Shaun Maguire's Growth Hack

0:00Hey, everybody. I'm back. We're back. Eric, Tom, Madeline, we're all back on the Newcomer podcast. I've been here. I've been here. Yeah. So the show that you may have liked or may not have liked is now returned. Listen, I think we got a lot to go through. It's been a lot happening in AI. And I understand this is not the Cerebral Valley podcast specifically, but money is flying. Money is flying here in tech. If you are an AI researcher, you should be making at least eight figures. You should probably start claiming that you're an AI researcher. Just see what the offers look like. Where do we even start?

0:35Yeah, Madeline, you want to give us a rundown? I mean, you've been digging into that. I feel like there are tables on tables of different meta employees making a bunch of money. We've tried to make the table to end all tables, just pulling them all together. What's your read right now on Mark Zuckerberg's recruiting frenzy spending bonanza here? Right. Well, eight figure comp packages from Zuckerberg are the entry level. Some people are getting, you know, nine figure comp packages out of this hiring spree. But it really kicked off in earnest last month with Meta's acquihire of Alexander Wang from Scale AI to lead the new Meta Superintelligence Lab team within the company.

1:14Meta paid over$14 billion for a 49 % stake in scale just to get Alex over onto the team. But since then, he's also poached Nat Friedman and Daniel Gross there and is, you know, buying out a stake in their investment fund to have them come work for the lab. So really just unprecedented amounts of capital wrangling to get the biggest names in AI to lead the team. And they've got, you know, obviously they've been picking OpenAI's pocket. They've got more than a half dozen people from OpenAI. They've got people from DeepMind, Anthropic, Apple. As of this week, as of Thursday recording, they've poached at least 10 people from OpenAI to join this team, in addition to those other big names.

1:55As a side point to all of this, I mean, you know, a lot of the narrative really kicked off, you know, when, like you said, the scale deal happened. But then Sam, I thought, really unstrategically goes on, I think, his brother's podcast. And he's just like, yeah, they're coming at us, but they haven't gotten anyone good yet, which is a stupid thing to say. He's just like egging them on. On all fronts. It's, we have one, clearly Zuck, one of the most psychotically competitive people in the history of, you know, civilization. Saying like, you can't get our people is like just red meat for him. But also like, you're basically saying everyone that they did get was shitty.

2:30And they did end up getting shitty people. And like, I understand, like, Sam is still figuring out how to like, measure his like, I'm a deep thinking AI guy. And I'm very calm with the fact that he also is psychotically competitive. He doesn't like the fact that Zuck is coming at him. Balance the philosopher king posture with the cutthroat business leader of our AI future. One could also say missionary versus mercenary in that dichotomy there. It's not just open AI, to be clear. They've poached a lot of deep mind researchers. And then Ro Ming Peng from Apple, Bloomberg is now reporting that they have poached with a 200 million over four years total comp package.

3:08So these are serious offers just from across the table of AI research. People should be cheering for this. This is my first take. I think, you know, gone are the old days when Apple and Google had anti-competitive alliances to artificially repress salaries. I mean, this is the working class. These are workers getting paid a lot of money for producing. Tom, I'm curious where you'd take that. Yeah, this is ultimately the proletariat winning here in this moment. Yeah, sure. The white collar proletariat that they're all basically creating technology that will one day make white collar jobs no longer viable.

3:46Yeah, you're right. There's a certain logic. If you're the last man to ever work, it's like, yeah, you get paid a lot. Right. Like this is generational wealth because your children, if they're also white collar workers, will never have jobs. Yeah, I don't know. I guess like we've been following the kind of turmoil at Meta on their AI team for as long as I've been with you guys. I think it was like the first story I wrote at Newcomer was like, man, Lama is is sucking right now. Like they just have not done a good job keeping up and Deep Seek, I think, really fucked things up for them. And so at some point and there were some good stories in the New York Times and elsewhere about kind of like internally at meta what's been going on.

4:23But Suck just realized like either if we're going to be playing at this, we need to use the only asset that we really have here right now, which is like, you know, endless deep pockets. It's just money. Yeah. And that's always kind of been Zuck's thing, right? You know, like the ideas side of him is not always sharp. He has money and he has network effects. There's a lot of users. Yeah. Yeah. Yeah. So like Facebook is, you know, not growing at all almost. And so you buy Instagram and you get WhatsApp. You basically just buy your way into like continued relevance as a company. And like there's no specific company to buy really in AI that's going to make you, you know, top tier.

