SpaceX IPO Date Talk + OpenAI Confidential Filing Rumors

Transcript

People hate AI. Since we're talking about the public markets, I think these companies waited extremely too long to go public. If you were advising your friends back home to choose between Anthropic or SpaceX, this year, which one would you do? Would you buy a portfolio or would you just buy SpaceX? [SPEAKER_03] I might buy Anthropic. [SPEAKER_03] I was going to say I'd buy SpaceX. [SPEAKER_04] I'd buy SpaceX. Over Anthropic? Space just seems more infinite. [SPEAKER_04] If you're an average person, it's an emotional hedge to invest in Anthropic. It's like I'm really mad about this and it's probably going to take my job. If you lose $1,000 on it, but it might not take your job, you might actually be happier. It's a 6-7. It's 6-7. [SPEAKER_01] Well, hello, friends. [SPEAKER_01] Welcome to More or Less. [SPEAKER_01] We can't find Sam. Hey, Jess. We can't find him. Surprise, surprise. [SPEAKER_01] We know he is going to make an appearance any minute, but I was too excited for this episode to wait a second longer. So we had to start. It's nice. [SPEAKER_03] We can get a few words in ourselves before Sam joins. I have so many topics to cover today with you all, as I'm sure you do as well. [SPEAKER_05] And Dell World Week. I have to close. [SPEAKER_01] I know, but it doesn't work with my... Doesn't work with my... [SPEAKER_01] Oh, Sam. [SPEAKER_04] Hi, guys. Brett said she was glad to get a few words in. But anyway, I was saying how excited I am for the IO IPO week plus Dell World. [SPEAKER_03] Yes. [SPEAKER_03] There's so much happening in tech. [SPEAKER_03] Everyone's trying to do the summer scramble. [SPEAKER_03] So to get all the news out before June, I think is the goal. [SPEAKER_04] Yeah, because then we're all going on vacation. I don't know. I hope so, but that's not happening. [SPEAKER_01] Well, and also we have SpaceX IPO. [SPEAKER_01] S1 is out confirming numbers previously reported by the information. [SPEAKER_01] June 12th, guys, is the rough date for liftoff. [SPEAKER_04] What's June 12th plus 180 days? [SPEAKER_04] That's what I'm marking my calendar for. [SPEAKER_03] Oh, they haven't set the lockup period yet. [SPEAKER_03] So we don't know. [SPEAKER_03] It's the end of the year. [SPEAKER_03] End of December is when all of the housing prices go up. [SPEAKER_05] Efficient market hypothesis says it's already happening. [SPEAKER_01] Breaking news from the information and other outlets is that OpenAI confidentially filing soon. [SPEAKER_01] The Wall Street Journal says as soon as this week. [SPEAKER_01] We cannot confirm or deny that timing. [SPEAKER_03] Question. [SPEAKER_03] Did we wait for the Elon Altman trial to be closed before they announced the IPO filing? [SPEAKER_05] No, they had to. I think they had to. [SPEAKER_01] This also is not an announcement. [SPEAKER_01] It's the next toe in the water. [SPEAKER_01] We're going to circle back to this because we have a lot. And I just think Anthropic must, which is, by the way, closing another round as we speak, must be in the mix. But everyone is feeling good. How are you guys feeling more? And I'm sorry about your untimely loss after one day of your fish, right? Because. [SPEAKER_04] Oh, God. [SPEAKER_04] Wait, what happened? [SPEAKER_04] Did you just not listen to the thing where you massage the tank? [SPEAKER_04] Massage? [SPEAKER_04] Britt didn't follow any of the rules. [SPEAKER_03] Oh, that's not true. It takes three weeks to get the tank ready. [SPEAKER_03] Okay, let your wife explain her epic fail. [SPEAKER_03] So our son really wanted fish. [SPEAKER_03] We were like, fine. We went to the fish store. [SPEAKER_03] We found these really cool fish called Tetra Glowfish. They actually use genetics from jellyfish. And when UV light is on them, they glow neon colors. And I was like, okay, I'm in. Now I'm a fish person because this is scientifically cool. So we pick out three fish. We name the fish. First, we set up the tank. We do what the man said. We set up the tank. We put the water in. We put the pH drops, whatever in. We let it sit. Then the next day we get the fish. We bring them home. We transfer them. It's been three weeks. What? No one told me this. We transfer them in. They're swimming around. We think it's great. I put my son to bed, wake up the next morning. The fish aren't moving. They're at the bottom of the tank and their eyeballs are staring at me. And I'm thinking, great. And now I have to not only tell him they're dead, but it's the first thing I have to tell him when he wakes up, because I know he's going to run straight to the tank. So we had a fish burial down the toilet and we failed, but we're going to do it again. A lot of the internet has told me now what's due right. And we don't give up. [SPEAKER_01] Guys, I have a woman for you. [SPEAKER_01] I think her name is Kate. [SPEAKER_01] She comes to your house and solves this for you. [SPEAKER_01] It turns out that taking care of fish. [SPEAKER_01] There's Uber for fish caretakers. [SPEAKER_05] It's hard. [SPEAKER_05] It's a real microbiome. [SPEAKER_05] There's a whole microbiome of the tank thing you've got to get right. [SPEAKER_05] This is the future of the economy: people that take care of fish. Let's start. Well, Dave, I was bothering you as I want to do with something I needed to ask you. And you were like, I just got off the stage giving a keynote at Dell World. [SPEAKER_01] A, what is Dell World? [SPEAKER_01] And B, what was your keynote? [SPEAKER_04] Well, also, before Dave gets into it, I just want to know, you posted on Facebook about this. [SPEAKER_04] I did. [SPEAKER_04] Or Instagram became Facebook. [SPEAKER_05] It's a real microbiome. [SPEAKER_05] There's a whole microbiome of the tank thing you've got to get right. [SPEAKER_05] This is the future of the economy is people that take care of fish. Let's start. [SPEAKER_01] Well, Dave, I was bothering you as I want to do with something I needed to ask you. [SPEAKER_01] And you were just off the stage giving a keynote at Dell World. [SPEAKER_01] A, what is Dell World? [SPEAKER_01] And B, what was your keynote? Well, also, before Dave gets into it, I just want to know, you posted on Facebook about this. I did. [SPEAKER_04] Or Instagram became Facebook. [SPEAKER_04] It's Dave Moore doesn't use social media like this enough that I got the notification a friend of yours who hasn't posted in a long time is active. [SPEAKER_04] And it's ah, Dave has been brought back to social media. [SPEAKER_01] I like how Sam is more interested in calling you a hypocrite than understanding why you're giving a keynote to 10,000 people, which seems really cool. [SPEAKER_01] So I know why. Well, I came back to Instagram after Stagecoach because I wanted to get my favorite country artists back in my life. And so, yeah, I went to Dell World. Michael Dell asked me to come speak about OpenClaw. I didn't realize how big Dell World is. There were 10,000 people in the auditorium. They bring together similar to NVIDIA's GTC conference. It's largely enterprise focused, largely focused on hyperscalers, people building big data centers. I didn't know this, but Dell actually was the primary builder of the Colossus superclusters for XAI. They're doing a lot of the big supercluster building. And they've also got some really great desktop grade AI computers. I've got a computer on my desk now that can run a 30 billion parameter model locally, tokens for free. It's pretty great. They've also got a desktop that can do a 1 trillion parameter model, which is pretty amazing. We had it on stage with us. So talking about the future of personal AI and inference on the edge, which I've talked about on this pod a lot. That's what I talked about on stage. This is funny. If you click on, if you go Google Dell AI computer, the sponsor result is Dell next-gen AI PCs. But then it just links you to Dell.com slash Qualcomm. So it's these are interesting. They're clearly marketing this, but then Qualcomm? Yeah, it's called the Dell Pro Max, if you just look at the Dell Pro Max. And it's similar to the GGX Spark on the low end. And then the high end, you have an NVIDIA 3800 in a desktop chassis. Pretty cool technology. It's actually, they had to create a liquid cooling system in order to make it work under the desk. And so it's actually pretty interesting cooling technology. So HVAC is the future. Yeah. So got to do that. It was great. Had a beautiful time at the Venetian. And the Dell team was great and really delighted to do it. And I haven't been on a stage that big in a long time. So it was quite fun. We know that we've in, we taught, we don't have to revisit the AI hardware theme that we talked about last week. Well, you know, the one thing I will say, Jess, to some of our conversations over the last several pods, the big topic at Dell World was tokenomics and how to think about token spend. Are you going to keep putting the, my favorite analogy that I heard was in the early days of the 80s and 90s, you could go to an arcade and play video games in an arcade, you know, to keep putting quarters into the machine. And this is what everybody's doing with Anthropic right now, you've got to pay Anthropic quarters to get the tokens out. The pinball wizards. Yeah. And then when Nintendo came along and put compute into people's houses, you could play as many games of Super Mario as you want in your own house. Enterprises and people are having to make decisions like, do you want to pay for the compute up front, get as many token runs out of it as possible? Or do you want to keep putting the quarters into the Anthropic or Open AI machine? It's just the state, the size and shape of where all the matrix multiplying is done, right? [SPEAKER_04] Like if it ends up being like the Apple story is going to be, it's all in the little handsets and consumers pay up front for the token, generate the token multiplying machines. And then maybe the spare cycles get resold. The desktop vision will be the same, SETI at home style, right? Like everyone's got their HVAC cooling plus matrix multiplying machines under their desk. And maybe you retail them back. The cloud version is no, it's all centralized. Well, but Sam, I'll push back and say at the enterprise level, and I learned this a lot this week, [SPEAKER_05] you're actually making the same decision. Are you going to rent it from Anthropic and Open AI? Or are you going