SpaceX’s Lunar Mass Driver, OpenAI Hires Meta’s Top Ad Exec, Zuck Builds CEO Agent | Diet TBPN

Over the weekend, you watched this live, right? Elon Musk, the CEO of Tesla, the CEO of SpaceX, got some folks together, gave a big keynote presentation about his vision for the future.

Basically something that looked like an Ultradome.

It did look like an Ultradome. And so you watched this, right?

I did watch it.

I watched a little bit of it. And overall—

And if you were taking a shot every time Elon said Epic. You were hammered by not in any, in any state to drive after about 3 minutes.

Yeah, he's got so much going on at this point. I mean, he's doing sports cars, Cybertrucks, Model 3, Model Y, just consumer cars, Model Y XL, space internet, space data centers, space launch capacity, point to point. He's going to take us from one on a rocket from New York to Tokyo in an hour or 30 minutes. Neuralink brain chips, tunnels, these Elon projects have just gotten bigger and bigger and bigger. Of course, he owns X and Twitter, xAI training foundation models, trying to make a consumer app where you chat with it as well as, you know, solve math and physics. Yeah.

This was super rough.

It was, I listened to it and it was a rough delivery. Like he seemed like, he seemed a little like, just like, whoa, okay, I got another thing to present. And so the cadence wasn't quite there and he wasn't able to get the crowd going. There was a lot of like waiting a little bit for the audience to like realize that was the applause line.

Key disclaimer, never bet against Elon. I'm not dumb enough to bet against Elon. I'm not dumb enough to try to short any of his companies. By now you should have learned not to doubt what his organizations can do.

Yeah.

And even though this TerraFab concept, you know, there's a lot of reasons that you would want to bet against it.

Yep.

I don't know that many people that are genuinely sold on, on the space data centers in the near term. It's not clocking for me.

Sure.

John? I'm a dad, I'm a husband. You're not getting it. It's not clocking to you. It's not clocking to you that I'm standing on business, is it?

It's not clocking for me that he's standing on business. He's standing on business and it's not clocking for you.

It's not clocking to me that he's standing on space data centers. Seemingly the first time he's put the Tesla XAI SpaceX logos together.

Sure. No, the tunnel project Hyperloop was a blog post that he wrote and it appeared both on the Tesla blog and the SpaceX blog at the same time. And everyone was like, who's going to be involved in this? And it turned out it was like a couple people from both sides. It became its own company at some point.

I'm not saying it's the first time.

No, no, but it does feel like egot megacorp.

Gearing up for the Elon megacorp. And yeah, so just overall the pitch is he's basically saying like, okay, we're gonna do TeraFab. It's this really hard project and the evidence for why you should believe that it's gonna work well, 'cause he's like, we've done a lot of hard things in the past. Talks a little bit about space data centers, and then he talks about needing something in the order of 1,000 times more compute than—

He says to reach a petawatt, we'll build an electromagnetic mass driver on the moon with robots like Optimus and humans.

No, but before the mass driver, he's talking about how—

Oh, we're chip constrained, which is true.

Elon chip constrained by 1,000x.

Sure.

And so we have to do this.

Yeah, we have to build that capacity.

But he doesn't lay out like a super compelling case of what all the chips will be used for. Right. Is it— are they going to be— is SpaceX going to be a Neo cloud? Is it just for their own internal products like, you know, self-driving, humanoids? Again, my criticism is not of any individual company as much as the presentation. I just came away— he kept saying like, if we can do this, it's going to be epic.

Yeah, yeah, yeah.

And —there wasn't a really strong justification. There wasn't a strong justification other than it's going to be epic. And then he just closes out with this render of the mass driver, which is, again, it's tight. But what was the takeaway on what we're even going to use the mass driver for other than to get to a Kardashev 3 civilization? Which, again, all this stuff just feels so, so, so—

So far in the future.

Far out. He did drop the line turning science fiction into science, science fact. Science fact. No way. Which is, I think it's a Josh Wolfe line. And it's funny because Josh Wolfe was like the hardest critic of Elon.

Oh, that's right. Yeah, that's right.

Like he basically had to stop at some point. Yeah, of course. But yeah, he kept saying these lines that he was like expecting like kind of laughter.

