SPEAKER_00: Young people have been told their jobs are going away. It might not be that they're scared. They feel like they've been double-crossed. That Google and Eric Schmidt representing Google has bad intent. They know that the horse has left the barn, and they understand that there just aren't going to be the number of jobs for them.

SPEAKER_01: Suddenly they're trying to get an apartment, pay for their own healthcare.

SPEAKER_02: This really just highlights the gap between business excitement about AI and the average consumer.

SPEAKER_04: A lot of them did their degrees in the age of ChatGBT. These people have had ChatGBT for two years. They've been using it. They understand it really well, I think. Hey, everybody. May.

Chamath Palihapitiya: Oh my God, it's May 18th, 2026. Ski season long behind us. Summer. Long gone.

SPEAKER_06: Right around the corner. My Lord, I've got a lot of summer plans. I got to be around the globe doing all-in stuff in Paris and Foundry University in Japan.

Chamath Palihapitiya: But we're going to have a twist all-star summer. We're going to do our all-stars again. The top 10 guests of all time. We're going to try to get them to come back and then reveal them. So over like six weeks of the summer, we're going to have the top 10 twist all-stars. If you want to just tweet at us, DM us on all the social platforms, give us your best suggestions. But I know people are going to want like Chris Saka back or, you know, Brian Alvey, my friend. He always like interviews me. And that's kind of fun. Just so many great guests we've had over the years who have just been Rory Sutherland. Everybody thinks I vibe with Rory really well. So we're going to try to do those twist all-stars for the summer. So we have something to look forward to when we're at the beaches or whatever.

SPEAKER_11: You know, you get your Kindle, you read a book, whatever. You can also listen to a great pod.

SPEAKER_13: We love a summer. We love traveling.

SPEAKER_15: And we're going to make sure everyone that twist does not stop, no matter what goes on. Three shows a week, Monday, Wednesday, Friday, the train rolls on.

SPEAKER_04: I mean, this thing has been gone for 15 years. I'm like, somebody's like, are you ever going to stop? I'm like, I like it. So no, it's a lot of fun.

SPEAKER_22: So let's keep going. Yeah, let's keep going. First though, Jason, I think we should give a quick applaud, applause, applause, there you go.

SPEAKER_23: We get applaud, applaud. And yeah, I'm just going to press here, get my haptic.

Chamath Palihapitiya: I wear my plod on my wrist some days because I don't like the way it hangs on my t-shirt. I prefer it on my wrist. When I'm wearing a suit, I prefer it on my suit. Okay. Well, different strokes for different folks. This thing has been incredible for me, hiking and rocking. Yesterday, I did massive rocking. I dropped my daughter off at a game show kid's birthday party. They didn't want the parents in there. So I got two hours to walk around the Dominion, which is like this incredible super mall in North Texas. Just like everything's there. Patsy Grimaldi's Pizzeria from Brooklyn opened an outlet there. There's an Apple store, obviously, and everything in between. But I'm walking around. I put my notes on. I walk. That's when I get all my great ideas. So I'm rocking with a 20 pound pack on my back. And man, I'm just giving all my ideas to my plod.

SPEAKER_25: Sync it up when I get home. Look at my notes. And I am back in business. No ideas left behind. Is this the Dominion Ridge shopping center, Chisholm, that you're referring to?

SPEAKER_18: I might pronounce it, Chris. Is it the domain or Dominion? There's domain in Texas.

SPEAKER_29: In Austin, I mean. And then there's the Dominion in San Antonio. Okay. Okay. Yeah. It's big, y'all. Trust me. I looked it up. It's large.

SPEAKER_34: It's giant. Uh, and, um, it is like to call it a shopping center is a huge understatement. What it is, is a town.

SPEAKER_00: Like it's almost like they made a town and the town has some streets that are open to cars. Most streets are not. Yeah. And they put a bunch of housing there and then they put a bunch of retail there. It's nuts how much stuff there is in this sprawling complex. It's about 15 minutes north of like 6th street.

SPEAKER_02: If everybody knows going to 6th street. Uh, you need a plod note pen S. And if you want to save a couple of bucks,

SPEAKER_15: when you turn your note taking life on its head for the better, go to plod.ai slash twist. P-L-A-U-D dot A-I slash twist. Use the code twist. Save 10%. Don't lose track of what matters. All right. Uh, Jason, the biggest news I think over the weekend that everyone couldn't stop talking about was the popularity of bringing up artificial intelligence at university

SPEAKER_02: and college commencements. People just ate this up. They couldn't help themselves. Yeah.

SPEAKER_42: It turns out, uh, AI surprisingly unpopular at commencement addresses. We're in that commencement season. And of course you want your most successful alumni to give a commencement speech or your biggest donors. I don't know how it exactly works, but, uh, Eric Schmidt, famous former CEO of Google essentially went to university of Arizona, I guess.

Chamath Palihapitiya: And he gave a talk that I guess I should just play here or we should play here. And I'll just give some feedback on it.

SPEAKER_44: Here's two minutes and I'll, I'll, uh, talk over it, I guess, but it's pretty wild.

SPEAKER_45: And I'll say pause if I want to interject. All right, here we go.

SPEAKER_47: So today we stand on this edge of another technological transformation. One that will be larger, faster, and more consequential than what came before. It will touch every profession, every classroom, every hospital, every laboratory, every person, and every relationship you have. Relationship. I know what many of you are feeling about that. I can hear you. There is a fear, a fear.

SPEAKER_53: Oh, that's a serious.

SPEAKER_47: There is a fear in your generation. Oh, if the future has already been written, that the machines are coming, that the jobs are evaporating, that the climate is breaking, that politics is fractured, and that you are inheriting a mess that you did not create. And I understand that fear. It's rational. And it's amplified every day by social media platforms with algorithms that are earned with great Facebook, TikTok, Instagram, Twitter, and their anxiety drives engagement.

SPEAKER_57: Yeah, okay.

SPEAKER_47: But I want to say something to you this evening as clearly as I can. Okay. To speak of the future as though it has already been decided is to surrender the one thing that actually matters. You are surrendering your agency. The future does not simply arrive. It gets built in laboratories, in dormitories, in startups, in classrooms, in legislators. And the people building it will be you, and people like you. The question is not whether AI will shape the world.

SPEAKER_62: It will. The question is whether you will help shape artificial intelligence.

SPEAKER_63: We do not know. We do not know. Okay. So let's pause here.

SPEAKER_67: This is, you know, and there is like a six minute version. The whole talk has not been released.

SPEAKER_42: Young people have been told that their jobs are going away, and they are hearing this, and then they're talking to people who've graduated ahead of them, right? So they were freshmen at one point, now they're seniors graduating. They've got friends who went into the job market already, and they're probably going into the job market. They might have siblings, and certainly they have parents.

SPEAKER_73: And they're seeing that big tech is laying off a ton of jobs.

Chamath Palihapitiya: The leaders of big tech are talking about UBI, about no work, about half the jobs going away. When they hear Dario and Anthropic or, you know, Elon talk about, hey, it's going to be a world of abundance and working is going to be optional.

SPEAKER_42: And this kind of stuff, you know, I think resonates with young people because they're thinking, well, what am I inheriting here? Like, what is the job that I'm going to have?

Chamath Palihapitiya: And the problem with his speech is he was kind of talking down to them a bit. It felt a little bit condescending.

SPEAKER_00: He's like, hey, I know you're scared. It might not be that they're scared. It might be they feel like they've been double-crossed. It might feel to them like you have bad intent, that Google and Eric Schmidt representing Google has bad intent. So instead of him saying, hey, the AI industry needs to listen, and the AI industry has to be thoughtful about this, and let me tell you all the ways this is going to be great. And then he kind of lectures them, hey, and you know what? You guys can't pretend like you're not going to have a role in this. You're going to have a role in this. And they know that's kind of bullshit. They know that they don't have a role in it. They know that the horses left the barn.

