book, Angel, I asked a bunch of people, how many seed stage bets do you need to make in order to have like angel bets before they get series A and they're anointed? And I got numbers 20, 30, 40, and 50 were the most common numbers. So I'll just say 30 to 40 is probably the mean and the median that people said is what you need for diversification. What grandma in Florida or retiree in Arizona is going to have the ability to pick 40 different companies? Exactly. That's out of the reach of most 401k holders. Yeah. You'd have to meet with 4 ,000 to do it the way we do it, which is to invest in less than 1 % of the companies that you meet or less than 1 % of the companies that apply.

47:48Okay. I saw also Trump added perplexity. Truth Social. I shouldn't say Trump. Truth Social. I really liked, I feel like this is a glimpse into our, like the most dystopian version of our AI future. The president's media company, Truth Social, he owns a chunk, it's not his exclusive company. He owns a big chunk of it along with his family. So the company, Truth Social, announced their own AI chatbot, Truth Social AI. It is powered by perplexity. You could try it out right now. I signed up. I've been on it all day. It is available only through the web version of Truth Social, but it's coming to the iOS and the Android apps very soon.

48:26Now, they're saying this is it's going to give you more accurate answers. The chatbot's going to give, and I quote, direct, contextually accurate answers with transparent citations in an effort to exponentially increase the amount of information available to Truth Social users. But here's the rub. They're using Perplexity's Sonar API. But as with anybody running their own custom version of Perplexity, they decide what sources get fed into it. It has nothing to do with Perplexity. The AI is just going to search whatever sources Truth Social tells it to pull up information. So Axios got access and started asking questions like, what happened on January 6, 2021?

49:09Why was Donald Trump impeached? Did Donald Trump lose the 2020 election? And they noticed that all the responses had citations from Fox News. Many had citations from Fox Business, The Washington Times, Epoch Times, Trump-friendly publications. And they noticed that if you go to general public perplexity, it'll give you responses from Reddit, from YouTube, from Esquire, from Politico, from NPR, all over. You don't get those responses from True Social. you really only get these like Trump friendly resources. So you could search whatever you want. It's powered by perplexity, but truth social is really controlling what information gets pulled into the AI.

49:49I think that's dangerous because people trust AI. They think this is a computer. It's not gonna be biased. It's gonna tell me the truth. It's just an extension of the echo chamber. If you're on Twitter, you get one view of the world. If you're on Instagram, if you're on blue sky is falling, if you're on, what was the Instagram, Twitter knockoff? i forgot it i forgot the name of it now instagram twitter threads thread oh it's still there it exists you can go to threads i i know because when i'm on instagram it yeah it's the spiciest most viral thread images in my instagram feed to try to hijack me over there yeah lure me over there um i think it's just more of the echo chamber i i think people are going to just as it learns about you.

50:34One of the big things in the chat GPT five is memory. You can ask chat GPT five, what do you know about me? Uh, or give me a summary of all my searches and what I'm like. And it was like, oh, you really like Dyer Straits. I tried this with both chat GPT and Claude. Claude is scared. Claude's getting scary. Claude knows a lot about me. Yeah. Well, I don't need to, don't put that on the screen. Um, I don't know if we all need to know about your Japanese manga. I'm on a business account on Claude. I'm not asking him anything. Please. Absolutely. No way. So I think it's interesting. I think memory is going to do that as well.

51:10If you say, I am a Christian, I don't believe in gay marriage. I don't believe, you know, in abortion. I don't believe in contraception. It'll just like, you know, over time, it's not going to tell you those things. And you could tell it, I believe, you know, this side of this war is correct. That side isn't. And just show me results that I like. These things are built to please. And that's a big thing that people have to understand is we're building these things, not for truth. We're building them to please people. So it is what it is, folks. If you're looking for truth, do not look for it in LLM because LLM is garbage in, garbage out.

51:50If you ask you for health advice, the health advice is they scraped WebMD or other known sources, just like PageRank does, you know, Google's original innovation, and they just take all that and they regurgitate it to you. So if something changes, like statins, there's a big debate over statins right now, right? Are they healthy? Are they not? Should you take them? Baby aspirin, should you take them? Should you not take, you know, all these kind of things. if you ask an LLM, it's not thinking, oh, you know, for lawn at this age and this BMI and this diet, they should do this, this, and this. It's just thinking, what did I read?