Yeah, it was really hard. I mean, doing like the actual delivery of a, speech on stage. I mean, Jensen just went through this with a keynote at GTC where he didn't have a teleprompter. And he was able to deliver moments and get the crowd going. And the crowd size at GTC is really big. And so even if you just have a couple hardcore fans that really get you and have listened to every Jensen keynote, they know. Yeah, they start clapping. And then everyone starts clapping because it's infectious. And then pretty soon everyone's clapping. This is how it works. So yeah, yeah. I mean, the comedic timing, the delivery timing of like when to clap, when to be, whoa, that was a little tough. I don't know, I actually know who was in the audience. It might've just been employees. It has a certain aesthetic to it. Interestingly about the tariff, this is something I was advocating for back in 2022. So when the Biden CHIPS Act happened, there was this question about like what's gonna happen with Intel. Intel should be at the like the center of the CHIPS Act because the whole pitch for the CHIPS Act was that America is not on the leading edge of semiconductor fabrication. And so we need to sort of pull our weight with our intellectual property laws to stop Taiwan from exporting them. Remember, Taiwan is its own country. They should, in theory, be able to sell their products anywhere. What say do we have over, you know, if Japan wants to sell PlayStations to China or anywhere, why do we have authority over them? Well, it comes down to the intellectual property. There's a lot of patents that have been licensed to ASML, to TSMC. And so we do have the ability to pull levers and strings to sort of limit those, especially if they're packaged by Nvidia, an American company. We have export rules, but it's a stretch. But the whole thesis was America's gonna lose to China unless we do this. And so Intel was a very logical one. And I remember there was like $60 billion or something up for grabs around the CHIPS Act. At the same time, Elon was marshaling around $60 billion for to buy out Twitter or something like that, $40-something billion. And then Intel's market cap, what, they're a $220 billion company now. I think that they were lower back in 2022.

I think it makes total sense for Elon to make chips. Yep. I think, I think it will make sense over time for XAI to start offering cloud services to other companies. Yeah, having compute is valuable. But, but still, this seems like a just kind of a blatant sci-fi pump going into the IPO. Maybe. I will take the other side of that.

The other side of this is that it is good to have a long-term vision, even if it is decades away. And that's what mass driver on the moon is. It's not going to be realized this quarter. It's not going to be realized this year, probably not even this decade. Elon, interestingly, didn't really put a date on it. Did he say something specific?

He said, I hope to see it in my lifetime. Yeah, in his— which is like, I mean, he's 54, he's got a while, right?

Yeah, if he's a radical life extension guy, he could be like 300 years from now, who knows? That's a very vague timeline, whereas in the past with point-to-point, with Roadster, with Cybertruck, with space data centers, all the other Elon pitches have been very focused on, we think we can do this in 5 years. And he always gets criticized for being Elon math, too aggressive, But then history's shown that, you know, he overpromises, underdelivers, but he typically does deliver eventually. And if the technology is completely new, no one else was gonna build it and he's the first one to build it, it's fine because Starlink works. It was the first LEO constellation for satellite internet, had turned into a huge business, wound up being a great technology. And so now the mass driver's a lot crazier. I was going back and forth, Tyler about this, trying to get him to nail down a prediction. We started with 100 years. Do you think it's possible to put a mass driver on the Moon in the next 100 years? And you said yes. So it's possible.

Yes. There's all sorts of qualifications. I think the thing, it was like 300 metric tons.