to build it on prem and make that investment up front? And a lot of it's going to come down to on versus, which is a whole Linux story and whatever, which is look, if you're Open AI or Anthropic, you would rather own all of it centrally as an asset, right? [SPEAKER_04] If you can't get the cheap financing to do it and then effectively mark it all up and sell it on out. Right. There's just more control that way for you. But if you're a company, right, and you can finance it and everything else kind of lines up the same, you'd rather arrest control. So there's going to be this big control. The only reason that companies will do the centralized thing long term is if the models are sufficiently better that they have to. I think. No, I think there's also a security consideration. People want to put the compute as close to the data as possible for certain use cases. You know, they may not want it going off prem for a myriad of reasons. [SPEAKER_05] Yeah. But obviously, there's lots of reasons to not want it off prem. I'm saying the only reason you're forced off prem is if the giga brain can do something you can't do by multiplying your own matrices. Right. Because it's closed source or whatever it ends up being. [SPEAKER_04] So it's interesting global tension, which then leads to my thing, which is it dawned on me recently that. So, Jess, you'll know this. I think there's also a security consideration. People want to put the compute as close to the data as possible for certain use cases. They may not want it going off prem for a myriad of reasons. [SPEAKER_05] Yeah. But obviously, there's lots of reasons to not want it off prem. I'm saying the only reason you're forced off prem is if the giga brain can do something you can't do by multiplying your own matrices. Right, because it's closed source or whatever it ends up being. So it's interesting global tension, which then leads to my thing, which is it dawned on me recently that the Hyperion thing Facebook's building with that 200 billion dollars spend. Is that the upshot? It's ridiculous. That's insane. That's a huge national security problem. Talk about decentralized versus centralized compute. If we're putting 200 billion dollars of capital in one place, right, that's like 10 battleships. That's 10 aircraft carriers. And so this is interesting. I'm very curious how the security will work out, because the other thing is going to be really interesting on the security. That's insane. It's literally taking 10 aircraft carriers, lining them all up, sticking them all in one place in one center. That seems like a huge security problem. What type of missile defense are these things going to have? Good question. OK, so missile defense for data centers on prem versus cloud. Let's keep on the developer conferencing and talk about Google I.O. Anyone make it to I.O.? This is Google's annual event. Didn't make it, but I watched the full keynote, followed along, actually watched it with our 11 year old Ansel. I know there are some Google haters out there and I know I'm a former employee. However, I think this was a really important event for Google because A, they touted some interesting numbers. For instance, I think they said they have over 900 million active users using Gemini. They also said their advertising revenues continue to skyrocket because of AI. They did 77 billion just in one quarter, which is up 16 percent year over year. When you look at it in the competitive landscape, you're like, whoa, Google's actually right on the heels of Anthropic and GPT. And by the way, their distribution continues to be massive with Chrome and YouTube and Gmail and everything else. [SPEAKER_03] And so what they did yesterday, I think, was really fundamental to getting up to par and starting to get ahead. They launched definitely a bunch of new models. They launched what's called Gemini Spark, which is a 24-7 personal agent. Obviously, everyone's starting to do this now. They launched Android glasses, starting to compete with meta glasses. But the two things I actually really loved, one, universal cart. I think this is huge. [SPEAKER_03] You've already lost me. Too many things, not clear message. I have no idea what's going on or what was good. I like honesty. Who likes honesty? Let's get up to par with what everyone else has. Fine. Not even up to par. [SPEAKER_05] The model's not even competitive. [SPEAKER_03] What's not competitive about it? It's not competitive. It's just less good than every conceivable benchmark. It's less good than 5.5 and 4.7. Meaningful. You wouldn't use it. [SPEAKER_05] Britt, end your point about universal search. [SPEAKER_05] All right. Universal cart. This is a big deal, I think. So Google is launching Universal cart. They have a ton of partners, including Shopify and every major commerce player you can imagine. Stripe, etc. Where there's a universal cart now across the web. [SPEAKER_03] This is important for agents. But also, Google's owning the back end of this. So I think this is their ads business reborn for the future of agentic shopping that I would pay attention to if I were you. And then they also had some really cool generative UI examples where search is basically becoming the front end of Gemini, which sucks for anyone who relied on SEO. But now, if you're like, what should I do this weekend with my family? It'll start literally generating a UI of your plans and all the events you can look at. They use Google Maps integration. They use all kinds of your Gmail to understand context about your family, calendars. Everything's automated and integrated. And it looks really cool, not just a chatbot. So I don't know. I give Google props. I find it confusing to read the launches by Google. It's very difficult. [SPEAKER_03] I mean, maybe that's the point is that they're launching stuff across their entire surface. They've got a billion users. AI is integrated into everything. There's not really supposed to be one way to understand it all. It's just supposed to be everywhere. [SPEAKER_03] I was expecting their model progress to be more aggressive than it was. [SPEAKER_05] If I'm understanding it correctly from reading the subtext of everything, DeepMind has opinions that LLMs simply aren't going to get us there. The way that Google appears to have bet is that world models and other new style types of models are the way to go with research. They're probably right, actually. So maybe it's a good signal that they're not plunging all of their resources into trying to compete at the frontier of LLMs. Maybe that's actually wise. And we should look to them for real innovation rather than just another horse in the LLM race to the CapEx to the moon. But I was kind of surprised that there wasn't more progress there. And they did launch some multimodal stuff, but it looked like a hobby to me. It didn't look like it was that important in the grand scheme of stuff that they're talking about, which is also, again, maybe that's good. There's plenty of people working on that. And we don't need that from Google. But anyway, those are some initial thoughts that I noticed. [SPEAKER_05] I'm somewhere in between. But I think this was Google showing us that its AI playbook is a Google playbook. And remember, there were people, including Google shareholders, who have eaten their words. [SPEAKER_05] And they did launch some magnetic stuff, but it looked like a hobby to me. [SPEAKER_05] It didn't look like it was that important in the grand scheme of things that they're talking about, which is also, again, maybe that's good. [SPEAKER_05] There's plenty of people working on that. [SPEAKER_05] And we don't need that from Google. [SPEAKER_05] But anyway, I don't know. [SPEAKER_05] Those are some initial thoughts that I noticed. [SPEAKER_05] I'm somewhere in between. But I think this was Google showing us that its AI playbook is a Google playbook. [SPEAKER_05] And remember, there were people, including Google shareholders, who have eaten their words, who thought ChatGPT would kill Google. [SPEAKER_05] And it hasn't. [SPEAKER_05] Search is growing. [SPEAKER_05] And Google's AI playbook has been, obviously, to use its distribution, which, by the way, is the most iconic interface in the history of the world, that is the Google search box. To be fair, you're right. [SPEAKER_01] Most people experience AI through Google stuff. [SPEAKER_01] And so I think Google 18 months ago would cannibalize search, blah, blah, blah. [SPEAKER_01] They're just cramming agents into that thing. [SPEAKER_01] So this is a dramatic overhaul to the search box. [SPEAKER_01] Shifting it. [SPEAKER_01] Britt, you nailed it. [SPEAKER_05] I instantly went to my publisher's group chat, editors of newspapers group chat. [SPEAKER_05] I'm like, what does everyone think about the Google search going agentic? [SPEAKER_01] This is the nail in the coffin of SEO. [SPEAKER_01] You're completely right. [SPEAKER_01] They should have known this was the direction all along. [SPEAKER_01] They were surprisingly not. [SPEAKER_01] So I thought that was interesting. [SPEAKER_01] And then, yeah, Dave, I thought the other stuff was a little bit, there was a lot. [SPEAKER_01] Keynotes pack in a lot, because they're also every chance for teams to show off what they're working on. [SPEAKER_01] And part of what Google is doing is like, hear a lot from OpenAI and Anthropic, but don't forget about us. [SPEAKER_03] So I think you felt a little bit of that. [SPEAKER_01] I had to say, I was inspired to try something that did not work well. [SPEAKER_01] So I was inspired, not to actually use anything that was announced, but to dip back into what AI capabilities I can add for our company and Google workspace by add-ons to Google Docs. [SPEAKER_01] And I had a really, really tough time. [SPEAKER_01] But I think when we talk about AI revolutionizing coding and all this, we also have to talk about the mind numbing hours that non-technical people like myself are spending cutting and pasting and repasting and reconnecting and being told that, oh, wait, I actually gave you the wrong thing. [SPEAKER_01] Just kidding. [SPEAKER_01] Over and over again. [SPEAKER_01] And I know I'm a complete novice, but my children were essentially late to school today because I refused to get my butt out of my seat until I solved this. [SPEAKER_01] And I still have not solved this. [SPEAKER_01] So for all of you people who aren't Sam, Britt, and Dave, whose lives are being