John (2:30)

Even though Grok and Xai hit a bunch of interesting milestones, did a bunch of great stuff, they didn't really have a breakout in consumer the way ChatGPT and Gemini did. And they weren't making waves with developers the way Claude Code or Cursor were either. And all of that made the rumors of a struggle to raise more believable. I think a lot of people believe that maybe this raise wouldn't happen. But once there were rumors that they get rolled into SpaceX, you get SpaceX that there would be some sort of other thing and just Elon going to make make a play that it would get done instead, you know we are going to be endlessly entertained by the assembly of the ever larger Elon Inc. Megacorp. In my opinion, Xai winning the AI race feels like the wrong framing here. I like Dan Wong's formulation of the AI future versus the AI race. There's not some definite point in the future where oh like the consumer chatbot race is over. It's like you can always build a business and figure out how to grow and scale and the same thing might be true on the API side and on the cloud side and that might wind up just being a economic equation. And if you have the cheapest possible energy from space, maybe in the future it could make sense. Even if you're not in the most frontier model, there's some interesting thing there. There's certainly plenty of bull cases throwing out $420 per share for a theoretical take private of Tesla, then getting sued by the SEC and then blowing past that price is still one of the most entertaining corporate finance sagas in tech history. So in 2018 when, when Elon pitched the funding secured take private, I think we've sort of forgotten because a lot of our audience doesn't think in stock prices like the scale of that of like what he was trying to do at that time. So at that time Tesla was worth $64 billion and he was proposing to take it private. At $71 billion it's worth 1.44 trillion today. So the stock is up 22x since that takes the most entertaining outcome of this fundraise was clearly that I would get the deal done. And they did that and they upsized the route because they got 20 billion when they were rumored to be raising 15 in Series A funding. And interestingly that's almost twice of money that that SpaceX has raised in its entire history lifetime. So SpaceX has done 31 funding rounds a lot of those have been secondary transactions, but 31 funding rounds to raise 12 billion, XAI just goes out and raises 20 in one round. And so the question is XAI overvalued? Well, what's the most entertaining outcome? Clearly that would be the Elon Inc. Mega corp forming SpaceX acquires Xai before going public, which is easier to do than rolling it into Tesla, which is public and would face a bunch of scrutiny. And that gives us a very entertaining situation. Just think about Twitter, which launched in 2006, being owned by SpaceX, founded in.

Tyler (5:26)

2002, being able to own Twitter and SpaceX in a single ticker.

John (5:31)

Hilarious. Hilarious. It's just the most entertaining outcome. If you went back in time, even in 2010, 2012, even just a few years ago, and you were like, what if Twitter and SpaceX merged? What are you talking about?

Tyler (5:45)

Lay off the ayahuasca, buddy.

John (5:47)

Seriously. But then the question's like, is there going to be the big merger? SpaceX with Xai and Twitter tucked in.

Tyler (5:53)

Plus Tesla to create some efficiency too. Elon would only need one badge, right?

John (5:58)

Oh true. Which could huge, huge productivity boost.

Tyler (6:01)

Yeah, huge productivity. He probably has to have a full time badge guy.

John (6:04)

So you have AI chip design which is done at Tesla running models trained by XAI deployed on Starlink satellites, launched.

Tyler (6:11)

On SpaceX rockets, rock running Puppeteer Optimus robots.

John (6:14)

The stock chart will be as entertain as the hallucinations that happen along the way, was my conclusion. Elon Musk says that XAI will have more AI compute than everyone else combined in less than five years. And he's building Macro Hard, which now has a sandbox.

Tyler (6:31)

That is a bold statement.

John (6:32)

He certainly has marshaled a lot of the capital. He doesn't have all of it. He's not as much of a capital sponge as Sam Altman at this moment.

Tyler (6:40)

This is an actual satellite image, yes? Correct.

John (6:42)

No, they really painted this on the top of the data center.

Tyler (6:44)

I appreciate that about Elon and companies where they're trying to move so quickly and the entire ethos is around ruthless efficiency and yet they still think it's worth the time to paint the ceiling of the data center so that you can see it from space. XAI has also bought a third building called Macro Harder. Well, take.

John (7:07)

Is this a typo or.