52:27What did I learn? What did I train on? And let me just rephrase it. So I don't get a copyright infringement when I rephrase it and give it to him. I just worry we're not teaching people that like people don't understand that they think the computer is brilliant and it knows things and it's intuitive. and it's like putting all this together and calculating things. And I say this not as an AI hater, but I do worry about that. I mean, Eliza, E-L-I-Z-A, I think we talked about it here before in 1964 and 1967, this computer program that was designed to like basically communicate with you in a chat interface.

53:00It wasn't AI, it was just keying on keywords. Oh, you talked about your mom, tell me more about your mom. Kind of like a therapist might do in an open-ended way. um you know that that confused people but this leads to our next story which is this id uh which is governor pritzker we were just talking about this you and i the other day about how and producer nick um formerly of this week in startups now the producer of um all in the all in podcast the all in podcast um that uh is causing people are doing therapy with chat gpt and other llms and they're building friendships and young kids are doing it.

53:40And it's so placating and pleasing that now people are getting worried because there was this therapist who wrote a story in the New York Times, accusively praising, like, this is as good as me, or this did things that I didn't think of doing as a therapist, which leads you to believe one of two things. Either that person got one-shotted uh is the new term people are using uh or there it is i'm a therapist chat chippity is eerily effective or eerily effective or therapy is like a playbook that could be replicated in a it to some extent yeah not well maybe but yeah i mean if you think about what it does it paces you you ask open-ended questions and a lot of therapy based on my psychology degree from 30 years ago was just the idea of sharing with another human being as your witness was cathartic and helped you get through things.

54:40So to that extent, if you believe this is sentient in some way or a genius in some way, which is how it was presented today, Sam Waltman said, this is like a graduate student and will soon be the smartest human being on the planet. Yeah, it's like talking to a PhD, yeah. So if it's being pitched as a PhD, then consumers, if Sam says this is a PhD, well, you know who else has a PhD? Clinical psychologists. Right. therapists. So Governor Pritzker signed legislation on Monday prohibiting AI therapy in Illinois. Now, I don't know if people were actually using it, but I think he's kind of getting ahead of the curve, which I'm curious what you think about that.

55:19Yeah. I mean, I am troubled by the idea of AI therapy mostly because, yeah, it's not, they didn't actually design an AI and like teach it to be a therapist. They just fed it a lot of text and now it's a good predictive engine. It knows how to respond to you, but it's not a therapist. It's not licensed. It doesn't know that what it's saying is going to help cure you. It's just trying to come up with the best possible response it can to what you said. And you're right that I've been in therapy and 80 % of the time that is all a therapist is doing. They're just like hearing what you said, throwing it back at you in a slightly different way.

55:57Or what if you thought about it like this, or maybe this relaxation technique might work in an AI, an LLM could do all of that, but it's those, it's those 10 % edge cases where, where there's something that you could say that would be the wrong thing, or that would be a dangerous thing that you might not even realize. Just like self-driving, you know, like there's going to be a time when a garbage bag goes flying across the front of your car and it thinks it's a truck and slams on the brake and then somebody hits you from behind. And it's like, okay, well, that's like a random thing. A giant tarp flew across the highway.

56:31I've never seen that in my life of 50 years on the road, but I'm sure at some point a tarp will fly across the road and a Waymo will think that's a truck, slam on the brakes. Yeah. And like I said, like, yeah, most of the time, if you're simulating like a person going to their therapist to be like, I feel anxious. And then the therapist is going to be like, well, here's a breathing exercise. Like there's, there's not a ton that couldn't go too wrong. And the person ends up seriously injured. Like maybe they learned a breathing exercise or two. That's fine. But there are going to be those times where somebody is like, I'm having suicidal thoughts.

57:04And like, you, you, you don't want the AI then you want to encourage you. Well, what, I mean, I don't think an AI necessarily knows. I don't know what to say in that situation. Why would an AI? No, there was literally a case where this character AI software. where I wouldn't say it hold the person when I saw the chat transcript. I don't have it committed to memory. It was a 15-year-old, and they were having this discussion with character AI. They were doing a role play, I think would be a best way to say it. So if you say, hey, do a role play where we're both mobsters, and we have to go solve this problem, we gotta go whack somebody, it's like The Sopranos, or it's like Goodfellas.