run by five agents in this agentic future, I see you and it's really damn frustrating. [SPEAKER_01] For what it's worth, Jess, you're right. [SPEAKER_01] This is, I think, one of Google's biggest challenges is that they have too much interface, especially on top of all their Google GCP cloud stuff. [SPEAKER_01] Extremely difficult. [SPEAKER_01] They wanted me in GCP, then they wanted me in App Store, and then they wanted me in Gemini, and then they wanted me in Studio, and I added credit cards. [SPEAKER_01] Gemini tells you in the friendliest voice that you're almost there. [SPEAKER_05] And I just wasn't. [SPEAKER_05] And I do have this experience using Claude as well. [SPEAKER_05] So this is not unique to Gemini. [SPEAKER_05] This one was just really bad. [SPEAKER_01] I think we have to ask, is AI taking over the world? [SPEAKER_01] Yes, in some respects. [SPEAKER_01] But this is the ground truth of what it takes. [SPEAKER_01] And I was highly motivated. [SPEAKER_01] And I still could not get an AI copy editing spell check thing into a template article doc. [SPEAKER_01] I actually think it's not taking over the world for this reason. [SPEAKER_01] The user experience is light years away from being easy enough for anybody that doesn't understand, at a deeply technical level, what's going on. [SPEAKER_01] And I think that's why you see only single digit millions of people using things like Claude Code and Codex, and the most advanced agent tools, because they're just impossible for people that aren't technical to use. [SPEAKER_01] And that's the biggest problem in the entire industry right now, is that we're not putting enough time into design and user experience, and simplifying how to use this stuff. [SPEAKER_05] And Google is definitely one of the worst offenders. [SPEAKER_05] It's just too hard to use all of this interface to figure it out. [SPEAKER_04] But again, the power that they have is like, I don't want to do this in something else because my company runs on workspace. [SPEAKER_05] And so if I'm going to build any of these assistants into our workflows, I'm going to do it in Google. [SPEAKER_05] I'm going to try using a different code generator and seeing if it will make the code from Google, because I don't know what's wrong. [SPEAKER_05] But it's infuriating. So that's my AI from the trenches. [SPEAKER_01] Pull my hair out. [SPEAKER_01] Can I just make two comments on this that I think are both interesting? [SPEAKER_01] One is, it is interesting. [SPEAKER_01] I think the interesting juxtaposition is the number one winning company right now in AI, I think is Anthropic. [SPEAKER_01] And they're actually shockingly focused, right? [SPEAKER_01] It's Claude. [SPEAKER_01] Very pure play. [SPEAKER_01] I understand they do consulting and all that stuff. [SPEAKER_04] But fundamentally, it's a very monolithic focus. [SPEAKER_04] It feels like. [SPEAKER_04] Google, on the other hand, is in the phase that ChatGPT or OpenAI was in before of where it feels like a lot of things, right? [SPEAKER_04] And so from a narrative-focused perspective, that makes sense, the focus. [SPEAKER_04] And for what it's worth, I actually think that is where there's so much leverage in AI. [SPEAKER_04] I mean, it's amazing just building what you want. [SPEAKER_04] I'm having an absolute blast with what I do with it. [SPEAKER_04] But to Jess's point, the thing I don't have is a team. [SPEAKER_04] We kind of have a team that's very small. [SPEAKER_04] But even the tools I built for myself, can I share them? [SPEAKER_04] No. [SPEAKER_04] It feels like a lot of things, right? [SPEAKER_04] And so from a narrative-focused perspective, that makes sense, the focus. [SPEAKER_04] And for what it's worth, I actually think that is where there's so much leverage in AI. [SPEAKER_04] It's amazing just building what you want. [SPEAKER_04] I'm having an absolute blast with what I do with it. [SPEAKER_04] But to Jess's point, the thing I don't have is a team. [SPEAKER_04] We have a team that's very small. [SPEAKER_04] But even the tools I built for myself, can I have them? [SPEAKER_04] No. [SPEAKER_04] I don't have a way to give them to you. [SPEAKER_04] And there's no way for me to collaboratively use them with you. [SPEAKER_04] So for the business cases Jess is talking about, it's going to fit in the G Suite. [SPEAKER_04] The permissioning is going to work. [SPEAKER_04] The tools that you're going to permission to dock with the da-da-da-da. [SPEAKER_04] And so it's a really interesting moment where all the people who can code or know how to build things are freaking out about how good the coding is. [SPEAKER_04] And it is, right? [SPEAKER_04] But then translating that into enterprise, it'll be really interesting. [SPEAKER_04] You get into permissions and privacy and all those types of factors who really wins out of the infrastructure of all of this stuff. [SPEAKER_04] The only thing I'll combat you on there, Sam. [SPEAKER_04] Yes, Google's doing a lot, but they have substantially more people that work for Google to do it with. [SPEAKER_04] Claude and OpenAI have single-digit thousands. [SPEAKER_04] I don't think that matters. [SPEAKER_04] I think it matters. [SPEAKER_04] It doesn't. [SPEAKER_04] Organizational inertia is a negative, I think. [SPEAKER_03] I think the fact that they have five products with more than three billion users that they can start implementing AI into. And even if it takes them longer to string it all together and unify the UI and the user experience across these things, they have the people to build all of that. [SPEAKER_05] And so I feel like they're going broad and they won't go as fast, but over time, they will converge to be the best interface out of at least the ones that do as much as they do. [SPEAKER_05] I mean, I want to talk because I think this is really relevant to Anthropic and OpenAI, which I'm sorry, dear listener, we need to spend some time on this week. [SPEAKER_03] Because what's clear is that OpenAI is trying to move up its IPO timetable. [SPEAKER_03] Anthropic, which had really been ahead of some of the preparations, but has been finalizing a major private round, it's still a little bit unclear. [SPEAKER_03] You have to believe that they're close by. [SPEAKER_03] And to me, you cannot deny the incredible revenue growth of these companies. [SPEAKER_03] And I think if they are confident to be moving up their timetables now, it's partially in response to the positive sentiment building around SpaceX, which could be overzealous or not, we'll know in a few weeks. But their businesses have to also be ripping. Because OpenAI would not have moved up their timeframe if their financials, which are instantly leaked, lag behind Anthropics significantly. And we know that Anthropics, because Dario said this at conferences, are continuing to see this exponential growth. So you have these two AI companies that in really short order are going to be trying to convince public investors they're worth more than a trillion dollars. And they're going to have extraordinary revenue growth to back that up. And to me, it's like, to be bearish on these companies, you basically just have to say that this is not... You have to acknowledge that they are winners in this paradigm. You have to question whether this is going to be the prevailing paradigm for how AI is going to be deployed, right? Or whether Google will erode it somehow with its consumer distribution, or different models local open source will take over. But I'm curious how you guys think about Anthropic OpenAI one trillion. Makes sense? Doesn't make sense? What do you have to believe or not believe? Because the convincing is going to start next week, pretty much. I think all you have to believe is that number go up. We're so past fundamentals. Yeah, but I think you do have to do a little bit of analysis, Sam. How do you unpack revenue number large, but there are real nuances around how large number gets. [SPEAKER_01] My bet is that the people who have done the most recent fundraising rounds, the late stage funds, are doing five seconds of analysis and just work has done access. [SPEAKER_01] I don't think anyone's doing any analysis. [SPEAKER_01] They're just number go up. [SPEAKER_04] By the way, then you get research reports being, we're going to do 15 quadrillion tokens by this year. [SPEAKER_05] They're probably right. [SPEAKER_05] I'd love to see the growth curve in electricity or whatever. [SPEAKER_05] But then the question is all margin, how good a business is it? [SPEAKER_04] I just don't think any of that matters. [SPEAKER_04] I think the reality is, these things go public at a huge number. [SPEAKER_04] I think people will sell less than people think they will, because it's that you're on the train. [SPEAKER_04] Getting off the train is a pretty dramatic move when you have to pay 50% taxes, unless you're a tax-free entity. [SPEAKER_04] So it might be a little bit of rebalancing, but I think it'll be less. [SPEAKER_04] I think people will stay in these things, which means the float and available shares will probably stay lower than people realize. [SPEAKER_04] And then all you need to believe is marginal buyer for a while. [SPEAKER_04] It doesn't matter what number go out for any of these things. [SPEAKER_04] It's literally just no analysis. [SPEAKER_04] This will be bigger in the future than it is today. [SPEAKER_04] And you'd be crazy to get off the train. [SPEAKER_04] We'll be kind of the pitch on them. I would only add one other thing, which is, what rate does big revenue number go up? [SPEAKER_04] Is there some kind of a point at which the circular economy going on around these tokens, slows down, right? [SPEAKER_04] You've got a situation going on now where Anthropic is sucking up all of the venture capital dollars being spent on tokens in the entire ecosystem. [SPEAKER_04] And their revenue growth rate is enormous. [SPEAKER_04] But they are providing little to no downstream value, right? I saw somebody tweeted yesterday or the day before that a platform is only a platform if the revenue being produced by the products being built on top of it is more than the revenue that the platform is generating. [SPEAKER_04] what rate does big revenue number go up? [SPEAKER_04] Is there some kind of a point at which the circular economy going on around these tokens slows down, right? [SPEAKER_04] You've got a situation going on now where Anthropic is sucking up