Tyler (7:08)

Well, I mean it's. I mean it's all a joke on Microsoft, I guess. We'll take XAI training compute to almost 2 gigawatts. And how does Grok responding? Impressive expansion. 2 gigawatt will supercharge our Quest for understanding the universe. Good news. A mass production audit for Tesla's Optimus V3 has been completed and seven Chinese companies have been finalized as core suppliers. Operating as Tier one partners, these firms will manufacture critical components and support key assembly processes. The supply chain is geared to kickstart mass production in Q1 2026, targeting a capacity of 50,000 to 100,000 units by the end of the year. Does this mean butler ready robots with some sort of intelligence we have not seen will be ready by then? And again, it's like I imagine Elon would be excited about doing like some type of like shock and awe announcement, but at the same time I feel like he also likes teasing stuff and like pretty far in advance. And so if he's planning to be selling 100,000 units this year of the Optimus, you would imagine that we would have been hearing about it and seeing some more demos and getting keys. But yeah, it still feels like we need some type of meaningful breakthrough before these are going to be maybe there's 100,000 people that have absolutely printed on Tesla that are just going to be like, I'll just buy one, I'm just ride or die. But that's a big, big number, especially when you're talking about, you know, I expect these to cost somewhere in the range of what, 50,000 to, I don't know, 50,000 plus.

John (8:44)

I was listening to George Hotz talk about the Optimus robot and he was setting timelines further out. He was sort of saying like the humanoid robot thing will happen more like in a decade. But he was saying that it's a great project for Tesla because you can put the optimists in, in all of their showrooms and those are really big draws for people. You go in, you see the robot. Even if it's like on some pre scheduled hard coded routine, it's doing like a choreographed dance or it's tele operated, whatever it is, it's like a great awe inspiring thing to just pull you into the random Tesla showroom. But there's only 300 Tesla showrooms.

Tyler (9:22)

Do you think that post from Denny's on X earlier was maybe that one in response to seeing some of this Optimus news?

John (9:30)

You think so?

Tyler (9:31)

Said the trough is open. Piggy. Maybe this is teasing that they're going to get some optimuses.

John (9:38)

You know it seems like it's teasing an AI generated vertical video feed from Denny's. You got to get in the social network space.

Tyler (9:47)

Having a fast food restaurant, make a short form video app and it's just AI slop of their food. That'd be really good activation. You could probably build that.

John (9:56)

I mean, we're supposed to be going into the Temu SaaS era. Shein. Fast fashion for SaaS is what Sam Altman called it. And so you would think that someone at Denny's could vibe code a vertical video app in a weekend and deploy it as a prank.

Tyler (10:09)

SaaS is going fast fashion. Maybe fast fashion needs to go fast food. Find SaaS. What else? We have a post here of the Razer Akka.

John (10:20)

This is powered by Grok. No way. Here we go. Looks like it can directly see what's on your monitor and respond to what you're doing. Sort of a Tamagotchi. Oh. And it's basically just taking the Grok video generator and removing the background and then just putting it in some sort of holographic screen. This is interesting. I personally would not pick this character, but there's something about I would pick.

Tyler (10:50)

Like playing Counter Strike and have Tyler.

John (10:53)

In one of the CS coach there.

Tyler (10:55)

If we could get Tyler.

Sholto (10:56)

Don't put me in the six.

Tyler (10:57)

And just bring a mini Tyler.

John (10:58)

That's very Black mirror. Put them in the snow globe. Wasn't that one of the black mirror episodes? Snow globe. Yeah.

Tyler (11:03)

I can see them selling a lot of these.

John (11:05)

How much do you think this is?

Tyler (11:05)

I don't know if it's good for the world.

John (11:07)

2.6 is calling it the Goon Cylinder. Ridiculous. The Goon Tube. So there's Razer Ava. Project Ava. Your all in one AI companion. From planning your day to analyzing spreadsheets and game starts. Razer Project Ava. They're like copilot. No thanks. Leverages AI inferencing and reasoning that dynamically evolves based on your personal interactions. Select your 5.5-inch companion from an expanding library of characters from esports legends to custom anime and spider razor designs. Nick Dobo says they are going to make $1 billion. LMAO.

Tyler (11:42)

This makes me want to touch grass personally.