SPEAKER_04: They see it because a lot of them did their degrees in the age of ChatGPT. These people have had ChatGPT for two years. They've been using it.

SPEAKER_00: Their students are like startups. They're looking for edge. They're looking for ways and tools to try. They've been using ChatGPT to get through school. They understand it really well, I think.

Chamath Palihapitiya: And they understand that there just aren't going to be the number of jobs for them. I know we're going to keep having this debate over and over again, and people are going to look at different numbers, but they're going to just have a harder time in the job force. That's my personal belief. That's their belief, clearly. Founders scale faster on deal. That's the deal.

SPEAKER_76: You can grow your company without borders, and you can set up payroll for any country in minutes. Hire anyone anywhere like a modern startup or large company does. And deal is going to get all the visas handled fast so you can get back to building. There's a great talent war that's going on right now. And you need people with superpowers for your startup to be competitive, to beat your competitors, to get your products to market. But anytime you try to grow your team with overseas hires, oh my lord, you've got to reinvent the wheel and you've got to navigate a tangled web of international laws, regulations, and you can't get these things wrong, folks. You want to onboard new staffers in other countries? You want to get them set up on your network, nice and secure, IT access, all that good stuff? You want to manage their benefits? Trust me. This is all a nightmare, unless you partner with Deal. They are the people stack for startups. They're going to take care of all the onboarding, payroll, HR, IT benefits, everything you need quickly, in one place, done perfectly. So visit deal.com slash twist. That's D-E-E-L dot com slash twist. And they think AI, I think, is also lame and inauthentic.

Chamath Palihapitiya: I think there's like a little bit of a cultural thing, but I don't know the nature of the school. I think there's also like the anti-billionaire thing. Like, why is a rich guy coming here telling me that like, my relationships are going to be mitigated by AI? That was another weird one. Like he's like, your relationships, like hospitals, everything's going to be AI.

SPEAKER_81: I don't think people want that future. Certainly these kids don't want that future.

SPEAKER_29: No, I think this really just highlights the gap between business excitement about AI and the average consumer, especially, I would say the recent college graduate, who is going into the workforce probably for the first time.

SPEAKER_15: Suddenly they're trying to get an apartment, pay for their own healthcare, and then they're coming to age in this kind of AI era. But Jason, this is not the only clip from this commencement cycle that shows a very public complaint about AI. So go ahead and give a quick listen to this. Okay, this is one I haven't seen. Yeah. University of Central Florida. And watch how surprised this woman is.

SPEAKER_85: Who is this woman?

SPEAKER_15: Gloria Caulfield. She's a real estate development exec. That was one of the speakers. Oh, it's right there on the screen. At the Tavistock Development Company, VP of Strategic Alliances. So a local business leader is how I think about her. Here we go.

SPEAKER_90: That said, we are living in a time of profound change. That's an understatement, right? Profound change. Change is exciting, very exciting. And let's face it, change can be daunting. The rise of artificial intelligence is the next industrial revolution. Oh, wow.

SPEAKER_95: She's shocked. They both were kind of taken back.

SPEAKER_98: Okay. I struck a chord.

SPEAKER_99: Yeah, you did. Yeah.

SPEAKER_103: She's a bit frazzled. There's one more round of booze.

SPEAKER_106: I was not a factor in our lives.

SPEAKER_108: Interesting.

SPEAKER_88: Okay. We've got a bipolar topic here, I see. Okay.

SPEAKER_90: And now, AI capabilities are in the palm of our hands.

SPEAKER_02: Wow. Can't believe it.

SPEAKER_15: She can't believe it. Also, I think she meant polarizing, not bipolar. But you know what? When you're extemporizing on the podium, I forget.

SPEAKER_34: Yeah. I mean, sure. Two different polars. I think we get the drift here. She's hit on something.

SPEAKER_42: I think both of these have hit on something. This generation maybe feels like this isn't in the best interest of humanity. By the way, it's kind of what we've all been saying in our own industry is like, is this good? Is screen time good? After what people saw from social media and screen time and the algorithm, which, by the way, the algorithm, which Eric Schmidt mentioned in his first one, that was kind of the first manifestation of machine learning that hit consumers in a major way that they became aware of as well.

Chamath Palihapitiya: Self-driving would be the other one.

SPEAKER_122: Yep. So when they think about AI, they think, oh, it knows what I'm thinking. It's listening to me. It's giving me targeted ads. It's sending me videos it thinks I'm going to be interested in.

SPEAKER_42: And then if I interact with them, it sends me more. And self-driving is like the other piece. And I think they're probably looking at it going, yeah, this tech stuff is toxic.

Chamath Palihapitiya: And it's not good for society. And we probably want to hang out with our friends and not have this in the palm of our hands. I think it's they're going to opt out of it. I think they might be looking at it going, if you don't want us as workers, we're going to

SPEAKER_01: figure something else out.

SPEAKER_89: It would be weird to have kind of a counter-cultural moment a la the hippies in the 60s, Jason, because of AI. But we could actually have one from that perspective.

SPEAKER_01: It feels like the anti-war movement. I think that's a very good observation.

Chamath Palihapitiya: It does feel like Vietnam or something. They're like, this is their Vietnam. Like, we don't want anything to do with this. You guys are shoving it down our throat. We do not want to be drafted into your AI army. I think that's it. Like, you're really interesting analogy between now and the 60s. I don't want to be drafted to go to Vietnam and go to this war halfway around the world. And I think these kids are like, I don't want to be drafted into some AI slavery where I'm using these tools. It's sucking my life out of me. And it's replicating me. And then I'm out of a job.

SPEAKER_132: Yeah. This is an interesting moment in time.

SPEAKER_15: I want to bring up a data point that I saw that I think is really salient to this conversation. So, the University of Utah's Kim C. Gardner Policy Institute. Put out a very interesting paper, Jason. And these are the critical charts. So, when the kids are talking about, you know, concern about the job market, I think this is a good data point. For those on the audio version, this is four charts. Essentially, we're looking at the construction workforce for all U.S. data centers through 2030, those in Utah, and then the U.S. data center operations workforce growth over time, and then the same for Utah because, again, it's from a Utah University. But what matters, Jason, is if you look at this, the construction jobs that data centers are supposed to create will dramatically decline after those projects are done, which kind of makes sense. But what about the jobs that will be created by them directly? Well, the answer is in the low automation case here, they total about 65,000 by 2030. Technology companies have already cut 100,000 jobs this year. So, all the data center jobs by 2030 are less than the layoffs we've seen in the first four and a half. And that, I think, just underscores why people are worried. And maybe rightly so, I don't think we can afford to not invest in the future.

SPEAKER_29: I'm still a free market capitalist guy, but I empathize with these poor students. I remember being an afraid college graduate, worrying about getting an apartment and buying forks.

SPEAKER_42: Yeah, and there's a socialist movement kind of moving, the democratic socialist movement. The government should take better care of its citizens. We should have universal health care. There should be more job loyalty.

Chamath Palihapitiya: All of this stuff is kind of coalescing in an anti-billionaire, anti-progress, pro-socialist kind of maelstrom. Yeah. And here we are, folks. I guess a commencement address is when the generations come together.

SPEAKER_137: And this generation apparently does not like to be talked down to and told how exciting this is, because they don't see it as exciting.

SPEAKER_13: No, no.