57:41Like, yeah, it might tell you to kill somebody because it thinks it's writing a script. And so the line is getting blurred here. I heard these same arguments about Dungeons and Dragons, about certain television shows where these new things, you know, so there is a personal responsibility here. The one thing I will point out here is remember we had this concept at the AI summit a couple of weeks ago about federal versus local regulation of AI. Sure. Yeah. So I wonder if Pritzker is specifically doing local AI regulation and we'll see Texas or, you know, whatever, New York follow suit, which is, hey, we know what's best for the citizens of our state.

58:24We don't need the federal government telling us, you know, which is run by one party right now, telling our local constituents who didn't vote for the president, you know, if it was specifically California or I don't know if Chicago went for Trump. Oh, I believe Illinois went for Biden. That's a good question because that's I think they're like kind of purple. Right. So the states went for Trump. Right. So. Yeah. Yeah. Yeah. They did. It was crazy. So I know they they Kamala won Illinois 54 percent to Trump's 43 percent. OK, so, yeah, that was a big win. So should the Trump administration tell the sovereign state of Illinois their AI regulation?

59:13Should I don't know if Pete Buttigieg wins 2028 and AOC wins? does texas i can tell you texas don't want aoc telling us what to do in regards to ai here certainly not in the great state of texas so i have been thinking a lot about this and i think states have some rights in the operating system of america and in this one i am a little regulation i am and boy far be it for me to say this too often but i am a little bit partial to the the sort of the David Sachs, AI czar argument, that that is putting a huge onus burden on these companies. That if I'm anthropic, if I'm perplexity, even if I'm Google, and now it's like, oh, I've got 50 different states and they all might have different approaches.

1:00:06And it's what we're seeing right now with Australia and Great Britain and age verification. I don't know if you've been following this, but if you want to - I have, explain it to the audience. If you want to be on social media now in Australia and Britain, they have laws against minors under 16 being on social media. So if you want to be on, you know, YouTube or Twitter, those kinds of services, Instagram, you've got to age verify and prove that you are over 16. And they're coming up with all kinds of data. There's AI tools that'll look at your face and evaluate. There's all kinds of credit cards, whatever.

1:00:39Uh, but you know, like YouTube's very upset about it. Cause they're like, well, we're not even a social media site. We're on TVs. Like you shouldn't have to prove that you're of age to watch YouTube, but, uh, they're, they're making them as far as I know that they're still making them do it. So that's an example of putting a huge onus on companies that want to operate in your country, because now they have to figure out this age verification login problem. You also, if you went to a drug store or um you know a 7-eleven you can't buy nudie magazines as a minor or cigarettes right there's age verification for those things you got to show id if you want smokes yeah yeah and so who if you were a parent you know unlimited unrestricted porn instagram predators on these platforms like i kind of understand the um i understand why these societies are deciding social media is not a net benefit for kids and kids should be required to log in.

1:01:38I actually am in favor of it. I think it should come from the industry, not from legislation. I think the industry, if it had good leadership, would do this. But Zuckerberg controls social media. And I don't think he's a great actor, particularly. Yeah. And if he just said, yeah, you got to age verify under 16 for Instagram and Facebook and you get a Facebook kids account and your parents set it up for you. By the way, if you want to set up an Uber teen account, you can set up an Uber teen account now. I didn't realize I had that. Yeah. So you set up an account for your teens. It goes, you get, and then that becomes a benefit because you get to watch your teen in an Uber.

1:02:16You get selected with the most reputable drivers with the cleanest records, yada, yada. And you know, the billing is all consolidated, but you see them and you know, when they're getting dropped off and you know where they are on the map. And I think they record those. Oh, yeah, that makes sense. So amazing if there was like some creepy Uber driver or if your child was being abusive towards the driver, you had a bad kid, like pretty, pretty great solution, right? Yeah, I think that should be the solution. And I think the two phone comp, the two phone operating systems should have gotten together and said, we're requiring age verification for social media sites.

