Meta's AI Agent Plan, Grok's Perversion, Prison Of Financial Mediocrity

0:00Michael Lewis here. My bestselling book The Big Short tells the story of the build up and burst of the US housing market back in 2008. A decade ago, The Big Short was made into an Academy Award-winning movie. And now I'm bringing it to you for the first time as an audiobook narrated by yours truly. The Big Short's story, what it means to bet against the market, and who really pays for an unchecked financial system, is as relevant today as it's ever been. Get The Big Short now at pushkin.fm slash audiobook or wherever audiobooks are sold. Meta is buying AI startup Manus to do what exactly? Grok undresses people and apologizes.

0:43Who's responsible? And why are so many young people turning to prediction markets and sports betting? Here's a theory. That's coming up on a Big Technology Podcast Friday edition right after this. Welcome to Big Technology Podcast Friday edition, where we break down the news in our traditional, cool-headed, and nuanced format. Matt, we have a great show for you today. As we kick off 2026, we're going to talk about why Meta is buying the AI agent startup Manus. We're going to talk about Grok. I wouldn't even say going rogue, but basically following the prompts of people on Twitter and bikini-ifying, if that's a word, and undressing people and then apologizing afterwards.

1:21Who's responsible for that? And then we'll also take a look at a great piece called The Prison of Financial Mediocrity and an explanation for why so many people are turning to sports gambling, sports betting, and prediction markets. And joining us as always to do it here as we kick off the year is Ranjan Roy of Margins. Ranjan, great to see you. Great to see you. Happy New Year. Where in the world are you, Alex? Tell our listeners. Happy New Year. I am in Peru. It's been a pretty wild couple of weeks in South America, a couple more days here, and then back to New York. How about you? How's your Christmas break Ben.

1:56It's been great. I'm back in New York. Finally, I was in I was at a very nice New England holiday vacation up in Boston, then New Hampshire skiing. So excited to get back to it, though. And we both had Meta Ray-Bans fails, which we're going to talk about as we go through this Meta segment here. And we'll get to that in a moment. But I think the biggest news over the past week has been that Meta bought the AI startup Manus for more than$2 billion. This, according to the Wall Street Journal, meta-platforms has agreed to buy Manus, a Singapore-based company with Chinese founders that conducts deep research and performs other tasks for paying users.

2:33The deal is being closed at more than$2 billion. Manus gained a wide following after previewing in March an AI agent capable of producing detailed research reports and building custom websites using AI models developed by companies such as Anthropic and China's Alibaba. The demo followed the release of DeepSeq, a made-in-China AI that rocked Silicon Valley because of its advanced capabilities, coupled with claims by its developer that it was developed with far less computing power than American rivals. I think for those who follow AI closely, we all remember the Manus AI agent as sort of the highlight of this company, with like the journal kind of winked at as it highlighted the China connection at the beginning of this story.

3:20But I'm actually curious to hear your perspective, Ranjan, on this one. What do you think the significance is of Meta, a consumer company buying Manus, which seems enterprise focused and AI agent focused? What's happening here? Do you have any theory of the case? Yeah, so this is very squarely in my world. And for longtime listeners, I work at Writer, which is an enterprise AI focused company. And almost six months ago, we were debating because I was saying, what is an AI agent is being redefined. It's no longer these kind of blueprints with blocks stitched together. But in reality, it's just kind of like defining tools, connectors, giving a prompt and letting the agent go do its thing.

4:05And I think this is the year – this was one of my predictions last week. This is a year that that vision of what agentic AI is is going to be realized, the power of it. So I think Manus is just another – they definitely had one of the more popular somewhere between consumer and enterprise. Still more consumer focused. You can get on there for like 30 bucks a month. But still, the idea that you can do things very quickly and define these workflows that actually start to work. And I think that's going to be the battleground. OpenAI is getting involved with all their enterprise focused hires recently.

4:44like I think this is going to be one of the biggest parts of 2026 and that's self-serving as well. But I think what's interesting is that last point you brought up is that why is Meta buying them? And honestly, I was racking my head and I don't understand how this gets integrated into the product. The product, it's like this isn't going to be helping consumers, you know, book their flight. Remember when that was what everyone said Agentic would be? This is something different. It is going to be more work and enterprise focused. So that part still leaves me guessing other than Alexander Wang and just went to Mark and said, this person's smart.

5:25You can buy them for two and a half billion and it's not just paying a researcher a billion dollars. Okay, so I have some thoughts on this. So first of all, the Manus team doesn't even seem like it's going to be reporting Alexander Way and it's going to be reporting to Meta's chief operating officer, which is something that the story said in the beginning. In the release, Meta said, we plan to scale the service to many more businesses. My belief is that is a head fake. I think Meta is making a big consumer agent play. They think the people at Manus know how to do this. They think that over the coming years, like especially if you're right, Ranjan, and that the AI agent stuff will start to work, consumers are going to want to go to an app that's going to be able to handle that travel booking scenario for them pretty well.