SPEAKER_29: Totally fine. There was a viral essay from the New York Times that came out over the weekend. It was called What AI Did to My College Class, written by a Stanford University senior called Theo Baker. Jason, I'm just going to read you a couple of quotes from this. Okay. And I think it explains why college students who are the most AI forward, I think, generation, correctly, as you said, are maybe a little bit skeptical.

SPEAKER_15: So, quoting from Theo here, AI is everything. We talk about it at the dining halls and in history classes, on dates and while smoking with friends, at the gym, and in communal dorm bathrooms. For some, AI has opened the door to staggering wealth. But for many who came to Stanford, when a degree seemed like a guaranteed ticket to a high-paying job, the door has been slammed shut. Going on, cheating has become omnipresent. I don't know a single person who hasn't used AI to get through some assignment in college. Moving on, in our tech-enabled newly AI-powered world, students were increasingly fudging just about everything. They would embezzle dorm funds to spend on their friends and lie about having COVID to get the Uber Eats credit the school offered to those in quarantine. Half of laptops in any lecture seem to be open to ChatGPT or Claude, et cetera, et cetera, et cetera. So, it seems that AI arrived all at once to universities and instantly dissolved the foundations of kind of a liberal arts higher education much faster than it crushed the workforce. So, these students and their cynicism are saying, we're up to our necks in this and we don't like it. And that makes me legitimately concerned about the

SPEAKER_139: feature. It's the worst nightmare of every founder. You've built a product, everything's working great. Then real users start flooding in and suddenly it all breaks. What a disaster. You need to get it back up and running and you got to do that fast. You're looking like an amateur. That's why you need a partner like Sentry. Applications can break in many different ways, but Sentry sees everything. You'll get all the relevant details like stack traces, commits, releases, and even the developers who push that problem code in the first place. With Sentry, you're not going to be jumping around between different tools trying to figure out what happened. And Sear, Sentry's AI debugging agent, uses all this data and context to identify the root cause of the problem and suggest a fix. It can even take a look at your code before it ships and warn you if any problems are likely. Try Sear and Sentry for free. If you're a This Week in Startups listener at Sentry.io slash twist, use the code twist for $240 in Sentry credits. Make sure you use that code, make sure you use that URL, Sentry.io slash twist, so they know your uncle J. Cal sent you.

SPEAKER_42: Will they get jobs? Yeah, I think they're probably most of them will get jobs. They're

Chamath Palihapitiya: super smart. There is a weird thing going on in the Bay Area where the people who did go work for the frontier models are like instantly being worth $10 million in liquid stock, that there's like a massive secondary. And so there's been a lot of hand-wringing about that.

SPEAKER_122: Basically, a have and have nots in Silicon Valley, where one group of people hit this generation's

SPEAKER_42: Google slash Uber slash, you know, pick the company that Meta, Facebook. And if you hit one of those,

Chamath Palihapitiya: you know, you made a couple of million bucks in your stock options. If you held them for a couple of years, you made maybe 10 million, 20 million bucks and you could buy a house. Well, now it's

SPEAKER_00: like really weird because the frontier models are in such competition for this specific type of person. The people at those are now they're the haves and they have a lot because these companies went to a trillion dollar valuation very quickly, which previously only happened with like a handful

Chamath Palihapitiya: of companies got to a trillion, right? We're talking about rarefied air there. So, yeah, there's a little bit of resentment between the 0.1% and the 0.5%. No, it's like literally like the top

SPEAKER_00: 10th of a percent versus the bottom 10th of that same first percent. Like they're all in the 1%, but there's some animosity there or some hand-wringing about it. And then there's also like, what is my, what is the point of all this? That's the P doom that people are feeling. What is the point of anything if the machines can do everything? And once you've experienced that, where you're like, I can do my homework with chat GPT and nobody cares. And my professors like reading the essays with chat GPT, this is all a farce. It's all performative, or I'm building something at my company. And there was an Atlassian employee who got laid off, who explained everything they built. And it's obviously super, uh, competent person. And he was kind of lamenting, like I did all this stuff.

SPEAKER_42: And then I basically trained AI to replace me. There's people who are well aware in our industry that the work they're doing replaces themselves. And so we're, we're getting to, yeah, there it is.

Chamath Palihapitiya: His former Aussie tech employee reveals what he built in viral video. It's a very interesting

SPEAKER_42: video. I watched like half of it. It's incredibly technical. He goes through step-by-step everything he built, and then he got laid off. And you're like, this guy's a genius. How did he get laid off?

SPEAKER_02: You fired that guy? Well, I mean, ultimately, I think now layoffs, Jason, have gone from a shameful act

SPEAKER_15: to actually one that signals to investors, two things. One, that you're finding, uh, efficiencies from AI automation already. And two, that you're conserving capital, AKA gunpowder, if you will, for the future to be more AI first. So I think investors dig it, but sucks for people who are suddenly cut loose.

SPEAKER_06: It's, it's brutal for them. This is a level of, uh, choppiness. I didn't exactly expect,

SPEAKER_01: uh, all at once. Pretty interesting. Can I actually underscore something you told me a few months ago?

SPEAKER_29: So we were talking about this issue because we've been talking about it now for 18 months or so. Sure. And you said that people should go out there and start a company. And I, I thought that

SPEAKER_15: was a little bit tricky because not everyone has resources, access, never, blah, blah, blah, blah. But I think the reason why I like that now is to me, it is the only way to escape the permanent underclass because if you are a worker for someone else, you're a cost center. But if you own your own company of any size, you're at least in control of your own destiny. So I don't know if that's the way to, you know, dramatic wealth as we consider it in technology circles. But I think if you own your own company in whatever it is you do at a minimum, you have some buffer or bulwark against these layoffs we're seeing. So I think I've come around to your thinking on that.

SPEAKER_06: Well, the writing was on the wall and I just extrapolated if this continues, if the layoffs

SPEAKER_42: continue and there's not a lot of hiring. So then what is the best thing to do now? There is this preconceived victimhood, like talking down to people like, oh, you know, not everybody can start a company. Not everybody has the motivation. Not everybody has the skill set. Like I've been doing this for a while. I've invested in six or 700 companies. Like I can tell you anybody can do it. Like anybody can do it. At this point in time, anybody can do it. Sure. Anybody can put up a shingle. Anybody can build a website. Anybody can start, you know, a sales process. All the answers are out there. ChatGPT and, you know, Claude and everything will do half the work for you. So if you are going

SPEAKER_122: to get laid off and you don't want to do this dance anyway, building a large number of small companies

Chamath Palihapitiya: companies and small companies means like making 500k to $5 million a year, not venture scale,

SPEAKER_00: but delightful scale for the people who own them. Okay, here we go. And in media, we've seen this happen. So if you look at the sub stack podcasting, everybody leaving big media, they've already experienced it. They experienced massive layoffs over the last 20 years since newspapers and magazines contracted BuzzFeed and Huffington Post and Vox, all the things that were supposed to replace them. They also imploded. They also either shut down or laid everybody off. Vice is the other one. You know,

SPEAKER_36: all those are gone. So what did they do? They went and they started independent publications,

SPEAKER_00: podcasts, consultancies to work. Those are the same people. Those are the same journalists in a lot of cases who were like, that's terrible advice. Jake, how, you know, not everybody can start a company. And I'm like, except you guys all just did that. Except you all just did that. And it's like, then people realize, oh, if necessity forces me to start it, then I will start it. And so that is when there's a fire, you know, and like, you're running out of money, you will get very creative. And the fear of starting a startup is a lot better than working at Starbucks as a professional,

SPEAKER_166: you know, like that would be the, uh, the, the disaster case for like somebody who is a writer

SPEAKER_29: at clock or advice. Why isn't someone right now talking about removing obstacles and barriers and friction to making small companies more viable? Like why haven't we talked more about healthcare

SPEAKER_15: portability? Why aren't we actually trying to support these people? Because I agree with 98% of what you just said, call it that. And I think that you're right. It is possible. But if you have kids and you're a one income family and you need to quit your job to go do a thing, that's a big barrier. I want to, I want to, I want to decimate it. I want to let people be free, but I feel like we're not addressing some of the structural things that would take your point and just turbocharge it

David Friedberg: and help a lot of people. If healthcare is the issue, just move to Europe and start a company there,

SPEAKER_173: you know, and then you get the healthcare and then you have to just pay more in taxes.