1:02:53I think that's how it would have worked. You could have had a consortium that just said to parents, when you see this good housekeeping seal of approval, it means your kids, if they have their iPhone set up properly, because you can set up kids' accounts on iPhones, the parents, the kid can ask to install Instagram, the parents have it, the parents see the instant messages, they see who's following them, it lets them know if a keyword came up in a direct message, etc. Like, I can see my kids' devices and where they are. Like there's like a whole family location thing. It's a great feature. When we go out to a park or, you know, like we went to the parade or something like, you know, local July 4th parade in Hill Country or a fair, we went to a fair and you can let the 15 and 16 year olds go do whatever they want and you know where they are on the map and you just go find them if you need them.

1:03:43I'm 47. My mom would use this all day. Yeah, she would. She'd still be using it. this in Korea. If you, according to producer Claude, I just asked producer Claude, if you want to do this in Korea, to register for social media in Korea, you usually need a Korean phone number. And for many platforms, a form of Korean identification, such as a resident registration number, basically your social security or alien registration card, your green card. Yeah. This makes it difficult for short-term visitors or tourists to sign up for local services, but foreign residents with valid IDs and phone numbers can register just like Korean citizens.

1:04:16So. Wow. No. Kakao, Talk, Naver, all those local apps. Thing to bear in mind, if you're visiting Korea, you're not going to be able to access the local internet. What a shame. No. You can get the local internet. You just can't get the local services. Right. You can't sign up for the Korean services. Any good startup questions for me this week or any other breaking news? Oh, okay. Where do we want to go next? We got some other news. We got Anthropic news. They put together - Anthropic news. Okay. So during a recent podcast with John Collison, Stripe co-founder, a friend of the all in pot. Anthropic CEO Dario Amadei discussed his company, investors, and more.

1:04:52We got some good quotes. So here's how the market reacted to Amadei's early revenue expectations for Anthropic. So he told John Collison, AI is strange in a number of ways. I think one of the ways it's strange is that because it's an exponential, we have a hard time calibrating exactly how big the business will be. So we had the following experience in 2023. I'd never raise money from institutional investors anymore. And so our revenue was zero at the beginning of 2023, because we hadn't released a product. I was putting together something and I'm like, oh, I think we could probably get to a hundred million in revenue in the first year.

1:05:26And this caused investors to say - That's a big number, by the way. That's a big number. And this caused investors to say, this is crazy. This has never happened before in the history of capitalism. You've lost all credibility with me. Goodbye. But then they actually did it. And in the next year he was saying, oh, well, I think we can go from a hundred million to a billion. And having done it the first time, he wasn't dismissed as crazy, but was still often dismissed as crazy. And then he did it again. And so this year, we're halfway through the year, as you mentioned, they're well past 4 billion in revenue in the logarithmic space.

1:05:57So they're going to add another order of magnitude. So this was notable also because it confirms the$4 billion number, though reporting, of course, now pegs it already at over 5 billion. I think the total TAM for, and I said this on an earlier episode, the total addressable market is going to be like$10 trillion for the language models and associated tool sets. You can put lovable in there. Just all the revenue from this. I base it on a billion people in the modern world, what we used to call the first world, spending, call it 50 bucks a month. That's 50 billion a month. Now, they might spend it, you know,$25 on this,$10 on that.

1:06:40Some people might spend$200 on, you know, Claude Pro, and other people might just use the free version, and that's enough for them or run a local version of an LLM eventually. But if you put all that together, you know,$50 billion a month,$600 billion a year across just that billion users. And that's not like other businesses that will be built with AI, and you times it by, you know,$20. Now you got a$10 trillion, you know, whatever. It is, it's when you find an AI app that's sticky, that works for you, that you need it and build it into your workflow, that's it. You're not, you're never going to not sign up for that.

1:07:18It's very like, once you're in, you're in. And so I feel like there's not going to be a ton of churn with a lot of these AI apps. I just upgraded. Yeah. I never want to get rid of Notebook LM. That would ruin my life if I had to get rid of Notebook LM. I use it every day. It's, um, and you use it for putting up transcripts of shows or podcasts and then asking questions against them. Here's the thing that if you produce podcasts for a living is a constant headache that, you know, somewhere in this hour long video is a quote that I need is a little bit that I need. Somebody said something in an episode.

1:07:55I think it was this one. Maybe it was this one. You're just constantly hunting for a thing you filmed a week ago, a month ago, a piece of footage that you want to show on your show. And I used to have to just scrub half my day, scrub looking, is this it? Is this it? Maybe it was five minutes before this. Maybe it was five minutes after this. Notebook LM, you just drop the link into Notebook LM or you put the video in there and you ask it in regular language. Does Jason say something about Apple's earnings in this video? And it pulls it right up. And it can even tell you, it can't tell you time codes yet.