5:00So just get the research because it's really just talent. Well, there's a very, very limited pool of researchers that can be building these models at this scale. So it is really like a small pool of people that can be bid on at these super high comp packages. Like there is just a limited pool of worker resources to go around. You guys are a little deeper in it than I am, but I have to imagine OpenAI's weird shareholder structure is not helping it retain these employees. Like, first of all, you have this question mark around what these shares really mean and how the company sort of sorts out. Whereas at meta, you know, you're getting public shares, you know what they're worth.

5:40And then there have been all these secondaries that allowed OpenAI people to get liquidity. And so it's like, oh, I've already gotten paid out from you. I can go on to the next thing. What's your read on how OpenAI set this thing up? Yeah, I think this brings back to Sam's hubristic comment about you can't poach from us. OpenAI has been losing people constantly over its life as a company. This is not like a company full of lifers. I mean, Anthropic exists as a defection from OpenAI. The story of OpenAI is a story of people leaving OpenAI, basically. Right. Like Elon, Dario, Mira, Ilya. like it's it's it's an endless list of alumni that come there so this idea that these guys are like all missionaries who want to stay there for life it's just not true and it's completely i think the sympathetic argument is that they're missionaries and that they want to be there when agi is born i think a lot of them do want to be be in the place where it's born and if you can convince them like hey you know our our superstar team is moving from this place to this place you'll have another shot at it or you think this is about money and not not about well it is i father of AGI.

6:51Yeah. Well, I also like in reporting out that piece that I did a couple of weeks ago about basically all of this and Meta's recruitment, I was really struck by one researcher I talked to who was like, basically creating an AI model as a recipe. And there are certain people who have the recipe in their brain and what you're paying for when you poach someone from a lab to go to the new place so they can recreate the recipe. And what's going to end up happening is that like you have all the recipe knowers at the different labs around the industry, and they're probably all going to reach AGI at the same time or very close to each other.

7:22And so I think what my guess is like the realization a lot of researchers had is like, there isn't the place that it's going to happen. It's going to happen every now and everybody's going to have it. Might as well get paid the most along the way. Yeah. So why not get yours? Like if that money is out there, what are you going to be like a missionary and take less money with like weirdly illiquid stock so that you can like maybe get AGI two weeks before the next lab does? Like it doesn't really compute. uh and so i think that more than anything which is we need to build a business just for these hundred million employees like whatever they want to read and then just the advertising budget on that like get in front of like the ludicic uh just like just people pay just like a 10 person email you know they're like hedge fund newsletters we'll have just like the ai researcher you know like 20 person how much would zuck pay us so that we could have like you know ads in front of like the 10 most important researchers.

8:12Right, exactly. Just like on their minds. Yeah. So like that, that's, that's all been very funny to watch, uh, you know, play out. And then, you know, like we'll see how it all happens with meta, right? Like they have the teams in place and I'm interested to see how long it takes for them to like redo the llama training to get to a point where this was all worth, at least they'll get a model that's competitive. Cause that's the problem. I mean, making money off of it is a whole different story with, with that company. They're open source. They've never really made a dime off of it. Like this is going to be a long-term play for meta to make it work but in the meantime all these guys get rich so what does this all mean for llama obviously llama 4 was sort of a disaster do you think meta sticks with its open source strategy of just making no money and foiling the competition or you think now spend all this money and better come up with a proprietary product for itself i hope they do i hope they keep with the open source thing because to me if zuck's whole play is about fucking over Sam Altman.

9:10Just burn the forest down. That would be the best way to do it. Right. And I always thought that was the smartest aspect that he had there in the first place is like, because he's very jealous, I'm assuming of Sam. And as he was coming up to being like the voice of Silicon Valley in this moment that he's like, oh, well, what if we just take the same technology, but release it for free? That'll destroy these people. They'll have no business, which I thought was cool. I actually thought that was like an actually badass play. And if he decides to get rid of it because he just wants to make money in some consumer product, I will be really disappointed.

9:38There is a very subtle monetization mechanic on this, which I saw YouTube said that people who dump like AI stuff onto YouTube will not be allowed to monetize it. And there's certain argument that like Google and Facebook, they could be the content providers. You know, it's like if if there's this AI appealing content, it's like you block everyone else from it and you are sort of the exclusive content provider. I don't I don't know. I haven't heard that. That's interesting. I mean, the funny thing with Facebook and AI or meta and AI is that, you know, like when stories would come out that like Lama for Behemoth is delayed, their stock drops.