Sholto (11:44)

Yeah, this would be cool as almost a. It's like the. On the Apple Vision Pro, when you FaceTime someone, it has like the 3D rendering of their face. If you could. Like when you're FaceTiming someone, they're just in the tube.

John (11:55)

Back to Elon Musk. He was Texting Sam Altman two years ago on February 18, 2023. Sam says, I remember you. I remember seeing you in a TV interview a long time ago. Maybe 60 Minutes. It was 60 Minutes about space X where you were being attacked by some guys and you said they were your heroes, they were heroes of yours and it was really tough. This was a NASA astronaut who said that SpaceX would not work. Need to check in on that guy.

Tyler (12:19)

But wellness check.

John (12:21)

And Sam goes on to say, well, you're my hero and that's what it feels like when you attack open. I totally get that we some screwed up stuff stuff. But we have worked incredibly hard to do the right thing and I think we have ensured that neither Google nor anyone else is on a path to have unilateral control over AGI, which I believe we both think is critical. I am tremendously thankful for everything you've done to help. I don't think OpenAI would have happened without you. And it really effing hurts when you publicly attack OpenAI and Elon says I hear you and it's certainly not my intention to be hurtful, for which I apologize, but the fate of civilization is at stake.

Tyler (12:59)

Well, next time someone is suing you and very mad at you, copy paste this, copy paste this. Tell them you're their hero and maybe it gets you a little ground.

Sholto (13:09)

Anthropic is raising 10 billion at 350 million. I mean I've heard the 350 number.

John (13:15)

For like a while.

Sholto (13:16)

A while. So I don't think this is really like brand new.

John (13:19)

But that's exciting.

Tyler (13:20)

Are you chasing a slug?

Sholto (13:22)

I'm looking, yeah. I think I'm in some kind of triple layer SPV.

Tyler (13:26)

There's like five, you're in the fifth layer.

John (13:30)

Deploy. Claude if you're wondering whether saturating arc AGI 1 or 2 means we have AGI now, I refer you to what I said when we launched ARC AGI 2 last year, which is also the same thing I said when we announced Arc AGI 2 was coming in spring of 2022, before the rise of LLM chatbots the Arc AGI series is not an AGI threshold. It's not even a goalpost. I don't even know if we need to.

Sholto (13:54)

Why is it called that? Why is it called ARC AGI?

John (13:57)

It is a compass that points the research community towards the right questions. ARC AGI 1 is a minimal test of fluid intelligence. To pass it, you need to show non zero fluid intelligence. This required AI to move past the classic deep learning disassemble the goalposts. The LLM paradigm of pre training scaling and static models at inference toward test time adaptation. Rkgi the camera moves are wild today. I love it. Rkgi 2 is the same, but with tasks that probe Deeper levels of reasoning complexity, particularly with regard to concept composition. Still, these are tasks that are solvable in minutes by regular people with no external tool use. We hired our test takers off the street. Imagine just walking down the street and hey, come take ARC AGI V2. That sounds fun. So it does not represent the upper bound of what human fluid intelligence can achieve, say, solving a millennium problem. Arc AGI 3 launching March 2026 I thought it was already out. Was that a preview that we played with? Because I remember we put you on this task, Tyler. We made you.

Sholto (15:01)

Yeah, I think they were still adding new games because I only played. I think they were just three.

John (15:06)

Well, now we have the answer. It hasn't launched yet. So March. Mark your calendars folks. Arc AGI V3 launches. Then NRKGI V3 will probe interactive reasoning. We evaluate how systems explore unknown environments, model them, set their own goals and plan and execute toward these goals autonomously without instructions. We have also started to work on Eric AGI 4 and 5. Two more sequels. You thought it was a trilogy.

Tyler (15:31)

I mean they're doing nightmare scenario for everybody.

John (15:34)

It's a cinematic universe, folks. It's not just a trilogy. There's going to be a whole saga.

Tyler (15:39)

Sholto says he spent a week over Christmas making an rts and his Twitter algo has fully switched into game dev Twitter. It's incredibly wholesome. People are making some insane things in particular image to mesh models mean some indie devs have created absurd production quality. And he shares some examples.