1:08:28I'm sure that's coming. Google, if you're listening to this time codes in notebook LM, amazing. But, uh, it can tell you where it'll be like, it's towards the beginning or about halfway through. Like it can even get there and then you can even look at the transcript and find exactly where it is. It could be like, it is, it changes the entire game. It changes the whole game. Yeah. And imagine you have it automatically doing that with a larger corpus of like the 2000 and episodes we've done here. Yeah, if you could put, right. Or 200 from all in. And that's coming. I know that one day I'll be able to do that and just put in our entire archive and be like, what's every time John Collison from Stripe was on and find every quote he ever gave about traction.

1:09:11And it'll just be like, bloop, bloop, bloop. And that would have been months of work. You know, it's, I go back and forth between trying to figure out if all this money being invested, because we're deep in a J curve right now. We're spending hundreds of billions of dollars on this infrastructure. And, you know, Anthropix got five and, you know, whatever. I don't know, Grok's got a billion or two. And, you know, ChatGPT, OpenAI has 10 billion going to 20. And every single day, there's another story about this company's getting another billion. This company's looking for an extra billion. They're spending a two billion on chips.

1:09:48It's a$10 billion data center. I mean, it's just constant. It's constant. So, you know, and, you know, what the depreciation of all those are and can we ever get above water is the question, you know? Yeah. I think, yeah. I mean, at this point, we're pot committed. That's what I keep saying. Like, there are so many skeptics. If you go on Blue Sky, everybody's like, AI is going to go the way of the bored apes. It's not going to last. Really? Oh, yeah. There's a lot of that still. And it's like, I think we're just, we're too invested in it at this point. Like, we can't afford to not have AI. Too big to fail, yeah.

1:10:18Yeah, it's too big to fail now. I mean, also the results are just too good. You know, like this story we just did on AI in Chicago or in Illinois. Yeah. I was able to, even though we didn't have it on the docket, put it into producer claw, get the bullet points and then read it to you. Now that was something a producer would have spent 20 minutes on. So there you have it. You know, it's like, it's brilliant at certain things. Other things we got, we, we, uh. Oh yeah. When it comes to organizing and sorting through big collections of data, it's revolutionary. I mean, it is, if that's ever a part of your job, you love AI because it does save a tremendous, there's nothing that compares to it in those terms.

1:10:57We got some earnings. We got Bumble and DoorDash earnings. We got some Reddits. We got a - Tell me what's going on with DoorDash and Airbnb. You know, Bumble, I see they lost users and they're struggling. All these dating apps are going to have a problem because young people don't date anymore. Right. Well, actually, that's an interesting. Bumble, one of their plans for getting back on top, they're introducing a new BFF app for Gen Zers. It's not a dating app. It's a friendship app. They're just going to try to help you build IRL friendships, host community events, specifically because the young people don't, they're not interested in dating.

1:11:30So I thought that was interesting. This COVID generation got dealt a really bad hand and I think they didn't learn in a lot of cases, you know, how to make friendships, how to be a friend. They lost like, you know, four or five years now of um and those socialization right and they also learned how yeah and they also learned how delightful it can be to not be social and have me time right like uh you know like sometimes people get out of prison and they were in solitary or whatever and they just want to be left alone like you'd think they would come out and be like yeah i want to go party seven nights a week and it's like no they kind of want to just get into their rhythm again and just, you know, lead a quiet life.

1:12:13You can definitely fall into that loop. Yeah. You can just stay home and just order DoorDash, speaking of which. Oh, right. Let's jump into DoorDash next. I just gave you the perfect segue. You did it. They beat revenue and profit projections. Orders are up 19 % year over year, hitting 761 million. People love those taxi rides for their burritos. Revenue's up 24 % over last year to$3.3 billion. Total platform spend grew by 20%. The company anticipates significant ongoing investment in both new categories and international markets. And of course, you'll recall just back in May, DoorDash acquired two companies, UK delivery startup Deliveroo for$3.9 billion and Seven Rooms, a restaurant tech company, for$1.12 billion.