all of the venture capital dollars being spent on tokens in the entire ecosystem. [SPEAKER_04] And their revenue growth rate is enormous. [SPEAKER_04] But they are providing little to no downstream value, right? [SPEAKER_05] I saw somebody tweet yesterday or the day before that a platform is only a platform if the revenue being produced by the products being built on top of it is more than the revenue that the platform is generating. [SPEAKER_05] Was that the Chamath retweet? [SPEAKER_05] I think so. [SPEAKER_05] Chamath by way of Gates. [SPEAKER_05] It's actually a good line. [SPEAKER_05] It was from the Facebook platform days. [SPEAKER_05] I remember it clearly. [SPEAKER_05] And it's just the truth. [SPEAKER_05] Anthropic is making all of the money and none of the companies being built on top of it are making revenue that exceeds their revenue. And so you've got a really serious platform problem. And will that number continue to grow at the rate that it's growing? Or will the venture capital music stop? [SPEAKER_05] Because people can't afford to keep paying these luxury token prices into the arcade, putting more coins in to build their stuff. [SPEAKER_05] Are they getting the actual efficiency out of it? [SPEAKER_05] I'm hearing from CIOs and CEOs that they're not sure. [SPEAKER_05] They're not sure they're getting the token efficiency and the productivity gains. [SPEAKER_05] And how long does that circular economy just keep revolving around? [SPEAKER_05] I like switching roles with you, Dave. [SPEAKER_05] It's fun. [SPEAKER_05] The counter to all that, and we just talked about Google and the Google search business. [SPEAKER_05] I think the thing that I, with Anthropic in particular, that I just don't know how this plays out, is the beauty of Google and Facebook, especially Google more than Facebook, as the best business model in the history of the world, is they get to turn the dials. You need some more money, just add another ad. You know what I mean? You have total marketplace control. The interesting story of Anthropic is, oh, you need to make your quarter? Just make it spend a few more tokens. You need your little bit too much capacity, not enough capacity? Tap down the tokens and make them a little more expensive. If they can get away with the dials themselves, it's an unbelievable business. Because even for all of my curmudgeonliness, you'll find ways to tune the dials to make money. The question is, how much of a marketplace is it? Because if they can't turn the dials, because it's an open-routery situation, oh, they're cheating me here, and I know they're cheating me, so I'm just going to use another provider of multiplying matrices, then it's terrible. I do. That's what I do. That's what you do. No, it's actually funny. It's what I say I do, but in practice, I just don't care and give away my money to Anthropic. I know how to do it a little bit, but here's the... [SPEAKER_04] And the reason why, Dave, is an interesting one. [SPEAKER_04] This goes back. [SPEAKER_04] For me, the incremental cost of, I just want it done. [SPEAKER_04] I don't care if it costs a little bit more money, and I think that is worth overspending on to a point. And I'm not on enough scale that I actually care. [SPEAKER_04] It was thousands of dollars a month. [SPEAKER_04] Maybe I care more. [SPEAKER_04] It's hundreds of dollars. [SPEAKER_04] But the question... [SPEAKER_04] But I don't know what the counterfactual is, because it's hard to measure the output of one model versus another in a lot of the types of tasks I do. [SPEAKER_04] So this is going to be a big game. [SPEAKER_04] Is this a thing where you're like, look, I have to just trust Anthropic, and I'll let them turn the dials on me, and they're going to extract enormous margin, and it's actually a enterprise lot. [SPEAKER_04] Or is it a totally flat, open, electron marketplace with no barriers and no regulation, in which case, terrible business? [SPEAKER_04] Although it still might not matter because numbers go up. [SPEAKER_04] Going back to normies, people are enraged by gas prices right now. [SPEAKER_04] I lived with a mom. [SPEAKER_04] She would drive across town to go to the gas station because gas was 20 cents less. [SPEAKER_04] Little did she know she was wasting gas. [SPEAKER_04] By the way, that's a good idea. [SPEAKER_04] I want a gas billboard for AI tokens. [SPEAKER_04] I'm going to have God make that right now. [SPEAKER_03] Oh, please build it. [SPEAKER_03] That's such a good idea. [SPEAKER_03] Right, and Americans are doing this. [SPEAKER_03] The majority of America are looking at dollars, and they're deciding what the trade-offs are. [SPEAKER_04] Yeah, I mean, literally, look at what was on stage at the White House yesterday. [SPEAKER_04] It was literally the Trump Rx thing. [SPEAKER_05] Joe Gebbia got on stage and showed a map of the lowest prescription drug prices, right? [SPEAKER_03] That's the interface that people want. [SPEAKER_03] I think you have to take this pretty seriously. [SPEAKER_05] I think there's also another thing playing out here, which is people hate AI. [SPEAKER_05] And the question, the more, I have to admit that the more I'm on the front lines of this, the more clear this is to me. [SPEAKER_05] And it's a serious problem. [SPEAKER_05] And as an industry, [SPEAKER_05] and showed a map [SPEAKER_03] of the lowest prescription drug prices, right? [SPEAKER_03] That's the interface that people want. [SPEAKER_03] I think you have to take this pretty seriously. [SPEAKER_05] I think there's also another thing playing out here, which is people hate AI. [SPEAKER_05] And the question, I have to admit that the more I'm on the front lines of this, the more clear this is to me. [SPEAKER_05] And it's a serious problem. [SPEAKER_05] And as an industry, we have really done a bad job with this. [SPEAKER_05] People genuinely hate AI. [SPEAKER_05] And I think you have to look at it from a couple perspectives. But since we're talking about the public markets, I think these companies waited extremely too long to go public. And the fact that people have no skin in the game and they have no ability to invest in these companies, if you wanted to participate in the AI revolution as a retail investor, you have had no option other than NVIDIA, Google. But Sam, the kind of returns that are going, the spoils that are going to the OpenAI and Anthropic investors are insane. [SPEAKER_05] You can't even invest in Anthropic right now if you want to. [SPEAKER_05] But Dave, what are you arguing for? [SPEAKER_05] You're preaching a creed or screed that has been true about the economy for a long time. [SPEAKER_05] How would you fix that? [SPEAKER_05] No, but I'm saying that I think this is playing into how angry people are. [SPEAKER_01] But Dave, I think the thing that you, I agree that people hate AI. [SPEAKER_01] First of all, I think I'd hate it anyway. [SPEAKER_01] And two, [SPEAKER_05] it doesn't matter because they're going to invest in it. What people are going to view is, I might be out of a job, so I better own Anthropic stock, right? It's going to be the man. And they're going to hate it, but they're going to do it because that's their hedge. So it is a hate invest, but they're going to invest. Okay, but where I was going, Sam, is the worst thing that can then happen is that the rate of growth in Anthropic's revenue bobbles and retail is left holding the bag. I agree with this. This is so funny. We've been talking this whole episode and everyone's talking about tokenomics, it's just a crypto flashback. [SPEAKER_05] It's we're still selling tokens. It's still tokens. It's tokenomics. It's who's the retail holder on the token. If number growth rate change bobbles at all and retail gets left holding the AI bag. But they're not, [SPEAKER_05] they don't have enough money to. [SPEAKER_05] The reality is these numbers are so big that retail can't hold the bag. [SPEAKER_05] They don't have enough money. Retail is a third of the market right now. [SPEAKER_04] Yeah, I think they could put in a significant amount of there. [SPEAKER_04] We'll see. [SPEAKER_04] The AI hatred narrative is partially just what it is. [SPEAKER_04] It's also, let's be honest, it's because Sam Altman in particular, but everyone leaned into it as marketing. [SPEAKER_03] This is so scary. [SPEAKER_03] This is so powerful. [SPEAKER_04] This is going to do things. [SPEAKER_04] It was a very explicit strategy for max dollars as quickly as possible. [SPEAKER_04] And you know what? It worked. [SPEAKER_04] But the cost of it working is that everyone hates it. [SPEAKER_04] And it took a long time for other platforms to be hated. This is instant hate. And what I'm saying is if anything goes wrong in the revenue execution path over the next 18 months and retail is left holding the bag, it's going to be double hate. And that's very bad. Maybe. [SPEAKER_04] But here's the funny thing, Dave, is because it's an emotional hedge to, if you're an average person, it's an emotional hedge to invest in Anthropic, right? [SPEAKER_04] I'm really mad about this and it's probably going to take my job. [SPEAKER_04] So there's a kind of funny irony to the hedging. [SPEAKER_04] If you lose $1,000 on it, but it might not take your job, you might actually be happier, right? [SPEAKER_04] So you have to, there's an emotion. [SPEAKER_04] It's a six, seven. [SPEAKER_04] It's a six, seven. [SPEAKER_04] Okay. [SPEAKER_04] Have we solved anything or are we just putting it out? [SPEAKER_04] Of course not. [SPEAKER_04] Why would we try to solve anything? [SPEAKER_04] But I tried. [SPEAKER_04] irony to the hedging. [SPEAKER_04] If you lose $1,000 on it, but it might not take your job, you might actually be happier, right? So you have to, there's an emotion. It's a six, seven. It's a six, seven. Okay. Have we solved anything or are we just putting it out? Of course not. Why would we try to solve anything? But I tried. Okay. So say I'm these companies are getting [SPEAKER_01] their public moment, which by the way, generation wealth event for blah, blah, blah. [SPEAKER_01] And Sam's like, all they need to do is not fall off a cliff in the near term. Does anyone have anything else to add? I mean, yes, that's just the truth, right? I think it is actually really good that finally retail gets access to it. But it's interesting. It's, would you, [SPEAKER_05] obviously we don't provide investment advice on this pod. Dave, would you tell