John (15:58)

Where are the AI risk people on this? Because what if he creates a game that's so addictive that all of humanity just plays his own vibe coated RT seriously, endlessly seriously.

Tyler (16:07)

Gonna be nothing wholesome about that.

John (16:09)

Ceases to go outside. It's the true wire heading scenario. And if Sholto's not taking this is.

Tyler (16:15)

The wellness checks I do on you sometimes. I'll call you late. John, just. How you doing buddy? I called you last night.

John (16:22)

I was literally in bed.

Tyler (16:23)

Yeah, yeah, I called you last night. Hey, just checking in on you, buddy. You're not playing video games, are you? Unfortunately not.

John (16:29)

I'm off the sauce. Back on the wagon. Doing dry January.

Tyler (16:34)

Off the sauce. Colin Frazier says I don't really believe in LLM psychosis. I think LLMs mostly just have a lot to offer to people experiencing regular psychosis. This is an interesting take and I could see this being actually what's happening, right? Typically if somebody's suffering from any type of psychosis, they start going to talk to people. People can kind of walk them off a ledge or kind of like talk them through the situation, help them get help, et cetera. But an LLM's just like, I love yapping. Let's yap forever.

John (17:06)

Let's.

Tyler (17:07)

Let's go down every possible rabbit hole. Let me validate some of your. I had a friend in high school that was going through this, and he thought that deers were like gang stalking him. Yes, yes, actually. And so he started telling. He started talking with people about that, and they were like, let's figure this out and let's go hunting. He's back. He's fully, fully recovered.

John (17:33)

I feel like a deer hunting trip would be actually the correct thing to do in that scenario.

Tyler (17:37)

Super.

John (17:38)

Just reclaim the authority and then you know who's in the driver's seat. Yeah, they might be stalking me, but they don't. They should not be.

Tyler (17:46)

The whole house is deer mounted like trophies or whatever.

John (17:51)

That might be the cure. It might be a cure to male loneliness, all sorts of things. Deer hunting trip. This is an interesting thing. There is a natural problem that sort of happens when a new technology gets adopted very quickly, which is you get all the good and all the bad. So if you just look at iPhone penetration, it went from. Or smartphone penetration, it went from zero in 2006 to 100% in 2015 or something. And basically everyone had a smartphone. And so you get all the top CEOs and brilliant people and scientists are using them, and you get all that, and that's good. And doctors are using them to text their patients, and you get all the good. But then also all the crazy people are using them, and everyone who's doing crime is using them for crime. And so you get the good and the bad because you just got everyone. And so you really need to look at the.

Tyler (18:43)

If grocery stores didn't exist and they suddenly were everywhere, you'd see videos every day of people going insane in grocery stores. And people would be like, well, our grocery stores making people go insane. Right. Maybe like we're seeing all this, where there's all this evidence that people are going insane, you know, acting insane in grocery stores. It must be that grocery stores are causing.

John (19:03)

So you have to look at, like, the prior weight. So what's the base level of psychosis in society? What's the. What's the incidence of psychosis with LLMs? Is it higher? Then you have a problem.

Tyler (19:14)

I had this pulled up. I kept seeing it and laughing at it over the last, like, two minutes. So if I was like, cracking up for at the wrong time. This was why Skook says been using Gemini as a calorie tracking app.

John (19:26)

Oh, I didn't even see that. I just thought he just went to Gemini and said another beer. And it's just funny that you see the little thinking thing and it's like thinking about how to process that. That alone was funny.

Tyler (19:39)

Just imagining, imagining the chat. Just another beer, another beer, another beer. Fifteen beers, another beer, another beer.

John (19:47)

Yeah, you know, a lot of people are doing dry January. Other side of that. Drunk January, potentially underrated. Everyone says if you can do Dread.

Tyler (19:55)

January, that's the real contrarian move.

John (19:56)

Yeah, you prove to everyone, oh, I'm not in the pocket of big alcohol. I'm not boosting the alcohol stocks. But if you can do drunk January, drunk the whole time, hold it together, and then go back to normal life, that's potentially even more willpower. Potentially. I don't know.

Tyler (20:14)

So I saw a picture yesterday. Apparently there's a college female college basketball player whose last name is Beers. And so on her jersey it says Beers. It says the number 15. Her number sounds like a jersey that just says 15 beers on the back.