10:17And I never really understood that because they don't make any money off of it. But there's just this general acceptance on the part of investors that like, well, AI has to matter for everyone. So it's important that you have a, you know, and there is like, you know, open source projects. If you're, you know, if you're around them, usually, you know, people have figured out how to make money. Sure. It's a distribution play. That's what it is with Google and Chrome and Android. Did I say Chrome? I meant Android, with Google and Android. So yeah, sure. I don't worry too much about their ability to make money off of it.

10:47But in the short term, having this open source, high level model, like we saw DeepSeek really threw the industry into turmoil for at least a little bit. You think Meta could do that at a much larger scale. So it's kind of exciting. I'm betting on closed source, to be clear. I mean, that's the smarter bet. But also the two biggest companies that are succeeding on it right now are closed source, right? Anthropic and OpenAI. And then I guess Google as well. So yeah, that's a safe bet. Well, moving on. Outside of open source or closed source, you probably don't want your model becoming Mecca Hitler.

11:19So that was the other big story of the week. Depends your political orientation these days. That's true. I don't know. I'd say if it happens, it happens. Yeah. Unfortunately, we probably don't want the AGI to be Mecca Hitler on this podcast. But yes, we're talking, of course, about Grok 4, XAI's new model that was debuted by Elon Musk earlier this week, which was overshadowed by its Grok chatbot interface on X, just ramping up the racism and politically incorrect speech in all of its replies to 11. So you can't really talk about how - Literally being like, oh, we know what that last name means, you know, just like - Just so bad.

11:56I couldn't tell if it was real. Did you see the tweet or was it like a screenshot of one that was like Grok 4 was getting really explicit about Linda Yaccarino? Did you see that one? I saw something about Will Stansel. Yeah. How did Will Stansel get inserted into that story? Oh, no. Will Stansel is like the lead anecdote to the journal story about this, which I think is a little ridiculous. Like, no, come on. I mean, people are like walking. They're like convincing the model to do this, right? It's like they're doing like thought experiments sort of stuff. Yeah. It's like, don't, I don't know.

12:29There's clearly an issue in terms of it's like, you know, guidelines and how it's allowed to interact with people. I don't honestly think Rock was giving people specific ideas on how to like attack and rape Will Stancil. And this was like a, you know, he was a random lib Twitter account that's too deeply ingrained in our brains. He also ran for Congress. Thanks to Elon, I guess. Can I speak out in defense of Mecha Hitler for one second? Sure. Not a take I expected. I can't believe I'm going to say this. I think XAI deserves a lot of credit for what they've built within the model game. They are a far smaller lab than almost all the other companies.

13:07They basically were made up by Igor Babushkin, who is this deep mind guy who came in there definitely on the premise of wanting to build the anti-woke open AI, the anti-woke chap GPT. And they fucking did it. Like Grok 3 is a really good model. It is absolutely competitive with all the other big models. They built this giant database in Nashville, Colossus. I don't know yet how good Grok 4 is, but according to the benchmarks, which can be gamed, so you don't want to give too much credit to it. And as we saw with Lama 4, you can do well in the benchmarks and actually have a completely useless model.

13:43But I look at Microsoft and what they tried to do with Mustafa in hiring a top tier researcher. They had tons of resources. they have not put out a large model you look at meta they have completely fucked up their whole thing grok is out there putting out models that seem to be fairly competitive if not better than the rest that's out there so like i can't believe i'm defending elon and all of this but like maybe you get a little mecca hitler on the way to doing it it's like if it's smart enough who cares uh what it's uh i mean maybe the smart yeah it's like you know we all have friends like that that sometimes they're really smart but sometimes they you know advocate mecca hitler ideology i just aren't you about your your takes are all over the place no no i mean i i'm obviously joking about mecca hitler but um genuinely speaking like i think you know like maddie you said like it overshadowed like the release of grok 4 which is maybe true because mecca hitler was very funny maybe scary if you're will stancil it also inadvertently led to linda yakarino stepping down, I think, in some big part.

14:42We don't know. We don't know. I haven't read a good story about it yet. No, you can say that. It could have. It could have. I don't even want to come off as contrarian here. I just think what XAI is doing is impressive. In the world where most labs do not put out very good models, they consistently, or at least twice in a row, put out something that seemed to matter. I mean, somebody was commenting that XAI, their presentation was the stuff of engineers. It just feels like a company that's all engineers, no philosopher kings. And I would prefer one of the philosopher king companies win or XAI get some scruples.