your mom and your sister in Montana to buy? [SPEAKER_05] I actually don't know. And I don't even know if I want to put it in my portfolio. [SPEAKER_05] Well, you clearly don't. Why would you put it in your portfolio? But Sam, [SPEAKER_03] this is really important, right? If you or I [SPEAKER_05] are, that's a pretty important signal. But to be clear, I'm also the guy who does keep not investing in Google for the same reason, which is number two big. It doesn't, I could, you can buy call options on it. When saying that a trillion dollars, it's just not, it doesn't fit my risk profile. They want way more risk or way less risk. Yeah. But for an average person that puts in five or 10K and they can double it in two, three years, they would think that's amazing. I don't know. There are lots of ways to make way more money. [SPEAKER_03] One of the funny things of the public market is, yeah, is it, I actually, is it SpaceX or is it Anthropic? That's actually the real, if you were advising your friends back home to choose between Anthropic or SpaceX this year, [SPEAKER_05] which one would you do? Would you buy a portfolio or would you just buy SpaceX? [SPEAKER_05] I mean, the cult of Elon, if you believe these are all just cult prices, forget the value of the companies, they're just cult prices. I mean, Elon is very good at this, right? The cult thing. Yeah, I was going to say, I'd buy SpaceX. I'd buy SpaceX. Guys, this is amazing. By the way, the AI token prices is up. It's live. It's on the internet. [SPEAKER_03] I'll send, it's currently at tokens.lessontech.co. Epic. It's actually fully live and looks awesome. I see it. I see it. It looks good. Oh, it's so good. That took one prompt in about 30 seconds. Wait, what does it do? I've actually, [SPEAKER_03] I've been building something while we're talking too. It's totally different. I'm shipping so much code. It's awesome. I'm glad you guys are shipping code. [SPEAKER_01] I am cut and pasting getting errors over here. [SPEAKER_05] So send help. This is why Dave and I hope to not be part of the permanent underclass. But look, here's the thing [SPEAKER_01] I don't know about you, Dave, is, I ship a lot of code and do a lot with AI. And then people are, I'm spending $100,000. I'm, how? What are you doing? Because, I'm using this shit really heavily. I think it's all just bullshit. I think they're bullshitting. What are you spending, Sam? I don't know. I think I would notice if it was more than $5,000 a month. Have you guys not used AI to create a centralized family dashboard for your finances? No. [SPEAKER_03] I mean, you guys should see what we're doing at OpenClaw. Like our OpenClaw token spend [SPEAKER_03] is pretty insane. I don't know. I just don't understand these tech people [SPEAKER_05] are, [SPEAKER_04] Sam? I don't know. I think I would notice if it was more than $5,000 a month. Have you guys not used AI to create a centralized family dashboard for your finances? No. [SPEAKER_03] I mean, [SPEAKER_03] you guys should see what we're doing at OpenClaw. Our OpenClaw token spend [SPEAKER_03] is pretty insane. [SPEAKER_03] I don't know. [SPEAKER_03] I just don't understand [SPEAKER_03] these tech people [SPEAKER_05] are saying, [SPEAKER_05] I spent $30,000. [SPEAKER_05] I'm like, [SPEAKER_05] doing what? [SPEAKER_05] I mean, [SPEAKER_05] I guess you could screen [SPEAKER_05] the entire stock market [SPEAKER_05] continually. There are ways I could invent to just blow that level of tokens. But it's very hard to think. I mean, I spend a lot of time [SPEAKER_04] thinking of ways [SPEAKER_04] to burn tokens. [SPEAKER_04] And it's just very impressive. [SPEAKER_04] Actually, [SPEAKER_04] Sam, [SPEAKER_04] here's a practical answer [SPEAKER_04] to your question. I see this at OpenClaw because I think we're running [SPEAKER_04] one of the most, [SPEAKER_04] probably one of the most [SPEAKER_05] sophisticated token engineering [SPEAKER_05] or agentic engineering [SPEAKER_05] operations [SPEAKER_05] outside of the labs. And the way that it really starts to go up is when you're running north of 10 agents at once. We've got people on our core engineering team that are running 50, 100 agents at once. And you're running a continuous integration process, very, very large volume. You can get numbers pretty high doing that. But individuals, I don't know. But Dave, your point about Google betting on other models, I think is also really important because LLMs are awesome. They're going to do what they're going to do. [SPEAKER_01] But obviously, [SPEAKER_01] there's going to be [SPEAKER_01] more models [SPEAKER_01] and more research. [SPEAKER_01] Hopefully. [SPEAKER_01] There's a lot [SPEAKER_01] of unknowns, [SPEAKER_01] actually. [SPEAKER_01] There's people [SPEAKER_01] have opinions. [SPEAKER_01] I talked to one [SPEAKER_01] of the founders [SPEAKER_01] of one of the really, I think, one of the coolest new labs this morning. There's a lot of questions right now about what the next frontier is. [SPEAKER_05] Everybody's got [SPEAKER_05] different opinions, [SPEAKER_05] but it's very early. [SPEAKER_05] It's been world models. [SPEAKER_05] All the fancy people [SPEAKER_05] like Fei have said [SPEAKER_05] it's world models for a while. [SPEAKER_05] There are a lot of fancy people saying a lot of things. [SPEAKER_05] Did you guys see [SPEAKER_05] Carpathia went to Anthropic? How does learning work? Right now, [SPEAKER_01] agents can't learn anything. [SPEAKER_01] There's a lot [SPEAKER_01] of memory projects, [SPEAKER_01] but two years ago, [SPEAKER_01] everybody was building, [SPEAKER_01] I don't know, [SPEAKER_05] a lot of pine cones [SPEAKER_05] and vector databases [SPEAKER_05] and lots of... [SPEAKER_05] What happened to pine cone? [SPEAKER_05] That's what I'm saying. [SPEAKER_05] Every year, [SPEAKER_05] there's a bunch [SPEAKER_05] of tooling that goes on [SPEAKER_05] at the layer [SPEAKER_05] right underneath [SPEAKER_05] these things. And then it gets blown away. And then there's real foundational things that models still can't do. So you've just got these big questions that require a lot [SPEAKER_05] of research work. [SPEAKER_05] I do think I'd rather [SPEAKER_05] own Anthropic [SPEAKER_05] at a trillion [SPEAKER_05] than SpaceX [SPEAKER_05] at a trillion. [SPEAKER_05] Is that crazy? [SPEAKER_05] Why? [SPEAKER_05] No. Well, to be clear, [SPEAKER_01] a trillion is the top [SPEAKER_01] of the current Anthropic [SPEAKER_01] price and by far the bottom [SPEAKER_01] of the current SpaceX price. [SPEAKER_01] I'd probably buy both [SPEAKER_01] if that were [SPEAKER_01] the exercise [SPEAKER_05] that I could do. And of course, to be clear, I can buy none of these. I mean, that is the question that I asked. [SPEAKER_01] What portfolio [SPEAKER_01] would you buy? [SPEAKER_01] Britt asked me why. [SPEAKER_01] Because I'm willing [SPEAKER_01] to believe Anthropic [SPEAKER_01] is a great business. [SPEAKER_05] I'm actually also willing [SPEAKER_05] to believe OpenAI [SPEAKER_05] is a great business. [SPEAKER_05] I don't think [SPEAKER_05] you're going to have [SPEAKER_01] more than that. [SPEAKER_01] I would find it hard [SPEAKER_01] to believe that there'd be [SPEAKER_01] 10 of those kinds [SPEAKER_01] of businesses. [SPEAKER_01] I don't think [SPEAKER_01] five years from now [SPEAKER_01] when we're talking [SPEAKER_01] about AI, [SPEAKER_01] I think we're talking [SPEAKER_01] about more companies, [SPEAKER_01] not fewer. [SPEAKER_01] I think they have [SPEAKER_01] a lot of pieces [SPEAKER_01] to continue to have growth [SPEAKER_01] even in doing [SPEAKER_01] what they're doing. [SPEAKER_01] But there are also is a great business. I don't think you're going to have, I would find it hard to believe that there'd be 10 of those kinds of businesses. [SPEAKER_01] I don't think five years from now when we're talking about AI, I think we're talking about more companies, not fewer. [SPEAKER_01] I think they have a lot of pieces to continue to have growth even in doing what they're doing. [SPEAKER_01] But there are also a lot of big risks that I think people don't talk about. [SPEAKER_01] On the consumer side, it's just the search box is the search box and is the stickiest UX known to man. [SPEAKER_01] And you could have a model that's 50% as good. [SPEAKER_01] And I think Gemini is better than that. [SPEAKER_01] By the way, be pretty free and win consumer. [SPEAKER_01] I think financially I can get the, I can totally understand the Anthropic bet over the SpaceX, blah, blah, blah. [SPEAKER_01] I'm fully okay with the fact we live in a memetic cult war thing. And it's going to be interesting. We've just seen a mini skirmish in the cult war of Altman versus Musk. I think the next cult war will be Dario versus Elon. No, no. Dario and Elon are BFFs because they are compute buddies. I guess we have to throw that in there, Jess. What's your portfolio if you have to choose an allocation between SpaceX, Anthropic, and OpenAI if they're all public this year? [SPEAKER_05] I would probably, because I don't know, de-risk and do something like a third, a third, a third because I really don't know. [SPEAKER_05] But here's the thing on SpaceX. [SPEAKER_01] SpaceX is an incredible company. [SPEAKER_01] They are a monopoly on a thing pretty much that seems super important. [SPEAKER_01] I love Starlink. [SPEAKER_01] I also think that it is possible that orbital economics do not go the way that Bezos and Musk think they will in terms of making sense for all this space industry. [SPEAKER_01] I think space will probably certainly be more important in defense because we're going to fight on every corner of this planet that we can. [SPEAKER_01] But I'm not yet convinced it's fully obvious that all this other industry will move into space. [SPEAKER_01] That is what fuels a big story about the rocket launching business. [SPEAKER_01] I think it's possible, but beyond Starlink. [SPEAKER_01] But I think it's possible, but I think it's also could not follow that trajectory. [SPEAKER_01] And I do think Starlink is so, so, so far ahead and obviously will be vertically integrated with the launch capabilities. [SPEAKER_05] It's a tough next doubling for them, I think. Doubling in what? In valuation or in business? [SPEAKER_01] No, in business, in business, maybe not doubling, maybe tripling. [SPEAKER_01] Look at the Tesla story. [SPEAKER_01] Do they just merge with Tesla? [SPEAKER_04] Well, that's for sure going to happen. I