OK. There was this interesting moment last year when everyone sort of came around to the same AGI predictions or the same superintelligent predictions saying 8, 10 years, George Hotz, Dwarkesh, Karpathy, Sam Altman, there were a number of people that were all saying in various words, like Sam said, a few thousand days. Dwarkesh had this probability distribution that said 2032. And so everyone was sort of in the 7 to 10 year range. And it was just this interesting moment in the AI industry, at least, where everyone had a clear definition of like what superintelligence would be. Simultaneously, everyone was like, yeah, it's not this year, it's more like 10. But there's way, like, we are so far away from consensus around mass driver timelines. I thought it'd be interesting to start developing thought process around what is a mass driver? What does success look like? And what are reasonable timelines here? We have to define what a mass driver actually is. In simple terms, it's an electromagnetic launch system fixed on the lunar surface that converts energy. So you put down some solar panels, you take energy, that's the electromagnetic. So you're not putting up fuel and then because that's extra cost, you, once you're there, you can just repeatedly get sunlight, turn that into energy, and it's going to successfully accelerate a payload to lunar escape velocity, which is 1/5 of Earth's escape velocity. You gotta be going 5,000 miles an hour to escape the moon's gravity well. You have to be going 25,000 miles per hour to escape Earth's. So from an efficiency, an energy efficiency standpoint, it's much better to leave from the moon than to leave from Earth. So then the payload needs to go somewhere useful. Like you make this thing, in this video, he's talking about making satellites on the moon. That's even farther away, but those have to go somewhere. So you can't just launch them into deep space. I mean, unless that's where you're going. Usually you would go to a lunar orbit or an Earth orbit or some point in between the moon and Earth.

Gabe says, what is this mass driver you speak of? SpaceX is getting in the supplements game. That's a great name. They should make a— They've done the Tesla tequila. Yeah, exactly. They should make a mass gainer that they release now and say, hey, if you buy this, you'll get priority access to the mass driver on the moon. I like this.

I like this a lot. I would definitely be a daily driver of the mass driver. And so that, yeah, there's this question of like, you set up this thing that's going to take forever, be really expensive. What are you actually pushing into space? You probably start with basic materials. So rock, Literally rocks, metal, oxygen, water, a lot of useful space infrastructure needs.

What if the asteroids that we get into our atmosphere is just an advanced civilization somewhere else just hurling—

Mass driving? Mass driving. I saw— We're like, "Yeah, we're downrange, guys. Relax." I saw a hilarious take that the first UFO that will arrive on Earth will just be a car guy from another galaxy who got into building a crazy supercar effectively. Taking it out on Earth. And somebody was like, "You can't get to Earth in that." "that thing, that's too modded, that's too crazy. You're going to break down, there's no way." And he's like, "I'll show you." "You're going to be too old by the time you get there." "I'll show you, I've modded this thing." You can also potentially get propellant from the Moon. If you can extract water from ice from the Moon, especially near the poles, that water can support people because you can drink it. Here are the 4 conditions I think need to be met in order to say, "Yes, we have a working mass driver on the Moon." This is how I'd formulate this bet. With Tyler Cosgrove over there. One, you need a permanently installed electromagnetic launcher, and it has to be on the lunar surface. This should be obvious. Two, it has to launch at least 300 metric tons over a 12-month period. This means it's driving significant mass, not just test payloads. Uh, let's aim for 95% mission success and at least 200 launches per year. Failures are going to happen, but we want to get this reliable and at commercial scale, and that means that it can't be breaking down all the time. Lastly, the mass launched from the Moon has to actually be put to use. You can't just have this demo unit that's just blasting rocks into the Sun. It's got to go to something useful. And that could just be water that's going somewhere else or, you know, rocks that are going somewhere else. But it needs to be like we're catching them somewhere. There's a reason that it's going there. Once all four of those conditions are met, it will be hard to argue that we're in some sort of demo phase or publicity stunt. How long will this take? My, like, best case—

You should put the goalposts. On the last driver because the way that you keep moving them. Okay. I'm moving the golf ball.

I put the over-under at like 15 to 20 years, which is about as aggressive as I could get. I think Tyler's going to take the other side of this. Well, I'm going to take the over.

You're going to take over 20 years is I think kind of crazy.

15 years is a blink of an eye.

I think I'm a bit more AGI-pilled than you, Tyler. Okay. So, okay.

Well, let's go through the 6 steps.

Just say, just say you lost faith.