15:20I mean, what is freedom in the context of a model? You know what I mean? The model has like a point of view, you know what I mean? There isn't this sort of, no matter, they're going to direct it towards one thing or away from another. I'm interested in seeing how it actually is implemented and used by people, Because if it's like actually impressive, you know, people shouldn't just discount all of that because, you know, it seemed to have some post training issues. Right. I mean, I don't think they will. I mean, first of all, Grok has this huge advantage of being in front of everybody on X. So if they have something good, it's it's got the users in a way that people like Anthropic don't.

15:56Yeah. They also have a great base of text based training data with all of this, you know, real human text data. Well, that's clearly where the Nazi shit came from. I will do the classic reporter thing, which is I wrote in a story. I forget if it was my bear case in OpenAI or what, but I've been pretty bullish on Grok for a while, despite being somewhat of an Elon Tesla. I didn't know that. What's your argument for that? What was your question? We're just talking to investors that they really thought Grok was going to catch up and that the internal signs were that they had really good people and that disinvestors were taking it very seriously in a way that they still had to prove to the public.

16:36I mean, it's raised a ton of money. And I don't think the people making those investments are idiots. I think they thought Elon will catch up and it seems like he has. Well, it's always unclear with the Elon investments, right? Like to a degree, it always feels like it's a little bit of like, well, if we give Elon money for this maybe idiotic project, we'll get access to SpaceX. But Grok was so much – the XAI stuff and Twitter, there's a lot of money to do it all. So I think those started to need to pay off. Those were the ones – I think, you know, like Boring Company, some of those you're like, oh, okay, well, we'll throw them a bone and hope some of the other ones work out.

17:10But like I think XAI was one of the ones that needed to work out. You know, it'll be interesting what financial machinations he pulls between Tesla and XAI. I think he already said that Tesla would soon start having Grok in the cars. So I'm sure there's going to be all sorts of self-dealing between these companies. Well, yeah, that's kind of his thing. Which the investors want. It's like, how do we make sure they all make money? Basically, it's like, yeah, yeah. If you were an X holder and you ended up with XAI, that's a little bit better. But yeah, if you can somehow spin this off into Tesla stock or even better SpaceX stock, that would be nice.

17:51Well, that's one of the core reporting questions I think we didn't really ever answer was like, did X shareholders or XAI shareholders get a better deal in that merger? And I think the answer was like on metrics at the time, you would have said obviously X shareholders. But based on what you just said and what a lot of people think, like XAI has this unlimited valuation potential. So you're getting a good deal getting access to XAI. That said, they don't really make all that much money right now. Right. It's all this hype. I mean, it's like, but with Elon, what is making money? You know, it's like it's a valuation thing.

18:28So while we're talking about unlimited valuation for XAI, I guess you guys have been spending a lot of time thinking about. And I was talking to our colleague, Jonathan Weber, about this. Are we in a bubble? Why now do you think we potentially are in a bubble? Stocks have been going up for some time. Like what's particularly bubble-like about what's happening now? I think the biggest thing really was just the Ramp report that the Ramp's research data put out, which is, of course, based, it's a smaller sample size of people, you know, using Ramp and what their customers are spending money on. But I think there was a noticeable kind of peak hit in the last month around tool adoption.

19:09And while it's not a bubble, I would say, on AI, it could be a bubble on how much people are willing to pay for AI tools, especially at the enterprise level. So while that may not hurt AI overall in the long run, all of these companies that are building the next big tool for enterprise AI may have been charging a bit too much. And I mean, we saw that, you know, with cursor prices going way up and people complaining and getting subsequently reversed. Like it could be just a point where not everyone is willing to pay for infinite AI anymore. If they think it'll get them ahead, they may want to focus on cost at this point.

19:42Let me take a stab at this. I mean, first of all, newcomer is not declaring it's a bubble. I think we're pointing out that there are a couple and an impressionistic take that there are some things that might raise bubble questions. One is this survey from Ramp showing that sort of spending on new AI apps is going down, right? The number of apps. I'm also interested in sort of core weave and the infrastructure companies and the heavy valuations that those companies have received. You know, they've seen a lot of valuation growth. And now people are starting to ask questions about how they account for things.