SPEAKER_42: And there's some regulations that it might take you six months to get your company fired up. So what you have to do is keep your company domiciled in America and then live in, and live

Chamath Palihapitiya: in Spain or Italy or France or Canada, somewhere where they have socialized medicine and the problem is solved. But of course people are also scared of moving, but you know, we've lived for many decades

SPEAKER_42: with a beautiful, safe, you know, employment environment that maybe we should have appreciated

SPEAKER_75: more when we were complaining about it. Uh, and now it's a bit chaotic and you're kind of, uh, on your

SPEAKER_177: own. You were on your own all that time. It just, um, you just had companies that would keep you employed

SPEAKER_76: for 20 or 30 years. Yeah.

SPEAKER_30: And Americans, when they came to America, guess what? They were on their own too. They had to go figure it out. So we just have like a hundred year delusion where it was just like really easy.

SPEAKER_00: Now it's going back to like it was for our forefathers and mothers who came here and it was

SPEAKER_173: hard. It's just going to be hard folks. And you're going to have to figure it out for yourself. Radical

SPEAKER_76: self-reliance. Self-reliance is what people need to learn in school. Self-reliance, like radical self-reliance. Like I will survive through sheer force of will. And you know, for some people it's

SPEAKER_180: like, I would rather survive through Mondami making the buses free and you know, whatever,

SPEAKER_183: you know, and then other people are like, I'll just move to Texas. I'll figure it out.

SPEAKER_31: The information reports that anthropic and open AI generate 89% of all AI startup revenue or

SPEAKER_15: basically revenue made by AI startups, Jason. And there's this particularly, uh, terrifying chart. I'm curious what you think about this and if this is bad news for startups, but as you can see here, going back to July of 2023, there's been a dramatic ramp in revenue from the major AI companies. And then if you see those little gray bars, that's pretty much everyone else

SPEAKER_65: in the AI startup game, kind of a shocking dataset. Sure. Uh, tokens are being sold

SPEAKER_67: at an alarming rate by two major companies. Cursor has a couple of billion in revenue, 2 billion, I think 3 billion cognition, and then 29 others are sharing like 7 billion.

SPEAKER_76: It looks like, uh, if this is correct. And this is a quarterly chart.

SPEAKER_187: Uh, I think this is a, as the data is released, I don't think we have exact month.

SPEAKER_40: Well, no, it's monthly. Yeah. It looks like it's monthly. Um, so the scale is monthly

SPEAKER_76: or is it one, two, three, four. No, it's quarterly chart. Um, are there indentations between the quarters? Yes. Okay. So it is a monthly chart. So it's a monthly chart based on estimates that the information pulled together from a total of

SPEAKER_04: 30 companies, 34, including open AI and anthropic. Yep. It's not your imagination,

SPEAKER_192: risk and regulation really are ramping up quickly. Your customers don't just want assurances from you. They expect proof of your security just to do business with you. That is why Vanta is a game changer for you and your startup for your business. Vanta's AI power platform automates your entire compliance process, whether you're preparing for a SOC 2 or you're running an enterprise GRC program, or you're doing an audit. Vanta is going to worry about the security. So your team can focus on building a great product, moving your business forward. Some of our favorite companies like ramp and writer, they're spending 82% less time on their audits right now because they work with Vanta. We've all heard the rumors about shady compliance and security companies. I don't need to get into the details. They were taking shortcuts apparently with their clients accounts. Allegedly, that's why you need a trusted partner like Vanta now more than ever. So get started today at Vanta.com slash twist. That's V-A-N-T-A.com slash twist. Okay. So there's a long tail of companies,

SPEAKER_00: obviously providing stuff. And, uh, this is only startups. This doesn't have the revenue from,

Chamath Palihapitiya: uh, Amazon web services, Google cloud and Azure, right? They're not included in here,

SPEAKER_29: not included in here, but it's worth noting that Anthropic doesn't take out the 20% or so it pays

SPEAKER_15: to certain cloud providers when reporting it's annualized run rate. So there is a little bit of that kind of hidden inside the data, Jason, but not enough to really change the overall picture.

SPEAKER_202: Yeah. So I guess the question is, is it a duopoly now? Uh, certainly is starting to look like one in

SPEAKER_67: terms of how people are using tokens. Yes, it would be, um, a duopoly and the companies that

Chamath Palihapitiya: the interesting thing is those other 30 companies, many of them are probably using, there's like a double counting here. So like, uh, some number of them are using Claude, our open AI as their backend. So maybe the revenue is getting double counted. There is the other possibility, but 80 billion in revenue is the total here or something.

SPEAKER_15: 80 billion in annualized revenue, about $6.6 billion per month. One more data point for people following along, uh, six months ago, the two major labs, Anthropic and open AI were actually four and a half percent less of the total startup AI revenue out there. So they're actually gaining share on these startups, which is a little staggering because we've seen companies like cursor grow quickly, cognition grow quickly, 11 labs is a rocket ship, Harvey, Legora, et cetera. Uh, and they're still losing ground. This actually has me worried, like legitimately concerned about startups

SPEAKER_42: in the app layer for now. It's really two different things. This is like combining the app layer and the infrastructure layer. It tokens feel more like infrastructure to me than the actual end product.

SPEAKER_06: So these companies are selling tokens, uh, at a massive loss right now. So the other way to look at

SPEAKER_67: this is, this is like Uber versus Lyft in those days where they were both losing a lot of money per ride. And then at some point they've got to, you know, make these profitable businesses, but they did kind

SPEAKER_06: of build a duopoly. And it looks like now then who's the Uber, who's the Lyft in this.

SPEAKER_15: That's interesting. So you think that their, their gross margins on inference today are negative?

SPEAKER_42: Of course. Yeah. With all the infrastructure they're building out all the,

SPEAKER_67: you know, I mean, the humans they have working for them, even if they're paid massively, uh, probably a small percentage of their costs. It's really the infrastructure. So they're, they're certainly losing money as they build up this infrastructure. And the infrastructure is questionable how long it has a practical use, you know, four years, five years, six years, core weave, Amazon, all we've, we've talked about this on the show. I've had this debate and discussion of like, what's the reasonable life. And then what do they do after their, you know,

SPEAKER_01: primary lifespan? Do they have a use or not? They probably do have some use. Okay. But the anthropic, uh, space XAI ideal, I think actually puts a good floor underneath

SPEAKER_15: the value of GPUs. Cause I think most of the processors in, uh, Colossus one from XAI are like H one hundreds, I think, which is now a pretty dated chip, frankly, but, uh, anthropic ones that they're going to power up the whole data center. I mean, they're two years old.