don't believe the Tesla, I guess also I don't want to own a meme stock. [SPEAKER_05] That would make me a bad investor, but it's just not my jam. Isn't that also where all the robots are going? Like all the humanoids are going to go on to SpaceX rockets to go to Mars, right? I do think Anthropic and OpenAI are very good businesses. [SPEAKER_05] I think they're a major unanswered question. [SPEAKER_05] Everyone's pretending that we have answers we don't have, but I also really big answers we don't have. But for them to fundamentally shift their trajectories now would require a series of massive crises, some of whom have already sustained massive crises and weathered them. So I think they are generational companies. [SPEAKER_05] pretending that we have answers we don't have, but I also, really big answers we don't have. But for them to fundamentally shift their trajectories now would require a series of massive crises, some of whom have already sustained massive crises and weathered them. So I think they are generational companies. And then the question is just how generational? And I think people are underestimating things like local compute, open source, other models. I think it's ridiculous to be dismissive of those. But I also, sometimes you do get lightning in a bottle. I don't know. That's what I think. As of today, I reserve the right to change my opinion tomorrow. Let's check it in six months. I don't know. That said, it is the cost piece I just keep coming back to again and again and again. [SPEAKER_03] And I think for many of these, whatever the frontier is, how many people really need to pay a premium for that. But you could make an incredible business by getting a very small group of people to really pay a premium for that intelligence. That could be an incredible business and that might be what they become. And Anthropic will, they have line of sight to a hundred billion in revenue, no question. They're not going to hit that this year or whatever, but that's not a crazy thing for them and their growth is accelerating. So that's how I feel. Big pie at this moment with the existential threat being, of course, that everyone hates AI, including potentially, do we have this executive order yet? I don't know if you guys have been. Oh, the White House has to review all the models before. Yeah, they're not. So what happened? And that's fake news. Yeah, that is fake news. Well, what happened? [SPEAKER_03] So what is happening? Let me get to the bottom of this for [SPEAKER_01] you. So, the information just published. Yeah, I learned about this from the information. What are you talking about? It's fake news. Depends on what you're talking [SPEAKER_03] about. What is happening, I think, is that no, this is not real. I thought it was non-real too until they started briefing these companies today on it. So something has happened that still may mean they don't do it, but they've definitely taken it to the next step. [SPEAKER_01] I think there's various forces that are at work here. You know, the more I understand the policy conversation, there's different agencies with different agendas, and some want it, some don't, some senators, blah, blah, blah. No, Dave, I completely agree with you. And I think the only thing that changed in the last three weeks, three months, is that those factions became a little more public and a little more discreet. [SPEAKER_01] Well, it seems like what's happening is that the power of 5.5 cyber and mythos are pretty up there. You know, as I understand it, these things can rip open binaries, there's no more closed source anymore. It's pretty crazy actually to think about. I don't even know how to think about software going forward. It's all software is open source now, I guess. Well, anyway, it's [SPEAKER_01] pretty up there. As I understand it, these things can rip open binaries, there's no more closed source anymore. It's pretty crazy actually to think about. I don't even know how to think about software going forward. It's all software is open source now, I guess. Well, anyway, it's gotten to the point where they held a briefing for the companies yesterday. So that is saying, talking about plans to pre-review, which [SPEAKER_01] obviously would be a sea change for this administration, although there's much nuance to it. Yeah, I just don't see it happening. They can't, well, I guess maybe they could. Why would the national cyber director have met yesterday with Open AI, Anthropic, and Reflection to walk through the models, walk through the changes? I don't know. [SPEAKER_01] Which agency is that person responsible for? What do they represent? The White House's office. I agree, Dave. In recent weeks, there have been four different drafts of [SPEAKER_05] this thing that we've been leaked and they're all different and they're all from different factions. [SPEAKER_01] So I am totally with you on the factionalism, but for the White House to host a briefing yesterday certainly suggests there's movement in that direction. They could back off. You don't call these companies for a meeting and fill them in on something you're planning not to do two days later. Yeah, it could also just be coordination, right? It's clear that there needs to be pretty serious coordination around these new capabilities and that matters to [SPEAKER_05] national security. It doesn't necessarily mean clamping down control. They specifically discussed reviewing advanced models before they are released. Isn't that already happening? That's the present, right? They're already [SPEAKER_01] taking these things. They're already controlling who gets access to these things. Not reviewing who gets access to them, but presumably reviewing something about how they operate. This is what the Biden executive order attempted to do, but never really did. [SPEAKER_01] Before we wrap, because dear friends, we have a business meeting. We're going to talk about the business of the more or less pot after this. That's what we should really streamline. This is an AI business. We just burn tokens and money. Well, Sam, you're the one who gets these unsolicited sponsorships you hide from us until you plug them. [SPEAKER_04] I think the company that sponsored my newsletter does raise a ton of money. I saw this news flash by. [SPEAKER_01] I was, wow, that's awesome. Let's only do sponsorships for equity going forward. [SPEAKER_04] I have to lose up. I think it was the Victor company that I just saw some headline they raised. And they've been [SPEAKER_03] messaging and they're the ones who had advertised in my [SPEAKER_04] newsletter. I probably should have asked them to invest. I actually don't really want equity in random AI startups. [SPEAKER_04] Why? [SPEAKER_04] I just feel we should have a portfolio of lots of things. I think that's the [SPEAKER_01] stuff that makes me nervous. The 101 billboard economy makes me [SPEAKER_03] nervous. I don't think those are the companies. [SPEAKER_01] Sam's literally bought all the 101 billboards. I have not. I bought one 101 billboard and it was awesome. Okay, we got to Phish. We got to IO, Dell World, IPO. We should mention everyone knows that OpenAI won a decisive victory in their Elon Musk trial that came down to the jury deciding Elon [SPEAKER_01] was too late to file this lawsuit. So they didn't really decide on the thing. [SPEAKER_01] They did not decide on it. [SPEAKER_01] It was awesome. Okay, we got to Phish. We got to IO, Dell World, IPO. We should mention everyone knows that OpenAI won a decisive victory. In their Elon [SPEAKER_01] Musk trial that came down to the jury deciding Elon was too late to file this lawsuit. So they didn't really decide on the thing. They did not decide on it. None of that. Who's trustworthy? Who's worth $30 billion? None of that. Just literally he was too late. The statute of [SPEAKER_01] limitations had run out. I got etiquette on the cover of the Wall Street Journal. That's amazing. Congrats, Sam. So good. The photos were hilarious. Who ever made those photos? Etiquette School had its moment. I don't know. What school [SPEAKER_01] had? Do you think it's passed? Once you're on the front page of the Wall Street Journal, [SPEAKER_01] you jumped the shark, Sam. But yes, what's next for Etiquette School? [SPEAKER_01] Not in my book. [SPEAKER_04] Up next is AI token prices as gas dashboards. This is pretty good, Sam. [SPEAKER_01] For literally three seconds of work, could we get better model names? [SPEAKER_05] It would be so much cooler if they were leaded and unleaded. I never know what all these numbers mean. What does Flash mean? Because a lot of engineers are naming models. They don't have people who actually understand brands. I literally don't know what any of it means. Which one makes code happen fast? Well, it is interesting just looking at this pricing that was just published by [SPEAKER_05] the bot. It is interesting this way. It's actually quite interesting this way [SPEAKER_04] because it just turns out you realize how incredibly expensive Anthropic and OpenAI are. Luxury tokens. Wildly expensive, right? Compared to 5x. 5x Google. So you're like, what's going on with that? [SPEAKER_05] I mean, are they really able, to your point, Dave, it's Anthropic reports a profit by charging four times as much for their tokens. Yeah. And when they control how many tokens they generate, etc. They're literally doing what you said, Sam. You know, luxury tokens. They're pricing at the efficient frontier. Well, I don't want to buy your shitty too. What's going on with XAI? Grok petrol 250 [SPEAKER_05] for an output token and Grok 4.3? Can you explain, just because I probably wasn't listening and I assume maybe others weren't because you move very fast. What is this thing tracked? The cost of tokens. The cost per million [SPEAKER_01] tokens. But it's standardized. Have you checked the numbers? Because I always just want to double check AI's numbers. I'll ask Grok to check whether Claw is right. Okay. So for those [SPEAKER_03] listening, this is a page that has for each of the model makers. It lists their model [SPEAKER_04] and the cost of their token. [SPEAKER_04] I see. [SPEAKER_01] And you just made them look like gas numbers. Well, it makes it easier to feel about them and internalize. It just is wild to see how expensive Exxon Anthropic. Why is OpenAI Mini so expensive? [SPEAKER_04] But these are the questions to ask. You're saying, oh, for Mini, Jess, because