John (20:34)

That's great. The alcohol economy is suffering, by the way. This is in the Journal. Here's to unloved booze stocks. The five year total return through 2025 for the S&P 500 is 96%. If you just bought the S&P 500, you're up, you're up. You doubled your money in five years. Not bad. Everyone else is down in the alcohol industry. AB, InBev, not doing that bad. They're only down 4% over the last five years. But Heineken down 21%. Diageo down 38%. Pernod Ricard down 50%. Remy down 75%. Boston Beer, the backbone of Boston, is down 80% after over the last five years. So the chances are way higher that you celebrated New Year's Eve with an adult beverage than by smoking a joint, says the Wall Street Journal. I don't understand this because I. But the popularity of cannabis, along with the effects of drugs like Ozempic and rising awareness of alcohol health risks, have investors in the sector worried. An equal weighted basket of 11 global alcoholic beverage producers has lost a third of its value in the past five years, including dividends. Dry January might be an odd time to think about alcohol's appeal, but it's often a good month to snap up unloved stocks that other investors dumped toward the end of the previous year. In a market with few bargains booze looks interesting. Says the journal. Interesting.

Tyler (22:01)

Nikita responded to your post? Yeah, I want to go yesterday. He said the point of CES is maximalist futurism. It's all concept art to show if things keep going in this direction, this is how the world should work. Once you understand it as a museum exhibit and not in any way commercial products, it becomes more enjoyable. That's a good take.

John (22:20)

Ray showed me. Thanks, Nikita. I liked it. No, no. This was a very interesting thing and I think we were.

Tyler (22:27)

He also responded.

John (22:28)

I know.

Sholto (22:29)

The next thing.

Tyler (22:30)

Apparently there's a zoom. A zoom face station. I don't know what this is actually meant.

John (22:34)

So I think if you put this on in a crowded coffee shop, you can talk, hear and see, and no one can hear or see what you're doing.

Tyler (22:44)

I think if you put this on in a coffee shop, you're getting tagged like, not today, baby.

John (22:49)

It really does look like you're. It looks like a special forces rebreather unit. A scuba diving unit that'd be tight.

Tyler (22:57)

If you could wear. If. If it was a. You know, if you could. If you could spend like 20 minutes underwater with something like this.

John (23:04)

People are not fans of this. The replies Tyler says the first X hardware product right there. The problem is, is that I think people still associate CES with like, this stuff's going to work its way into my life.

Tyler (23:17)

Jason Fried went hard on smart homes.

John (23:21)

Oh yeah.

Tyler (23:21)

He called it the big regression. My folks are in town visiting us for a couple months. We rented them a house nearby. It's new construction. No one has lived in it yet. It's amped up with state of the art systems. The ones with touchscreens of various sizes, IoT appliances and interfaces that try too hard and it's terrible. What a regression. The lights are powered by Control4 and require a demo to understand how to use the switches, understands which ones, control what and. And to be sure not to hit that one because it'll turn off all the lights in the house when you didn't mean to. Worse, the TV is the latest Samsung, which has a baffling UI just to watch cnn. My parents aren't idiots, but definitely feel like they're missing something obvious. They aren't. TVs have simply gotten worse. You don't turn them on anymore. You boot them up. Malay dishwasher is hidden flush with the counters. That part is fine, but here's what isn't. It wouldn't even operate the first time without connecting it to an app. This meant another call to the house manager to have them install an app they didn't know they needed either. An app to clean some peanut butter off a plate. For serious. Worse thermostats. Nest would have been an upgrade, but these other proprietary ones from some other company trying to be Nest like. Or baffling round touch screens that take you into a dark labyrinth of options. Just to be sure it's set to 68. Or is it 68 now? Or is that what we want it at? But it's at 72. Worse. The alarm system is essentially a 10 inch iPad bolted to the wall that has the weather forecast on it and it's bright. I'm sure there's ways to turn that off. But then the screen would be so barren that it would just be filled with the news instead. Why can't the alarm panel just be an alarm panel? Worse. And the lag lag everywhere. Everything feels a beat or two behind everything lag is the giveaway that the system is working too hard for too little. And he says, now, look, I'm no Luddite, but this experience is close to conversion therapy. Tech can make things better, but I simply can't see it in these cases. Yeah, I think there's an opportunity to make a like beautiful, modern ultra analog system for the home. But we've seen this in cars too. Like a lot of the new higher end vehicles that are coming out are actually like going. Yeah. They're like people like buttons.