SPEAKER_214: So if they're two years old, uh, yeah, it's definitely less time that I thought, I guess,

SPEAKER_67: I guess they're, they're probably in year three of their own life, somewhere between year two and three, maybe they're at like 24 months, 30 months of lifespan. So they probably have half of it left

Chamath Palihapitiya: to go. Um, yeah, there is going to be a consolidation in this. And then this obviously doesn't show open source tokens or other, uh, services. So it's, it's, it's an interesting,

SPEAKER_67: um, it's an interesting mental model to start building is where is the revenue going. And right

Chamath Palihapitiya: now it's on tokens. People want to buy tokens. And that means there'll be many people selling tokens

SPEAKER_29: soon. You've talked a lot about how, you know, there's a risk to building on top of other people's models because you're at risk of training them with your own use about what they're. Um, so,

SPEAKER_15: so when I look at this charges and I wonder if these major AI labs now have access to so much information from startups that use them, that actually application layer startups today that are building on top of these models are just going to fall backwards down the slope because look at how fast coworkers advancing. Look at how fast codex is improving. They're building open glass into codex. I mean, it feels like if you're a startup, you're just hoping that your market is not big enough to warrant a new set of anthropic cowork plugins that could just rip the ground up from underneath you. And they basically bombed that startups seem to be at such a disadvantage to

SPEAKER_220: these companies. If you own, uh, the tokens and you have the foundational model building an

SPEAKER_67: application like Harvey into open AI or building codex into open AI, like, yeah, that is a significant risk, um, to those other products. You do need to actually think, are they going to in their search for revenue and profits going to go direct and they'll be open AI for lawyers, open AI for accountants, open AI for creatives, et cetera. But you know, this doesn't account, this doesn't account for Gemini's revenue because it's not a startup. So it's not, this is not a complete picture. This is defined as

SPEAKER_224: startups and it removes many of the players. What happened to Google's AI prowess? Do you remember

SPEAKER_15: the end of last year when, uh, uh, uh, Gemini three preview dropped and everyone was like, oh, they're on top. Google's run away with it back from the dead. And then they just didn't say

Chamath Palihapitiya: anything for like six months. Yeah. Um, people, well, they're, they've got, I think more users using AI than open AI, um, because of putting it at the top of search. So I think they have

SPEAKER_67: more technically users, not dedicated AI users, but this really is just about selling tokens

Chamath Palihapitiya: and you have two token providers and then you have applications all put into a bucket of startups. Startups design, I guess is divine defined by not yet public. So this is a look at

SPEAKER_67: private companies, not, and it excludes the public ones, which gives it, you know, a bit of, I'm sure if you put Google and Amazon and Azure's revenue from this space, it would look much different.

SPEAKER_15: Yeah, actually that's fun. I'll do that for Wednesday. Cause I think we got some better, more concrete AI revenue reporting from the, uh, the hyperscalers. So I'll see what I can dig up for

SPEAKER_42: us. Are they breaking that out? I wonder if they're breaking out like their, what percentage of,

Chamath Palihapitiya: you know, their revenue is tokens being purchased from them because they're, if they're selling H one hundreds for six bucks an hour, H two hundreds for six bucks an hour, that's different than providing

SPEAKER_42: tokens. So this is like apples and bananas and, uh, you know, a fruit bowl all being counted in the same kind of the, you know, bucket. It is interesting to look at though, for sure.

SPEAKER_15: Two companies are running away with the revenue. A lot of things like Google saying revenue from products built on gen AI models grew by 800%. That's not so helpful. Yeah. They're not going to break it out. All right. Now, Jason, um, I want to do a shout out to everyone who's done our bounties. We have two bounties running. We have the AI sidebar, which by the way, we just got a demo turned in that takes into account your, your changes, just getting the two people talking through, I played with it this morning. Incredible. Jason, I almost wanted to declare the winner.

SPEAKER_42: We wanted to do one thing just so we are clear, just do the fact checking. I think I made that decision on the last one because we had so many different personas. I can't even follow it when it's doing it. So we said, just do fact checking. I see here, the personas should have a troll and a comment around it. That's not actually accurate. We said last week, so Jacob, make sure this is

Chamath Palihapitiya: correct. And everybody understands that just do the real time fact checker, one simple solution, one simple part of this that we should, uh, then judge everybody by, and we're going to give everybody another week to just finalize what they're working on. But you actually used one.

SPEAKER_02: Yeah. And was it doing the fact checking in real time? It was doing the fact checking in real time. And it had these simply amazing customization features.

SPEAKER_15: You could say, only give me answers after 80 words. Only give me responses after 160 words. You could change the personality. You could change how they talk. And it's a website. You don't have to download it. I didn't have to download a GitHub repo, build a codex, install it on my damn Mac. Awesome. Friday, we're going to have another round of demos for that. And then also, Jason, tell people about the annotated.com bounty. This is the one that we're all very excited

SPEAKER_202: about internally. Sure. And we have this week in startups.com slash bounties, plural or bounty,

Chamath Palihapitiya: singular bounty, either one. Yeah. Or you can go to bounties. So I always tell everybody make both in case we make a mistake on air. So if you type in bounties or slash bounty, it should work. This is like a little tip from J Cal to employees who, um, don't think about the users. If you think about users, like they may hear bounty and type in bounties, or I might say bounties instead of saying about me, why not have both redirect to the right page? So we have a page here. We're going to keep doing these, I think on this week in startups, cause it's fun and it's interesting. And we get to meet people who are builders, uh, for annotated. I wanted to build a service that let you clip something on the web. And so at the webpage, if you click on it, um, you know, the concept here, whoops, sorry, I'm on page, um, is to create a side ball anywhere you are on the web, you highlight something. Let's say it's a YouTube video and it says, okay, what do you want to clip from this YouTube video? You pick this starting point, this ending point, you know, these 30 seconds of this podcast. It then makes a landing page with a piece of that media. It could be text from

SPEAKER_67: a New York times article. It could be the transcript in the video of Eric Schmidt. We talked about earlier

Chamath Palihapitiya: on the program, uh, when he was getting booed and then you could put comments under it. So you log in with your Twitter, your Google, whatever you have a sidebar, you can clip it. It makes a landing page. So annotated.com slash Alex slash Eric Schmidt commencement. And then if it always links to the

SPEAKER_122: original author's work. So it's not stealing it's fair use because you're doing commentary. So this is a

SPEAKER_240: fair use. This isn't like archive is or other places that like let you remove a paywall. The intent

SPEAKER_42: is not to be remove paywall.com, which, you know, listen, everybody uses a very popular. If you don't

Chamath Palihapitiya: have like some niche St. Louis Herald subscription and people want to jump the paywall, I think people are kind of cool with it. Um, but we don't want to be like wholesale stealing. What we want is you take

SPEAKER_122: a small piece and you annotate it and you put your opinion, you put your thoughts on it. Now, if Alex and I both put in the URL of the same, uh, New York Times story, now we can see this New York Times story was

SPEAKER_42: also annotated by these other users. And then you can see what they annotated, but nobody can take the whole story. You could take a paragraph. I take a paragraph that I could respond to yours. You could

SPEAKER_173: respond to mine and the original it's clear who the original author is, but you get to do some commentary

SPEAKER_00: on it. And this is came from my frustration with places that just steal content. And also very often

SPEAKER_36: I want to take something and make a new landing page and comment on it and then share it with somebody. Yeah.

SPEAKER_132: Kind of like a bookmark, but a little bit more with these modern tools. That's my goal. Now, if I go to, let's say that I have the, the, the Chrome extension sidebar installed for annotated.com

SPEAKER_29: and I go to a website that I've annotated and you've annotated. Do I automatically under your vision, see both annotations or do I have to go into the app itself to kind of see the collection of notes about that page?