The 6 steps. 6 steps. Okay, so first off, we gotta get to the moon. We can't really get to the moon right now. We're like, we're pretty close with like really small payloads. We need reliable, heavy lunar launch capacity, and that's gonna take some time. So Starship needs to be refueling in orbit, landing on the moon with and without crews, 'cause you're probably gonna need some humans up there, some robots, and we need to start routine cargo flights to the moon. So this is straight up like 3 to 5 years away, like minimum. Then you need to build the power infrastructure. So that's another couple of years because once you're actually going up and back, you have to put down a bunch of solar panels, store it in batteries, and then the batteries need to be able to deliver a lot of power all of a sudden. Now this is what the Tesla does, right? Robotic construction capacity. You got to get autonomous construction rovers at least up and running to lay the electromagnetic track. Then you actually have to build the mass driver on Earth. The good news is that once you build it, once you've built it, like, The benefits are really good. Like you don't have an atmosphere, there's no drag, no weather, no air resistance. And of course gravity is a lot less. So you still need a plan for where the mass goes after it's driven. You gotta catch it kind of. And that's a whole nother problem. But, and when you add all these projects up, you start easily getting into the decades. I mean, if you think about this in like the most grand project you can imagine, which is like the Dyson sphere, basically solar panels directly around the Sun, all the way around the Sun to capture all of its energy. You're going to need to send a lot of solar panels to the Sun. Making those on the Moon is lower cost potentially than making them on Earth.

That's the theory. I'm seeing a tourism angle when the mass driver is pointed back at the Earth to be in a little like bobsled type thing, a little pod to be launched at the Earth. And then you kind of break through the atmosphere and then you parachute. Yeah. That could be good for like gender reveal birthday parties, you know? Potentially. Yes. This needs an Elon-esque figure to accomplish. Yeah, I'm thinking more in that 50-year range.

Over 50 or under 50? We gave Tyler exactly.

I changed. I'm going 30. 30? What?

I'm fucking 30. What changed?

Well, I was just like, AGI is like basically here. There's no way it's going to take that long. There we go. It's like everything will be so cheap.

I mean, I was thinking 5 years. 5 years. Head over to Suspended Cap. What do they have to say? Are they bullish on— is Suspended Cap bullish on the presentation?

Suspended Cap says, "TeraFab is a blatant pump. It actually disgusts me. There will never be leading-edge chips from them. There will never be data centers in space. It's such a blatant effort to IPO SpaceX on the back of some pie in the sky." Sky BS, and then merge with Tesla, do a humongous raise, and keep blah, blah, blah, blah, blah for 5 years. Not a fan. DeSell. Suspended caps of DeSell. There was an obituary in The Wall Street Journal for David Simon. He was a mall king who was both feared and admired. He had ruthless negotiation tactics, an obsession with details, and a sheer force of will that turned his family's real estate business into the largest mall owner in the country. He passed away on Sunday after a cancer diagnosis in 2024. He was 64 years old. During his more than 3— his more than 3 decades as Simon Property Group's chief executive, the inveterate dealmaker gobbled up competitors and defied critics who said malls were obsolete. Over a decades-long buying spree, he cobbled together an empire that spanned 250 properties and 2,000 or 206 million square feet giving Simon control of more retail space than anyone in the world. His success helped propel the family to the wealthiest echelons with an estimated net worth of $11.6 billion, ranking number 38 on Forbes magazine's list of the richest American families. At a time when malls were failing and retailers were filing for bankruptcy, Simon invested in his properties, beefed up their luxury offerings, and added non-shopping tenants such as high-end fitness fitness centers, high-tech mini golf courses, and upscale residents. I feel like that has been the secret to success with the modern mall. There's a lot of malls that have sort of mid-tier or low-end retailers where it is actually pretty comparable to shopping online. You go in, there's no one that's really going to talk to you, the stuff's just there, it might be on the floor, it's messy. But then there are certain malls that have leaned into the higher-end experience and, oh, would you like a drink? Like, let's talk about what you're thinking. You know, show you all this stuff, try this on, and it's much more— it's just in a completely different tier than e-commerce shopping. And so I feel like those properties and those chains have done very well. Along with partners, he snapped up ailing retailers including Aeropostale, Nautica, Eddie Bauer, JCPenney, Forever 21, Lucky Brand, and Brooks Brothers. The arrangement propped up the chains, prevented large vacancies in Simon malls, and turned out to be moneymakers for the company. Effective Monday, Simon's board appointed Eli Simon, one of David Simon's 5 children, as its CEO and president. The mall king was both feared and admired and could be both charming and terrifying. Simon unleashed a tirade at Coach executive Todd Kahn during their first meeting in the decade of the 2000s after the handbag maker tried to renege on deals to open stores in Simon malls. There was no shortage of profanity, Kahn said, The two came to an agreement and became friends. Nothing like a screaming match to kick off a lifelong friendship. David's word was his bond. You could have bang-out meetings, but if you came to a resolution and you shook hands, it was golden. That's a good way of doing business.