that's a reasoning model. It uses more tokens. [SPEAKER_01] I just asked Grok, which now has their [SPEAKER_04] version of Claude Code, which actually seems pretty good. I just started messing with it to check Claude's work. So I'm going to put this on a Claw [SPEAKER_04] and it updates every hour. Yeah, the new Grok is actually pretty good. The new Grok's pretty good, Sam. The normal Americans are going to go with the Google prices of this gas chart. Or the Grok prices. [SPEAKER_01] which now has their [SPEAKER_04] version of Claude Code, [SPEAKER_04] which actually seems [SPEAKER_04] pretty good. [SPEAKER_05] I just started messing [SPEAKER_05] with it to check [SPEAKER_05] Claude's work. [SPEAKER_05] So I'm going to put [SPEAKER_05] this on a Claw and it updates every hour. Yeah, the new Grok is actually pretty good. The new Grok's pretty good, Sam. [SPEAKER_04] The normal Americans [SPEAKER_04] are going to go with [SPEAKER_04] the Google prices of this gas chart. Or the Grok prices. And it's really [SPEAKER_05] funny because [SPEAKER_05] it's open source. [SPEAKER_03] It's owning a Tesla. [SPEAKER_03] What was the old [SPEAKER_03] Tesla? [SPEAKER_05] A Toyota Camry? Your Toyota truck? Prius. A Prius. That's what I was [SPEAKER_04] thinking of. [SPEAKER_04] Thank you, Brett. [SPEAKER_04] So you have Prius pricing. Oh, I went with Together's Cerberlis. I like Together. [SPEAKER_03] Good for them. [SPEAKER_04] It wouldn't be a [SPEAKER_04] more or less if we [SPEAKER_04] didn't at least create [SPEAKER_04] one business [SPEAKER_04] while we were on here. [SPEAKER_04] It's not a business [SPEAKER_04] if it takes one time. [SPEAKER_04] I also just [SPEAKER_04] finished and published [SPEAKER_04] a CLI for set list so that your agent can get the most [SPEAKER_01] recent set list [SPEAKER_01] from your favorite bands [SPEAKER_01] and then auto generate [SPEAKER_05] a Spotify playlist [SPEAKER_05] of the concert [SPEAKER_05] that your favorite band [SPEAKER_05] played last night. [SPEAKER_05] Is CLI replacing MCP? [SPEAKER_05] Yeah, CLIs are [SPEAKER_05] way better than MCPs [SPEAKER_05] in almost all contexts. [SPEAKER_05] Dave, I do want to [SPEAKER_05] point you and Britt [SPEAKER_05] to LessonBoyPodcast.ai. [SPEAKER_05] I already looked at it. [SPEAKER_05] It's really good. [SPEAKER_05] We have to [SPEAKER_01] promote Ansel's game, Britt. [SPEAKER_05] Ansel has a new game [SPEAKER_05] in the App Store [SPEAKER_05] called Tide Shift. [SPEAKER_05] Does Ansel want [SPEAKER_05] LessonBoyGames.com [SPEAKER_03] to promote his podcast? [SPEAKER_03] Yes. [SPEAKER_05] That would be great. [SPEAKER_05] Ansel can do some cross promo in the app [SPEAKER_03] if that helps. [SPEAKER_04] Our children are [SPEAKER_04] very into the economy [SPEAKER_04] and making some money [SPEAKER_04] thinking of their values. [SPEAKER_04] Our Millstone [SPEAKER_04] wants to be a garbage man though. [SPEAKER_03] No, he wants to be [SPEAKER_03] a plumber. [SPEAKER_03] He thinks he... [SPEAKER_01] Great. [SPEAKER_01] He's listening. [SPEAKER_01] He must be listening [SPEAKER_01] to the AI podcast telling him where the jobs are going to be. [SPEAKER_01] I don't think the jobs [SPEAKER_01] are going to be there. They just showed [SPEAKER_01] humanoid robots [SPEAKER_05] cleaning a hotel room [SPEAKER_05] in two minutes. [SPEAKER_05] And I was [SPEAKER_05] well, standard [SPEAKER_05] household cleaners. [SPEAKER_05] Britt, plumbers are [SPEAKER_03] going to be great. [SPEAKER_03] They're set for life. [SPEAKER_03] Dear listeners, [SPEAKER_03] thank you. [SPEAKER_03] We hope we gave you [SPEAKER_03] some meat as well [SPEAKER_03] with the fluff [SPEAKER_03] of this episode. [SPEAKER_03] And you can guarantee there'll be more news happening in tech and that we'll have [SPEAKER_01] opinions on it. [SPEAKER_01] So we'll see you [SPEAKER_01] back here next week [SPEAKER_01] for another episode [SPEAKER_01] of More or Less. [SPEAKER_01] Bye. [SPEAKER_01] Bye. [SPEAKER_01] If you enjoyed this show, [SPEAKER_01] please leave us [SPEAKER_01] a virtual high five [SPEAKER_01] by rating it [SPEAKER_01] and reviewing it [SPEAKER_01] on Apple Podcast, [SPEAKER_01] Spotify, [SPEAKER_01] YouTube, [SPEAKER_01] or wherever [SPEAKER_01] you get your podcasts. [SPEAKER_02] Find more information [SPEAKER_02] about each episode [SPEAKER_02] in the show notes [SPEAKER_02] and follow us [SPEAKER_02] on social media [SPEAKER_02] by searching for [SPEAKER_02] at more or less [SPEAKER_02] at Dave Morin [SPEAKER_02] at Lesson [SPEAKER_02] at Jay Lesson [SPEAKER_02] and as for me, [SPEAKER_02] I'm at Britt. [SPEAKER_02] See you guys next time. of unknowns, actually. You know, there's people have opinions. I talked to one of the founders of one of the really, I think, one of the coolest new labs this morning. There's a lot of questions right now about what the next frontier is. Everybody's got different opinions, but it's very early. It's been world models. All the fancy people like Fei have said it's world models for a while. There are a lot of fancy people saying a lot of things. Did you guys see Carpathia went to Anthropic? How does learning work? You know, like right now, agents can't learn anything. There's a lot of memory projects, but two years ago, everybody was building, I don't know, like there's like a lot of pine cones and vector databases and like lots of... What happened to pine cone? That's what I'm saying. Like every year, there's like a bunch of tooling that goes on at the like layer right underneath these things. And then it gets blown away. And then there's real foundational things that models still can't do. And so you've just got these like big questions that require a lot of research work. I do think I'd rather own Anthropic at a trillion than SpaceX at a trillion. Is that crazy? Why? No. Well, to be clear, a trillion is the top of the current Anthropic price and the by far bottom of the current SpaceX price. I'd probably buy both if that were like the exercise that I could do. And of course, to be clear, I can buy none of these. I mean, that is the question that I asked. Like what portfolio would you buy? Britt asked me why. Because I'm willing to believe Anthropic is a great business. I'm actually also willing to believe OpenAI is a great business. I don't think you're going to have, I would find it hard to believe that there'd be 10 of those kinds of businesses. I don't think five years from now when we're talking about AI, I think we're talking about more companies, not fewer. I think they have a lot of pieces to continue to have growth even in doing what they're doing. But there are also a lot of big risks that I think people don't talk about. I mean, on the consumer side, it's just like, it turns out the search box is the search box and is the stickiest UX known to man. And you could have a model that's 50% as good. And I think Gemini is better than that. By the way, be pretty free and win consumer. I think financially I can get the, I can totally understand the Anthropic bet over the SpaceX, blah, blah, blah. I just like, I'm fully okay with the fact we live in like a memetic cult war thing. And it's going to be interesting. We've just seen like a mini skirmish in the cult war of Altman versus Musk. I think the next cult war will be Dario versus Elon. No, no. Dario and Elon are BFFs because they are compute buddies. I guess we have to throw that in there, Jess. What's your portfolio if you have to choose an allocation between SpaceX, Anthropic, and OpenAI if they're all public this year? I would probably, because I don't know, de-risk and do something like a third, a third, a third because I really don't know. But here's the thing on SpaceX. SpaceX is an incredible company. They are a monopoly on a thing pretty much that seems super important. I love Starling. I also think that it is possible that orbital economics do not go the way that Bezos and Musk think they will in terms of like making sense for all this like space industry. Like I think space will probably, will certainly be more important in defense because, you know, we're going to fight on every corner of this planet that we can. But I'm not yet convinced it's fully obvious that all this other industry will move into space. That is sort of what fuels a big story about the sort of rocket launching business. I think it's possible, but beyond Starlink. but I think it's possible, but I think it's also could not follow that trajectory. And I do think Starlink is so, so, so, so far ahead and obviously will be vertically integrated with the launch capabilities. It's a tough next doubling for them, I think. Doubling in what? In valuation or in business? No, in business, in business kind of, like maybe not doubling, maybe tripling. Look at the Tesla story. Do they just merge with Tesla? Well, that's for sure going to happen. I don't believe the Tesla, I guess also like I don't want to own a meme stock. Not, that would make me a bad investor, but it's just like not my jam. Isn't that also where all the robots are going? Like all the humanoids are going to go on to SpaceX rockets to go to Mars, right? I do think Anthropic and OpenAI are very good businesses. I think they're a major unanswered question. Everyone's sort of pretending that we have answers we don't have, but I also, really big answers we don't have. But for them to fundamentally shift their trajectories now would require like a series of massive crises, some of whom have already sustained massive crises and kind of weathered them. So I think they are like generational companies. And then the question is just how generational? And I think people are underestimating things like local compute, open source, other models. Like I think it's ridiculous to be dismissive of those. But I also, you know, sometimes you like kind of do get lightning in a bottle. I don't know. That's what I think. As of today, I reserve the right to change my opinion tomorrow, but. Let's check it in six months. I don't know. That said, like it is the cost piece I just keep coming back to again and again and again. And I think for many of these, you know, whatever the front, front, front, frontier is, how many people really need to pay a premium for that. But you could make an incredible business by getting a very small group of people to really pay a premium for that intelligence. Like that could be an incredible business and that might be what they become. And