SPEAKER_42: Great question. I mean, I, I don't have the answer. I do know when you're on that New York Times story, it should show in the annotated sidebar, like the number seven. And if you click that, then when you're w when you're walking around the web and you're on a YouTube page and it's like, here's some new audio track, uh, here's the new, you know, or here's a dire straight song that three people have commented on. It shows you the number three. So it's, it's basically a global annotation system for the web for podcasts, et cetera. So during a Joe Rogan podcast, you can put in the link to Spotify or the YouTube link or wherever it's hosted and you can see multiple comments on it, but it's not on Spotify. It's not on YouTube. It's on annotated, not the whole podcast, just the part

SPEAKER_122: you want to comment on. And then if that gets taken down, eventually this acts as a mini archive

SPEAKER_15: of that moment. Yes. Which by the way, really fricking matters. All of five 38.com just got taken offline. Talk about link rot. Just gone. Yeah. Just gone. That's so depressing. Now,

SPEAKER_254: while we're talking about bookmarks, Jason, can I show you something really cool, really quick?

SPEAKER_06: Of course. Okay. And I don't know if people are building this anyway. And then I reserve the right to split the prize. If I think there's like multiple good winners. Fair enough. It is

SPEAKER_15: literally the Jason bounty. So I believe you are in charge of it. Um, there's a company called, uh, think with mark has their website. Think with mark.com. And they made an AI bookmark about a year ago. It was like $129. And I forgot about it entirely after it came out there back with a new one. I want to show this to you because I think this is the first real AI gadget apart from plod that I

SPEAKER_259: really want in my life. Take a listen. Last year, our viral mark launch blew the internet away. A bookmark that kept track of what you're reading. But amidst all the noise, we wanted to create something that truly felt like it was built for you. It highlights passages, records your thoughts, and builds a personal library of everything you choose to keep. And with customization, it's a great way to show the world your taste. Highlighter is where magic happens. It has a scanner allowing you to save your favorite quote

SPEAKER_262: and a voice recorder for your favorite ideas, allowing you to engage with your readings without

SPEAKER_15: leaving the moment. Yes. So you can literally just take half this bookmark, take your book or your e-reader or whatever it is you're reading in a hundred languages, scan it. And then you don't forget the cool quotes. Literally. I just had to return my copy of Gulag Archipelago to the Athenaeum up the street. And that means that I lost all the sections that I had marked for later reference, but it was overdue. So I had to give it back. Yeah. I would have literally used this in my life like 20 minutes ago, $159. And it's an AI gadget that rocks. See, it's possible.

SPEAKER_34: I think there's like personal gadgets that are down the long tail. Like I wouldn't use this

SPEAKER_42: one myself. Sure. Why? Because I buy books. I don't ever want to rent books or use a library. I like to buy them and have them because they're so cheap anyway. And then I have a stack of them. And, um, I use my Kindle. My Kindle has all those type of features in it. You can take notes in your

Chamath Palihapitiya: Kindle, you know, and the same thing with audible. So since my primary place is those two places, this doesn't super appeal to me, but I do think for people who are readers, they're going to love this just like plod, you know, and these note takers, they have unlocked, you know, a very specific use case for a very specific group of people. If you're a leader, if you're a writer, if you're a

SPEAKER_00: researcher, you need these kinds of tools. And I could see this one, um, getting better and better. Now there is this, uh, axiom. The best camera is the one you have with you. Uh, and so the question

Chamath Palihapitiya: is always, you know, what do you have with you at the time? And if you have your phone, what would be nice with this device is if you forget your device or you're doing something else, you can take a picture with your phone of that page. And it also allows input from that. So I'm always thinking of like multiple ways to get input into it. But what would be great about this is putting it in your brain eventually and being able to have your agent have all this knowledge. And that's really something

SPEAKER_00: I've been thinking about. There's something called obsidian that people are obsessed with, which is some sort of bookmarking knowledge tool for nerds. Uh, I've been thinking about this with the domain name, begin.com where I just an annotated is kind of like in my same thinking, which is, I just want to have all my bookmarks in one place, but I use the bookmarking system in Tik TOK and I use the bookmarking system in Instagram and the bookmarking system in Twitter

Chamath Palihapitiya: with folders. I want all of those to be collected into one location. And then my agent to have that knowledge when I'm like, you know, I really, I was, you know, in this book, I remember talking, they were talking about this specific, uh, designer and I also bookmarked that designer and their, you know, perfect t-shirt. Okay. Tell me about that. And it knows it's like, oh yeah, well, you read this book and you heard about Calvin Klein and Calvin Klein made this t-shirt and then this

SPEAKER_36: Japanese company was the origin. Boom. It all comes together in one place. I think that's what we're all trying to use AI for is to build like a little backup brain before Neuralink kind of does this

SPEAKER_13: for us. I, I agree. I kind of a mind palace of our own, uh, before we move on to flock safety

SPEAKER_15: and preventing tragedy, I just want to say libraries do have a, um, have a use case that we didn't mention. And I, I agree with you. I mean, look behind me, this is a fraction of my family's library. We're huge fans of books. We buy more than we can read, but the best part about a library is you're browsing for free. Like you just walk around. I found like six or seven books about Soviet economics and I just took them all home. I read like 10 pages of some of them and I read the whole book,

SPEAKER_22: but I just don't care because it's free. Yeah. Anyway, it's wild. I, I, what did you learn about Soviet economics? It's fucked. Did you know that central, that centralizing authority in one country as part of a larger union doesn't work? I mean, for like 10 billion different reasons. If you want to get really, really opposed to, to centralized planning, just go study the Soviet economy. It'll, it'll radicalize it in a funny way. Uh, all right,

SPEAKER_271: Jason Garrett Langley, uh, the founder and CEO of flock. Uh, he's been on the program. He's been

SPEAKER_15: on the program several times. Uh, you talked to him episode one, two, four, nine, July 21. I talked to episode two, one, three, four in June of 25. So last year now what's going on? Well, this is the thing that you flagged and it's him discussing how his group flock security to make cameras and such, um, helped solve a crime while looking at data from a town that wasn't where the

SPEAKER_173: crime was committed. So why does this capture your eye? So there was a shooting, um, in Austin

SPEAKER_76: yesterday when I was doing my rocking and I was at the domain, I got like these alerts. My daughter

SPEAKER_276: has a phone now. It came up on her phone and it was like, Hey, stay out of this area. There's police

SPEAKER_42: activity that gave the all clear or whatever. And apparently there was a shooting. Now block is very controversial because privacy. And do you want your license plates and where you're going in a database? No, is everybody's answer. Like just, of course I don't. But then if there is a shooting or a robbery in your neighborhood, do you want the criminal's license plate, uh, recorded so you can capture them and get your money back or arrest them before they harm anybody or rob another house? Of course you want that. So most people have a hard time deciding my personal privacy and where

SPEAKER_279: I was and where my license plate got picked up or the edge case, which is a certainty in the United

SPEAKER_06: States. You're going to have a crime at some point, unless, I don't know, maybe you live way out in Montana, but if you live in a city, there's crime happening all day long. So if you assume there's,

Chamath Palihapitiya: you know, tens of thousands of crimes in a city, there's hundreds a day. You then have to make a very simple equation. Am I willing to give up my license plate data

SPEAKER_122: and somebody could abuse that and then publish it or it could get hacked in every single place I've been

SPEAKER_00: every single location. Let's say you were at a casino. Let's say you were at a, an adult entertainment venue. Let's say you were at, I don't know, a place you're not supposed to be in the middle of the day. I don't, you know, pick your vice that somebody was involved in. I don't know, somebody was scoring drugs in a, in a drug alley. Um, or you just were out late at night. Maybe you don't want people to know your, do you, are you willing to give up that privacy and risk that abuse for catching criminals more quickly? Now, how does this relate to Austin? Austin is incredibly democratic and woke in the middle of Texas. So we do have some of the same issues that San Francisco has