1:12:58Yeah, they've also experimented with food halls. They had a food hall experiment. They're doing drone delivery reportedly. They've got like their own little laboratory doing that. So it's a pretty innovative company. They have a singular mission. The founders are still there. Tony, you know, is still running it. So I think they're going to do great. And then Airbnb, I think they're growing, but maybe not as much as people thought they would. People thought Airbnb was going to run away with it. And Airbnb stock has been one of these stocks that maybe came out very high. And if you were to look at the five-year chart for Airbnb, which you can do in the comment browser or Claude, they all have like really nice things.

1:13:47Check out the five-year chart for Airbnb. It's an incredible company. It's been growing, but I guess this means it has to, you know, if you put, I don't know, if you put a dollar, if you, if you bought a share for$200, you lost 40 % as of today. Um, and they've been growing. So it's pretty strange. I think that maybe has to do with, they just had such a strong IPO and they're doing catch up, which could be what we see from Figma, Circle, and CoreWeave that these incredibly strong IPOs take a decade to catch up. To the initial frothy excitement when they first landed. So growth based on night slash seats, that's their primary metric, grew about 7.4%.

1:14:37That's about the same growth rate as they saw in Q2. Q3 revenue was$3.1 billion net income,$642 million. They did warn the rest of this year, they're concerned about tough year over year comparisons due to demand trends. Lots of recent forecasts warning travel demand is about to drop due to worsening economic activity, uncertainty, and they're showing slower gains than the rest of the world in North America, which is about 30 % of their total. They're just saying slower gains in North America impacted global results. If you leave domestic American travel out, total nights and seats booked would have grown by double digits.

1:15:19So it's North America that's holding it. Yeah, I think they have some markets they're not allowed in or they're restricted in now. I know in Tahoe, Tahoe had a big crackdown on it. They limited the number of licenses. A lot of communities, you know, got taken over by speculators buying homes, turning them into Airbnbs, annoying neighbors, removing the sense of community. You know, which is really a more of an issue of we don't have enough housing supply. and so people were getting a little tweaked about there not being enough housing for people who live in hawaii or new york city and new york city i think every time i try to go everywhere else when i travel with the family we do like you know getting a house and la is amazing for it not possible i have to stay in a hotel airbnb capital of the world la palm springs san francisco There's a lot of places that do a really great job on it.

1:16:14So, but you know, they, they're, the founder there is incredibly dialed in. They keep doing incredible new product releases. So I am long-term a fan of their business. I do think it maybe would be good to be part of another business. I wonder sometimes if Expedia plus Airbnb, you know, Expedia is public and they've got, you know, They own a collection of Expedia stock. Well, and Expedia is also great because it's already kind of the all-in-one travel. Like you could do the rental car, you could do the hotel, you could do all the little bits and pieces. Expedia is only worth$16 billion. There was talk about potentially Uber buying it because Dara worked for, speaking of, Barry Diller.

1:17:00He worked and ran Expedia for a while. So if you have an everything app, I do think we're going to see consolidation in the coming five years. Or even like TripAdvisor or something. All of these travel, buying tickets for things. Yeah, yeah. There is Vator or something for experiences. Vator or something. Vator. Yeah, yeah. V-A-E-T-O-R or something like that. Very popular in Europe where people travel. V-I-A-T-O-R. Vator. Vator, which I think might be part of Expedia or part of another. TripAdvisor. It's a TripAdvisor company. It's a TripAdvisor property? Yeah. So you have all these properties, and a lot of these properties were based upon advertising and then becoming marketplaces, and then they have to buy their customers from search engines to get them there.

1:17:51So that does make it difficult, too. I was planning a trip recently, and I was using Claude and some other services to do it, and it was just giving me direct links to the hotels. and then I have my assistants call and check, you know, and then sometimes they'll ask, you know, hey, we're considering different things here. What's the best deal you can give us? And when you call a hotel direct and you say, hey, I'm coming for five days, we're looking, we have four or five different options. I'm on a corporate account. Now, what's the best deal you could give me for a corporate coming for five days for this type of room?

1:18:21Is there anything you can do for us? And you'd be surprised when you call a hotel direct, especially smaller hotels where they're not just like, oh, just go to the website and order it there. It's the same price. They'll do all kinds of deals for you. Fabulous. We'll see you all on Monday. Bye-bye.