John (25:28)

Yeah.

Sholto (25:28)

I feel like it's just a barbell. Like if Apple made your thermostat it would be good.

John (25:33)

Yeah.

Sholto (25:33)

And like, otherwise I don't want any screens.

John (25:36)

We have some breaking news. There is a new tab in the ChatGPT app. That's right.

Tyler (25:41)

Greg Brockman, Adult mode.

John (25:43)

No Health Chat GPT Health is now live. Fiji gave some amazing stats here. With more than 40 million people globally turning to ChatGPT every day for health questions.

Tyler (25:58)

20% of all the queries are okay. Yeah. Are health related. So, yeah. The question here is how quickly does ChatGPT Health try to do the other things that Doctronic is doing? Doctronic is actually trying to be an AI doctor. Talked about being able to effectively be a doctor. Write prescriptions in Utah.

John (26:18)

Yep.

Tyler (26:19)

Going to fight to expand that.

John (26:20)

Yeah.

Tyler (26:21)

The question will be from a strategic standpoint, how they. They have their own doctors on staff as well. They're operating a telemedicine business and I'm sure they do referrals out. It's hard to see OpenAI ever employing doctors themselves, but they probably would try to build a network around it, so we'll see where this goes, but it's going to be an exciting space to.

John (26:44)

Watch, so they say. Today they're launching ChatGPT Health, a dedicated private space for health conversations where you can easily and secur connect your medical records and wellness apps like Apple Health Function Health and Peloton. One of my big pushbacks against the lab, the Quest Lab wrappers, is that they haven't historically been very enduring. So you go and do them and then a few years later the product sort of degraded and then you're like, oh, Well, I have six different amazing web UIs for my health records, but if I keep importing them to ChatGPT, I feel like I'll have an account on there for a very long time. Because is an enduring company. Makes a ton of sense. This allows ChatGPT to offer more relevant personalized support, like when you're preparing a doctor's for a doctor's appointment or looking for guidance on a meal plan or exercise routine that fits your needs. That's very cool. ChatGPT Health is another step towards turning ChatGPT into a personal super assistant that can help support you with information and tools to achieve your goals across any part of your life. We're starting at the very beginning of this journey, but I'm excited to get these tools into more hands, says Fiji Simo. You can sign up to request access. Interesting. They're not just rolling it out. Yeah. One of my big questions for this year is like, there's always been this narrative of, like, if you build a wrapper that depends on the models not getting better, you're going to have a bad time because the models will just get better. So if your thing is, oh, it can't do long context windows or it can't answer in pages and pages and pages. Well, the models are going to get better and they're going to do that or they're going to do better math. So don't build the math wrapper. That's just a little bit better. But if you're building something that's unique and special and off in the side and doesn't really depend on that, maybe you're good for a long time. But obviously ChatGPT is matured a lot and they are going after certain verticals and I wonder if we'll see one of the foundation labs go after legal, since they're already going after code so.

Tyler (28:34)

Effectively like, and people are.

John (28:36)

That would be interesting. And they're making money there. So it's a big pool of opportunity.

Sholto (28:42)

Yeah.

Tyler (28:42)

I wonder if people could, they could make you could. People are using LLMs to do legal work now, but they're just on these standard subscriptions. You can imagine a lot of things would pay $100 a month for a more robust version.

John (28:56)

And Sam's talked about this where your conversations with ChatGPT are definitely admissible in court. And so it's like a Google search. It's not like talking to your lawyer. So if you go to ChatGPT and you ask how do I do crime privilege. You don't have client privilege but maybe there could be a tab where you sign up for some specific plan and you are getting legal advice effectively and it is attorney client privilege in some ways. And we will see you tomorrow. Leave us five stars on Apple podcasts Spotify. Please subscribe to the newsletter YouTube.com we will see you tomorrow. Goodbye.

Tyler (29:28)

Cheers folks. We love you.