Well, RIP to the mall king.

OpenAI is offering private equity firms a guaranteed minimum return of 17.5%, as well as early access to models not yet in public release. There was a lot of chatter on the timeline going back and forth like, is this some sort of weird guarantee? Is this not normal for private equity dealing?

Market Participant had broke it all down to take you through it.

So why don't you read through some of this?

Many are saying OpenAI's $10 billion private equity joint venture is giving a desperate feeling. A lot of people were throwing up the red flag emoji, saying vendor financing dressed up as venture capital guaranteed returns that smell like a Ponzi. A company burning $14 billion this year buying distribution it can't earn organically. I'd push back. TPG Bain Capital, Advent, and Brookfield put up $4 billion. They get preferred equity with a 17.5% minimum return, board seats, and early access to OpenAI's newest model. OpenAI gets instant distribution into hundreds of portfolio companies. No traditional enterprise sales cycles, no selling company by company. The part people are missing: a huge number of software companies are privately held. Thoma Bravo alone owns 75 software companies generating $30 billion of annual revenue. PE firms collectively did $1.8 trillion in buyouts last year. These companies are shifting to heavy token consumption. AWS knows this. They run a dedicated PE sales function. Talk about a dream job.

That— oh, I would kill for that.

Yeah. Yeah. If you love private equity. Yes. Hard to imagine anything better. Advisory programs, transformation consultants, the whole apparatus. Azure and GCP do the same. OpenAI is running the cloud distribution playbook. Wholesale tokens. To PE portfolios at scale. The difference is exclusivity. AWS doesn't ask PE firms to put up $4 billion to take board seats. They give the advisory away. OpenAI is asking for capital commitment, which buys preferential access, dedicated forward-deployed engineers, and a financial incentive for the PE firm to actually push adoption across its portfolio. A PE firm that invested $4 billion in this JV isn't running a parallel RFP with Anthropic. That's the point. The 17.5% return looks like a red flag until you look at it closely. Its preferred equity hurdle, not a coupon on debt. This is the main thing that the hundreds of posts I saw are missing. Priority on returns before common shareholders. So it's preferred equity plus a return on top of that. OpenAI is buying captive distribution channel into enterprises they'd otherwise spend years trying to reach. Anthropic is running a parallel deal with Blackstone, Hellman Friedman, and per Mira, $1 billion common equity, no guaranteed return, more of a Palantir-style consulting venture. Both companies doing this simultaneously tells you where enterprise AI is heading. The bear case is real. 95% of enterprise AI pilots fail to deliver ROI. I've already gone— Yeah, that's not true anymore. But yeah, that was one study. And again, it was every agent coding. $14 billion in projected losses this year. Profitability not expected until 2029. All true. And OpenAI's enterprise business is already $10 billion of $25 billion in revenue, 40% and growing to 50% by year end. The distribution problem is what's left to solve. Um, and anyways, goes, goes on and on. But yeah, overall, I think like people are taking one data point out of context, kind of spinning it into something that, that looks quite extreme and concerning when in reality is just giving them a preferred return. Financial Times had some reporting Saturday, OpenAI to double workforce. Force as business push intensifies. I love a business push. So this is fantastic news. Matt Slotnick says, TBH, the biggest risk to incumbents may be that labs absorb all the top talent. Labs already get great engineering talent, but this will spread to product, sales, marketing, et cetera. When they become the optimal destination to solve the problems of your customers, it's going to get ugly. Yeah, basically, you know, you have a big enterprise instead of going and doing and trying to figure out like the right SaaS solution for a specific problem. You can go to a lab and get something similar. As part of that workforce push, OpenAI tapped former Meta executive to lead their ad push.

Yeah. Why is no one talking about Dave Dugan? The Duganator.