Chamath Palihapitiya: like a prosecutor who won't prosecute, you know, crimes as, uh, maybe it, you would get in Dallas or Houston or San Antonio. Sure. Put it all aside. Austin city council got rid of flock

SPEAKER_42: for privacy reasons. Yep. Intellectual people, whatever the shooting happened. They couldn't find the guys driving around who were involved in a multi shooting, uh, spree. I don't

SPEAKER_76: know what the motivation for this spree was, but they went to multiple locations and discharged a

Chamath Palihapitiya: firearm multiple times. They then went to, was it Lockhart County? They went to a county outside of Austin, got picked up on flock immediately because somebody had the license plate number and they were captured immediately. So here yesterday, 12 shootings reported across Austin led to a manhunt that involved 200 officers, including SWAT air and canine support over several hours. The suspects were found and arrested as they entered flock supported manner. I'm sorry. It wasn't Lockhart. It was manner. M A N O R. Um, if flock safety wasn't able to aid manner PD and Austin PD, in this case, how many more could have been harmed? So that's basically

SPEAKER_75: the message from Garrett Langley. I thought I would bring it up because I am conflicted on this issue

SPEAKER_29: myself. I am too. I just thought I'd go ahead and show some people what this looks like in practice. So there's a, um, how to phrase this, a community activist group of people who are finding these

SPEAKER_15: cameras and putting them on a map. So here is a data set from, uh, D flock, um, showing my town of Providence. And these are all the cameras that have been reported. Now, these are not all flock. There's 177 cameras in this shot of my local area, 155 of them are from flock. So mostly when we're talking about license plate readers and such, we're talking about flock. Their argument to Jason, as I had this talk with the CEO about it is if you're on a public road, you have no expectation of privacy. And that to me is actually a pretty good argument. The problem is I also share your unease

SPEAKER_29: about this. I don't really want to be surveilled and I don't mean to judge up old, you know, 1776 quotes, but the idea that those who would give up, uh, Liberty for security deserve, neither, uh, does come to mind. Why, why did this happen without a national conversation is what I'm kind of stuck on

SPEAKER_298: a bit like AI earlier. Yeah. Technology moves quick and everybody had security cameras. So then if you

Chamath Palihapitiya: have a security camera and AI comes out, um, AI, all of a sudden is like, I can tell you what type of

SPEAKER_34: car that is, what model it is, what color it is, what the license plate is. And I can tell you who's

Chamath Palihapitiya: driving it. And literally my security systems at my residences. Sure. Have a, uh, plate reader, face reader, car reader, animal reader. It reads everything,

SPEAKER_42: puts it all into a database. And I know when a license plate comes onto the ranch,

Chamath Palihapitiya: if that license plate or not is, um, in the database and approved. So I have approved license

SPEAKER_122: plates and then you can take these things a step further. Now the gate can open or close based on license plates. You can open and close doors and gates based on facial recognition. Now on a private

SPEAKER_276: property, you can do whatever the heck you want. I was about to public property. Okay. Um, you can

SPEAKER_00: also do this. So this is a technology that didn't exist when the forefathers thought of all this stuff. So what is the concern? The concern is the police state abuse, the Stasi nightmare scenario. How do you

Chamath Palihapitiya: solve for that? Well, with new technologies, it's a very easy solution. You track every single request that gets made and authorized and you audit it. And then the data is kept for six months or a year.

SPEAKER_42: And then it is destroyed and destroying data pretty hard these days because everything's backed up and, uh, in cloud. So even when people think they've deleted something, is it actually deleted?

Chamath Palihapitiya: We saw that with Savannah Guthrie's mother being kidnapped. They didn't have a subscription to the Nest camera and Google was able to recover the clip of it. So you're like, wait a second. It's like,

SPEAKER_42: okay, Google is running the paid version on everybody's camera, even if they don't have it, or maybe they're running the last week. Yeah. Recover is doing a lot of work in that,

SPEAKER_15: in that room. I saw the same word that you did. Um, I will say that there does seem to be a public

SPEAKER_29: backlash to this stuff. Do you recall the Superbowl advertisement from Ring that was going to call the cops if dogs were out and about in the neighborhood? Yes. My friend Jamie Siminoff's company, yeah. There was enough blowback to that. Ring is actually, Amazon, the parent company, has scrapped that police partnership because people were so mad. Their pitch for public safety became a privacy nightmare. So there is still a limit about what people will take and won't take. And I, you know, we were talking about states rights a week or two ago, and this is going to be one of those issues, but I think it's reasonable for people to be concerned about this. I think that there's

SPEAKER_15: been a lot of people that are like, oh, you just want murders. Then you're a murderer.

SPEAKER_75: Calm down, everybody. It is. I have a, I have, I have a, another concept here. So number one is

Chamath Palihapitiya: there has to be an audit trail. There has to be a cost paid. If you abuse it, just like Facebook had to deal with, and Google had to deal with people spying on their exes, right? Yep. A pretty well-documented cases of that they put audit trail. So an audit trail basically means if somebody looks at something they're not supposed to, they get fired, they get sanctioned, et cetera.

SPEAKER_42: And then retention is the other piece. So there has to be a retention policy. Then there's another idea. If crime is above a certain level, these things are deployed. If crime is below a certain level, maybe it's not as necessary. So, you know, if you lived in the tenderloin and you went to anybody who lived in the tenderloin. A neighborhood in San Francisco that's famously rough and tumble. I mean, as rough and tumble as this kid from Brooklyn in the seventies and eighties

Chamath Palihapitiya: would tell you. It's bad. Like it was scarier to me than the scariest places in New York I'd been.

SPEAKER_42: Because there were hundreds of people out at night high as F and chaos and no police presence. And just, you would literally feel like you were in like literally hell, like anything could go wrong. If you told a hundred percent of the people who are not addicts in that group and not homeless, we

Chamath Palihapitiya: want to put up these cameras to like, you know, dissuade this stuff and have facial recognition, they

SPEAKER_173: would say, thank you so much. I appreciate you so much. So I think people's belief in these systems

SPEAKER_36: versus giving up privacy depends on the chances of a crime happening to them.

SPEAKER_15: Yeah, I think that's right. Because people, the people's preferences will shift dramatically based

SPEAKER_29: on what they're seeing and what they need. Now, when you think about where it's being deployed, though, Jason, I'm going to show this map again and show you, Texas. This is a D flock. This is the D flock map, which we're using because I don't think there's as public a one from flock itself. This is roughly Texas. You can see Dallas. And I think this is Houston down here,

SPEAKER_15: tons and tons of cameras. I'm a little surprised that in Texas, the, you know, the Lone Star Freedom State, there's this much surveillance and it's acceptable by people that are there. I would have thought there'd be more of a pushback. So I wonder if I'm actually over indexing on people's aversion to these cameras, given that they seem to be all the rage in a place that's famous for independence and straight shooting.

SPEAKER_06: Here's the thing. These are deployed locally. So this is bottoms up, not top down. Your local

SPEAKER_42: community makes the decision and pays for these. It's not like the state of Texas deployed 21,000 flock cameras, if that's the right number you have, that means manner and lock heart and tripping springs and Westlake. And, you know, if it was the tenderloin, whatever, each of those communities

SPEAKER_122: has to opt into it. So that actually, to me means people are, have agency and are making their own decisions in my community. I want to know what's going on and I don't mind security cameras. Everybody

SPEAKER_319: accepts. So then is it AI enabled security cameras becomes the issue. Oh, and the database.