Anthropic will, they have line of sight to a hundred billion in revenue, like no question. Kind of. Like I don't, I mean, they're not going to like hit that this year or whatever, but like that's not a crazy thing for them and their growth is accelerating. So that's how I feel. Big pie at this moment with the existential threat being, of course, that everyone hates AI, including potentially, do we have this executive order yet? I don't know if you guys have been. Oh, like the White House has to review all the models before. Yeah, they're not. So what happened? And that's fake news. Yeah, that is fake news. Well, what happened? So what is happening? Let me get to the bottom of this for you. So, oh, the information just published. Yeah, I learned about this from the information. What are you talking about? It's fake news. Depends on what you're talking about. What is happening, I think, is that No, this is not real. I thought it was non-real too until they started briefing these companies today on it. So something has happened that still may mean they don't do it, but they've definitely taken it to the next step. I think there's various forces that are at work here. You know, the more I understand the policy conversation, there's different agencies with different agendas, and some want it, some don't, some senators, blah, blah, blah. No, no, Dave, I completely agree with you. And I think the only thing that changed in the last three weeks, three months, is that those factions became a little more public and a little more discreet. Well, it seems like what's happening is that the power of 5.5 cyber and mythos are like pretty up there. You know, as I understand it, these things can like rip open binaries, like there's no more closed source anymore. more. It's pretty crazy actually to think about. I don't even know how to think about software going forward. It's like all software is open source now, I guess. Well, anyway, it's gotten to the point where they held a briefing for the companies yesterday. So that is saying, talking about plans to pre-review, which obviously would be a sea change for this administration, although there's much nuance to it. Yeah, I just don't see it happening. Like they can't, well, I guess maybe they could. Why would the national cyber director have met yesterday with open AI, Anthropic, and Reflection to walk through the models that walk through the changes? I don't know. Which agency is that person responsible for? What do they represent? The White House's office. I agree, Dave. In recent weeks, there have been like four different drafts of this thing that we've been leaked and they're all different and they're all from different factions. So I am totally with you on the factionalism, but for the White House to host a briefing yesterday certainly suggests there's movement in that direction. They could back off. You don't call these companies for a meeting and fill them in on something you're planning not to do two days later. Yeah, it could also just be coordination, right? Like it's clear that there needs to be pretty serious coordination around these new capabilities and that that matters to national security. It doesn't necessarily mean like clamping down control. They specifically discussed reviewing advanced models before they are released. Isn't that already happening? Like that's the present, right? They're already taking these things. They're already controlling who gets access to these things. Not reviewing who gets access to them, but presumably reviewing something about how they operate. This is what the Biden executive order attempted to do, but never really did. Before we wrap, because dear friends, we have a business meeting. We're going to talk about the business of the more or less pot after this. That's what we should really streamline. This is like an AI business. We just burn tokens and money. Well, Sam, you're the one who gets these unsolicited sponsorships you hide from us until you plug them. I think the company that sponsored my newsletter does raise a ton of money. I saw this news flash by. I was like, wow, that's awesome. Let's only do sponsorships for equity going forward. I have to lose up. I think it was the Victor company that I just saw some headline they raised. And they've been messaging and they're the ones who had advertised in my newsletter. I probably should have asked them to invest. I actually don't really want equity in random AI startups. Why? I just feel like we should have a portfolio of lots of things. I think that's the stuff that makes me nervous. The like 101 billboard economy like makes me nervous. Like I don't think those are the companies. Sam's literally bought all the 101 billboards. I have not. I bought one 101 billboard and it was awesome. Okay, we got to Phish. We got to IO, Dell World, IPO. We should mention everyone knows that OpenAI won a decisive victory. In their Elon Musk trial that came down to the jury deciding Elon was too late to file this lawsuit. So they didn't really decide on the thing. They did not decide on it. None of that. Who's trustworthy? Who's worth $30 billion? None of that. Just literally he was too late. The statute of limitations had run out. I got etiquette on the cover of the Wall Street Journal. That's amazing. Congrats, Sam. So good. The photos were hilarious. Who ever made those photos? Etiquette School had its moment. I don't know. What school had? Do you think it's passed? Once you're on the front page of the Wall Street Journal, you kind of jumped the shark, Sam. But yes, what's next for Etiquette School? Not in my book. Up next is AI token prices as gas dashboards. This is pretty good, Sam. I mean, for literally three seconds of work, could we get better model names? It would be so much cooler if they were leaded and unleaded. I never know what all these numbers mean. What does Flash mean? Because a lot of engineers are naming models. They don't have people who actually understand brands. I literally don't know what any of it means. Which one makes code happen fast? Well, it is interesting just looking at this pricing that was just published by the bot. It is interesting this way. It's actually quite interesting this way because it just turns out you realize how incredibly expensive Anthropic and OpenAI are. Luxury tokens. wildly expensive, right? Like compared to 5x. 5x Google. So you're like, what's going on with that? Like, why are, I mean, are they really able, to your point, Dave, it's like Anthropic reports a profit by charging four times as much for their tokens. Yeah. And when they control how many tokens they generate, etc. They're literally doing what you said, Sam. You know, luxury tokens. They're pricing at the efficient frontier. Well, I don't want to buy your shitty too, like what's going on with XAI? Grok petrol 250 for an output token and Grok 4.3? Can you explain, just because I probably wasn't listening and I assume maybe others weren't because you move very fast. What is this thing tracked? The cost of tokens. The cost per million tokens. But it's standardized. Have you checked the numbers? Because I always just want to double check AI's numbers. I'll ask Grok to check whether Claw is right. Okay. So for those listening, this is a page that has for each of the model makers. It lists their model and the cost of, oh, of their token. I see. And you just made them look like gas numbers. Well, it makes it easier to feel about them and internalize, you know? It just is wild to see how expensive Exxon Anthropic. Why is OpenAI Mini so expensive? I don't know. But like, these are the questions to ask. You're saying, oh, for Mini, Jess, because that's a reasoning model. It uses more tokens. I just asked Grok, which now has their version of Claude Code, which actually seems pretty good. I just started messing with it to check Claude's work. So I'm going to put this on a Claw and it updates every hour. Yeah, the new Grok is actually pretty good. The new Grok's pretty good, Sam. The normal Americans are going to go with the Google prices of this gas chart. Or the Grok prices. And like, it's really funny because you know what, you know, it's kind of like open source. It's kind of like owning a Tesla. What was the old Tesla? Like a Toyota Camry? Your Toyota truck? Prius. A Prius. That's what I was thinking of. Thank you, Brett. So you have like Prius pricing. Oh, I went with Together's Cerberlis. I like Together. Good for them. It wouldn't be a more or less if we didn't at least create one business while we were on here. It's not a business if it takes one time. Ooh, I also just finished and published a CLI for set list so that your agent can get the most recent set list from your favorite bands and then auto generate a Spotify playlist of the concert that your favorite band played last night. Is CLI replacing MCP? Yeah, CLAs are way better than MCPs in almost all contexts. Dave, I do want to point you and Britt to LessonBoyPodcast.ai. I already looked at it. It's real good. Oh yeah, we have to promote Ansel's game, Britt. Oh, Ansel has a new game in the App Store called Tide Shift. Do you want, does Ansel want LessonBoyGames.com to promote his podcast? Yes. That would be great. Ansel can do some cross promo in the app if that helps. Our children are very into the economy and making some money thinking of their values. Our Millstone wants to be a garbage man though. No, he wants to be a plumber. He thinks he... Great. He's listening. He must be listening to the AI podcast telling him where the jobs are going to be. I don't think the jobs are going to be there. They just showed humanoid robots cleaning a hotel room in two minutes. And I was like, well, standard household cleaners. Britt, plumbers are going to be great. They're set for life. Dear listeners, thank you. We hope we gave you some meat as well with the fluff of this episode. And you can guarantee there'll be more news happening in tech and that we'll have opinions on it. So we'll see you back here next week for another episode of More or Less. Bye. Bye. If you enjoyed this show, please leave us a virtual high five by rating it and reviewing it on Apple Podcast, Spotify, YouTube, or wherever you get your podcasts. Find more information about each episode in the show notes and follow us on social media by searching for at more or less at Dave Morin at Lesson at Jay Lesson and as for me, I'm at Britt. See you guys next time.