It's, it's crazy. Dave Dugan, a former top advertising executive at Meta, is going to lead sales, ad sales. Dugan, who announced earlier this month he was stepping down from his role as Vice President of Global Clients and Agencies at Meta, has been named Vice President of Global Ad Solutions for OpenAI. He will report to Brad Lightcap, the Lightcapinator. The high-profile hire underscores OpenAI's urgent push to generate new revenue streams to support its enormous funding requirements for its computing needs.

People were complaining that like the advertisers that bought the first ad campaigns on ChatGPT said that process with low tech and that they haven't received much data showing if their ads worked. Two executives at agencies working with early ChatGPT advertisers said they haven't yet been able to prove the ads have driven any measurable business outcomes for their clients. Eric Sufert over at Mobile Dev Memo shared two thoughts on reporting that early ChatGPT ad campaigns lacked performance data. One, this is to be expected. Any early offering will feature limited functionality. Early adopters adapters should be prepared to fill measurement gaps themselves with polling, you know, or any other sort of pixels or tracking. You should be able to work this together. There's been plenty of times— this was the early podcast ads, you know, you'd run a bunch of podcast ads and be like, okay, some people use the coupon code, but a lot of people don't. What multiplier should I apply to that? And then you wind up surveying your customer base and being like, wait, we spent 10% of our marketing budget on podcast ads, but 20% of our customers said they found out about us from a podcast. Maybe we should spend more there, right? Measuring incredibly at asserting performance is precisely what ad agencies are paid to do, says Eric Sufert. He says, if an ad agency says it can't provide performance data because the channel doesn't provide it, it is merely a media buying intermediary. If you're an ad agency, you got to figure out whether or not the ads that you pitch to your clients are actually working. Mark Zuckerberg. Big scoop.

Huge. From Megan over at the Journal. Mark Zuckerberg is building a CEO agent to help him do his job. Employees are also adopting AI agents and AI tools internally, namely MyClaw.

MyClaw. MyClaw. My dude. And Second Brain in a bid to speed up work as they get graded on AI use.

Tyler, where's our MyClaw and our Second Brain?

Yeah, I don't know. I'm so interested to know what Mark Zuckerberg is prompting.

Hope's revenge says ZStack. Yeah, it is ZStack. Honestly, ZStack is like, it's like God mode.

I mean, it does make sense in his position to have a model that's fine-tuned on the internal KPIs, the internal org chart, all this information that's private. And he probably doesn't want to hand that off to another lab that's just going to Maybe look at the data and be like, oh, okay, so Mark Zuckerberg just asked, how do I poach from all the other labs? But there's gotta be so many other questions that he's asking all the time. When you're walking into a meeting with executives, you wanna know, well, how is this division performing? How much money am I spending on Meta Ray-Ban displays? What's the turn rate? Who are our biggest partners? There's a million questions.

Yeah, we're joking around, but ZStack does make sense. It does make sense. You should be able to have God mode.

Yeah. And I mean, this, this has the— there are companies that are starting to do this where they're pitching their agent stacks for CEOs. Bordi is sort of one version of that, right? But there's more and more people that are like, okay, we will come in and fine-tune something for your organization specifically. And maybe do that on the fly. Maybe that's just ChatGPT Enterprise or Anthropic Enterprise or whatever. But Zuck clearly wants something built from the ground up. And it's also a great product demo because if it works really well for him, then the lessons and the learnings from there will probably apply to other people in the ecosystem. So the agent, which is still in development, is currently helping Zuckerberg get information faster, for instance, by retrieving answers for him that he would typically have to go through layers of people to get. Zuckerberg's agent project reflects a drive across the 78,000-person company to accelerate the pace of work, eliminate layers from its organizational structure, and change the day-to-day jobs of its employees to remain competitive with AI-native startups with much smaller staffs. The company views AI adoption as critical to the future of success. Zuckerberg has been spending more time coding recently. He previewed some of the efforts at the company's earnings call in January. Quote, we are investing in AI-native tooling so individuals at Meta can get more done. We're elevating individual contributors and flattening teams, he said. If we do this, then I think we're going to get a lot more done. And I think it'll be a lot more fun. I love that. Leave us 5 stars on Apple Podcasts and Spotify. We are now the number 2 technology show on Spotify. Thank you to everyone who supported us. Who's ahead of you? All In. Have a wonderful evening. Massive stream tomorrow, 11 AM Pacific. Send us reaction to this.