SPEAKER_29: I think the problem is that it's not the Stasi. It is a third party that is, that is feeling the

SPEAKER_15: Stasi. So there's two potential boogeymen here for people to worry about. It's not just the central

SPEAKER_321: government. Then flock shouldn't have access to the data. So that that's an easy solution. I don't

SPEAKER_42: remember from my conversation with him, but he told me everybody sets their local standards. So

Chamath Palihapitiya: the retention policy is local, et cetera. And I think I did ask him, like, do you have like a minimum or a maximum? And he's like, yeah, no, we don't. And we have to think about that if I remember correctly. But it was very early on. This was like five or six years ago.

SPEAKER_06: Yeah. So today, you know, he could just say, listen, we don't have access to this. It's encrypted.

SPEAKER_173: You're the only people with the keys we do have. We do enforce a log system. So you have to have a log.

SPEAKER_00: You have to use a biometric. So our software is designed that in order to look at the reader, you got to put in your thumbprint and you got to do an IR scan. If you don't want that,

Chamath Palihapitiya: you can't have the system. That's our safe part. So they could create like a very low benchmark or, you know, for basic use of their system. You can't keep data over 36 months. You can set it from zero, one month to 36 months. So they probably will self-regulate their own system. And then the government will probably come to them at some point if they're not self-regulating and then regulate them. And that's what you really want in a vibrant, high functioning democracy is that local folks can have input and make decisions and then people self-regulate to the point at which citizens feel like they've done a decent job. Nobody's complaining about the rating system for movies or explicit language on music. When I was a kid, that was the entirety of the discussion in the 80s. This album, you know, NWA or shocking Indiana Jones and the Temple of Doom, you know, that's where the PG-13 rating came from. Now people are like, yeah, there's a rating on it. Just read the rating and yeah, smoking is dangerous. There's a label on it. So, you know, at a certain point, society just labels things properly. People have decisions and there's some personal freedom. So I don't see the, when I see that chart of Texas, I don't see people not having personal freedom. I see people having personal freedom. They're making a decision in their community and they're electing to put these cameras in. I may not like it as somebody who's privacy concerned, but it is what it is.

SPEAKER_13: I don't, I just, from a very high level, I don't love that I don't recall voting for cameras in

SPEAKER_29: Providence. And also I don't like that my location information on my phone is so widely commercially available. It feels like I am now not able to functionally live in society and maintain a private location. And that to me is a bit of a bummer, but also I don't know how we could have not gotten here given our generally permissive approach to technology and the pace of its expansion,

SPEAKER_15: but it does make me slightly leery. Anyways, Jason, putting that aside. Be vigilant. Be vigilant. Do you have time for a couple of node-y game questions? Sure. Okay. So Channing Hodges, and I've cleaned up his question just a little bit here for you. Sure. What does Jason think about startups using AI to help file patents instead of using expensive

SPEAKER_34: lawyers? Oh, that's a great question. Patents are a pretty sophisticated area of the law. It's not like

SPEAKER_224: just filing a trademark or incorporating a business or doing an employment agreement or doing a convertible

SPEAKER_00: note. So if you were to look at all the things you could possibly work on as a startup, that would be litigation and patents would be amongst the most high risk areas, high stakes areas where you wouldn't

Chamath Palihapitiya: want to rely on AI. You could use AI with human in the loop. In other words, you could do all you want

SPEAKER_36: with AI and then go talk to an attorney. Now, if you didn't have money and you thought,

Chamath Palihapitiya: hey, I can just file a bunch of patents and this AI is making it work and it's the only option you have. Okay, fine. Do it. You just might need the wisdom of a patent attorney at some point. And if people are filing patents, it tends to be a lot at stake. Like you're building some rocket or some incredible

SPEAKER_146: technology and some, you know, something that really needs to have IP protection for some period of time like a drug. And so, yeah, that would not be the place to try to save money. But most startups

Chamath Palihapitiya: don't need patents. That's the other piece is like these business process patents, all these patents, like I almost never see them come into play in business. It's the world's most annoying vanity metric.

SPEAKER_13: Do you know who has the most patents per year? IBM. IBM does a lot of core research, yeah. They have

SPEAKER_15: a lot of patents. Yeah, big impact. All right. Russell, again, translating here, says, what does Jason think about people buying high-end or lower-end hardware to run AI models at home?

SPEAKER_89: Or is this a fad, essentially, the MacBook minis, the Mac studios? Mac studios. It's not a fad,

SPEAKER_06: it's the future. So it's my personal belief is these models are going to be so good,

SPEAKER_67: you're going to want to run them without looking at the price tag of the tokens. And until tokens become, you know, much cheaper, it's going to be much easier to just run these things locally for a certain group of people, depending on the application. And I don't see a time when you wouldn't

Chamath Palihapitiya: want a corporation to have the privacy and not educate the, you know, to our first story today about, or second story about all the money being made by the top two token providers. You also have the privacy issues. Like, do you want your venture capital company secrets and all your investing secrets and all your documents and knowledge being uploaded to Anthropic? Probably not. People are doing it. But you know, now Anthropic understands the entire venture business or understands your documents. And then when the next person makes documents, they just draft off of yours. It's kind of, you got to be thoughtful about it. And then data leaking, something leaks into one of these LLMs. It could be really problematic. It could be incredibly problematic.

SPEAKER_13: All right, Jason, before we let our friends go, do you want to do a quick row.co plug?

Chamath Palihapitiya: Uh, sure. Row.co. I take it to get GLPs. You can go to row.co slash twist and they'll tell you if your insurance will cover it. And here's the good news. When I started on my weight loss journey four or five years ago and lost 40 pounds, um, I did it through fasting. I did it through rocking. I did it

SPEAKER_11: through getting better sleep and I did it through GLPs. At that time, they were $3,000 a month, two or

SPEAKER_352: $3,000 a month. Really? Yeah. It was absurdly expensive to come out of pocket. They were only

Chamath Palihapitiya: for these drugs were only for people doing, who had diabetes and the off label use five years ago was, hey, maybe these work for weight loss. That's when I started using them. I heard Kevin Rose talk about

SPEAKER_67: it. Now you can get into them for 150 bucks, 300 bucks a month. It's literally five, 10% of the cost

SPEAKER_76: of five years ago. Don't be scared about paying for it out of pocket and your insurance will probably

SPEAKER_11: cover it. I think you have to probably have a BMI. I think my BMI at the peak was 32, which was crazy. And I was technically obese. And so it got covered or some portion of it got covered. And now you don't have to worry about that. So row.co slash twist. Absolutely. I highly recommend looking into it.

SPEAKER_15: Hard agree. And for everyone who's a big twist fan on Wednesday, we have two absolute banger guests coming on the show, Jason, including one in the fintech banking space,

SPEAKER_159: let's say that you're not going to want to miss. So we'll see everyone here live on Wednesday. All right. Bye-bye.

SPEAKER_360: Thanks for watching This Week in Startups. If you liked this episode, check out more. If you're a startup founder, Founder University cohort 13 kicks off this fall. It's a 12-week program that provides guidance on building your product, launching to real customers, and pitching to investors. Top startups receive $25,000 or $125,000 in investment. Apply now at founder.university.twist. Already have traction? The Launch Accelerator invests $125,000 and connects you with 500-plus investors to help you raise your next round. Apply at launchaccelerator.co. If you're an accredited investor looking to gain access to quality deal flow, apply for Jason's Angel Syndicate at thesyndicate.com. We find two to three deals a month. And check out This Week in AI, Jason's experts-only roundtable with top AI founders and operators every week. Find it thisweekinai.ai. Check out The Twist Ticker, our daily newsletter, at thisweekinstartups.com. Thanks again to our sponsors for making today's show possible. Follow the show on Instagram. Follow the show on x.com. This Week in Startups publishes three days a week, Monday, Wednesday, and Friday at 5 p.m. Central Time. You can submit an audio or video file question by emailing it to thisweekin.com.