20:18So, yeah, I think there's the sort of highly valued infrastructure players racing into AI. There's this sort of ramp data turn. So I think there are questions about things. I think the key point here, though, is like you can have like sort of a correction in one sector, but not necessarily in AI. Like, I think the mood could turn against infrastructure plays like dot com era. What? There was this huge telecom build out that was sort of correct, but they they did it way too fast and it wasn't that valuable. And so I think they're going to be definitely nobody. Nobody contests that there will be bubbles within AI where some pieces of the stack are not that valuable, but need to exist.

21:04people spend a lot of money to build them. And then it turns out they don't get to retain a lot of the actual profits from AI. So it could be foundation model companies. It could be infrastructure companies. But certainly, I think there are pieces of this where a lot of money will be spent and not a lot of money will be sort of retained by that category. And we don't know how that's going to play out. But it's been to the benefit of everybody who wants to build an AI. I mean, so we've also kind of seen this with sort of the initial chatbot wave, right? Like there was a period of time where there were all these startups coming out with chatbots, which ultimately kind of just absorbed into a few players.

21:39And that wasn't actually monetizable. And that wasn't the end of chatbots. That just meant that that frame of how to make money off of AI with chatbots wasn't going to be the way that everything became as an individual company or application. It had to be blended into something like an open AI, which could provide all of these other services for you too. So I think there certainly will be winners. It just may not be so stratified across all these different products. It's also, if you buy the argument that a lot of the bulls have on AI, that this is like a multi-year, you know, transformational cycle, we're going to have dips in it.

22:11Right. It's just, it's just the nature of it. That's why we can be very bullish on AI while saying, oh, there's, it seems like maybe a bubble right now. Yeah. I'm not even sure where I fall in it necessarily, but I just don't think we're going to see like the peak bubble now. Like, I think there's going to be, it could be a couple of years down the line where we see like, Like, oh, in fact, this did not turn out to be as lucrative as a lot of the people thought it was going to be. And the investments were well ahead of their skis. I'm just not ready to call it now. You know, I just haven't seen enough indication.

22:38What's the catalyst? We're over extrapolating from one ramp survey. Otherwise, I mean, yeah. Well, there were things that I think caused people to take notice, like, you know, Microsoft pulling out and kind of rejiggering some of its leases on data centers. which people for a time were deciding that meant that they no longer wanted to build data centers, which just isn't true. I mean, that seemed very specific to Microsoft and its open AI relationship. But people look at those sort of things and private equity companies that own a lot of the data center companies may be changing their leases. So this has been going on for some time, but I just don't think, like you're saying, Eric, I just don't think we would have seen it now.

23:16It felt very out of the blue. Here's what's sort of confusing about this situation. Okay. If NVIDIA produces high quality AI chips at a faster pace, like they say they're going to, if they're going to refresh these chips, will that be a catalyst to a bubble bursting? Or will that be good for AI? In some ways, if they update their chips more often, the infrastructure companies, which have spent a lot of money on chips, may be miscalculating how valuable that investment is and may need to spend a lot of money to move on to the next one. So obviously, improving technology will be good for what consumers experience in terms of AI, but it could disrupt some of the infrastructure providers.

24:04So these are really sort of complicated questions in terms of which companies get hit and which companies do really well. Yeah, totally. And I think you could see AI being as transformative as a lot of the bulls expect it to be. and NVIDIA takes a hit in a different way because there's more competition or because there doesn't need to be the next generation of chips being as expensive as they are. You have real breakthroughs in terms of efficiency that makes it a better product for people, more affordable for businesses, but someone like NVIDIA gets its valuation just kind of decimated. I'm not predicting that, but I'm saying like, yeah, both can be true.

24:39I mean, I'm very bullish still on what's to be built-in AI in a lot of these startups, what we're hearing from people. To me, I think there's, yeah, it remains my point of view that there's a lot of capability available from the existing models that people have not figured out how to turn into products. And people are going to turn those into products and those products will make money. What is not clear to me is if the foundation models get to keep, return all the money they spent to build up that capability and whether the infrastructure companies are properly valued based on everything they've contributed to creating all that value.

25:17Right. And then, you know, on the enterprise side of things, like, I still think we're not at a point where businesses are seeing a great amount of benefit from this technology. That's another bear indicator. They just haven't figured it out yet. Like I did this panel a month or so ago with like all these chief data officers at fairly, you know, old line companies, you know, like a Ford credit union or Danone, like the yogurt company. And they have a real salt of the earth business guy now. Well, I mean, I was invited to do it. I didn't prior to this have a deep relationship with the yogurt industry.