6:14And I think Meta is doing something really smart here, which is it realizes that's going to be a competitive edge. It's going to be a differentiator. The people that get the credit cards first are probably going to be the ones that start to embed that consumer behavior in their AI agent. And that's what they're going after. And notice how the statement is very interesting. We plan to scale the service to many more businesses. Maybe they'll do that. But ultimately, I think the interesting thing here is what they do, what they didn't say, which is that, yes, they might scale it to some more businesses.

6:47But ultimately, this is a way of them getting into this, you know, the space where Apple intelligence wanted to play, build the AI platform, build the thing that people want to use that gets things done for them and tell the world that, yeah, this is more enterprisey as you come out the gate. really, though, developing the consumer application here. And if you own that platform, then you're going to win. I love this. OK, you know what? I'm starting to I'm starting to feel that a bit more because to me, one of the most kind of like amazing parts of how quickly seeing agentic AI evolve in this way is like it just makes so many basic things I do regularly a lot easier.

7:31Like when you receive an email from this one recipient, go research this person, go research this company, send me a Slack notification. Like that's all stuff I have fully built into my life, but professionally. So there's so much opportunity around that on the consumer side. But yeah, you're right. Who has the actual like economic incentive to actually push this within their product? Open AI keeps saying the word Greg Brockman again tweeted, enterprise agentic AI adoption and scientific research are the two most important trends of the year. So everyone is just leaving consumer agentic to the wayside.

8:13And you're right, Apple intelligence, I mean, we all know. So, OK, I like that. And Manus is very, very slick in terms of its UI. It's very user friendly. So if something could be more consumer than enterprise, I think it could definitely be them. I like it. And you need that. You need that specialized knowledge to be able to go do that. I mean, Meta had brought this whole super intelligence team aboard. And like their task is like build the best model. But ultimately, this is again going to product. This is how you productize these things. And I don't think it's any coincidence that this is happening as Fiji Simo, the former head of Facebook app, is sitting inside OpenAI and thinking about how to build an ad platform.

8:55And not only that, when you start to research things in OpenAI, within ChatGPT, there are new, I don't know if you would call them maybe like cards that will help you do more browsing for things that you might make a purchase decision on. I know for myself personally, and I'm sure anyone that used these tools, like tools like ChatGPT to plan their winter breaks. I mean, for me, going through South America, ChatGPT was super important. It was a pivotal tool. And the one thing that was missing was for me to just say, all right, that sounds like the right option. Go ahead and book it. And if Meta's models can get to the point where they're helpful in this way and the agentic side of things can get to the point that they're useful in this way, then all of a sudden Meta's AI effort goes from something that's kind of laughable, something that Jan LeCun, even today in the Financial Times is saying, you know, they fudge the benchmarks in a way for Llama 4.

9:55They go from that to becoming a player. Now, there's a lot of ifs here, right? I said if a lot of times, but ultimately it seems to me like if this is what they're doing, it's a sound strategy. Yeah, even more. I think let's not even talk, but are you actually booking that flight to South America or that tour? Or like if you start to think about it, there's so many elements with shopping, with travel, with anything consumer focused, with health, with like personal health, where you're asking these kind of recurring routine questions or have these ongoing conversations. And rather than having to go to ChatGPT and remember to type it in, you're already scrolling Instagram and Facebook.

10:38And now you just have one additional block that's like, oh, yeah, do you want to pick up where we left off? I actually just found this other article that would be useful for you or this other flight alert that do you want me to go book it? So like having you already locked into that interface does start to get pretty interesting. We're going to get into Adam Massari and what he was talking about, like AI within the feed. But if this is the direction, I'm starting to find this pretty interesting right now. All right. One more thought about this. You mentioned the agent going proactive and suggesting, hey, remember, we were talking about this.

11:14What do you think about this? What does that sound like? Isn't that a great ad business waiting to be built? And Meta can definitely build it on the foundation that it has. And Meta has already stuffed their feeds full of plenty of ads and everyone is totally okay with it. And it is not slowed down. Even as ad load increases, everyone just uses the products more and more. So if anyone can do it, if it's already native to what that platform they're using versus I would be horrified the moment ChatGPT says that in the context of a chat, even more so, it seems like that's a good place for Meta to get involved.

11:52Did we just – I really hope this is a strategy because this is so sound that if it's not, I'm going to be disappointed. And weirdly, you're right. Meta is at a better starting point to introduce the business model here. Yeah. It's like its detriments are its benefit in some way. Yeah. I mean, everyone is completely used to advertising. And their advertising is good. Let's not take that away. They have made very, very good, relevant advertising that's so good all holiday season. All my cousins and everyone. Now everyone just openly jokes like, oh, I'm getting tracked. Ha ha. ah, oh, we were just talking about that.

12:31Now I'm getting tons of ads. We clicked on it once, but everyone's good with it. So I think go consumer, become the entire consumer agentic layer, disrupt every travel and shopping and personal, any kind of consumer website. It's an opportunity. That is probably what's going to happen. Now the question is what happens to Meta's feeds as we see more and more AI? And you remember not long ago, it seemed like they were encouraging Shrimp Jesus and other forms of AI slop in their feed and the type of conversations that they encouraged. Now, the head of Instagram, Adam Mosseri, is coming out and talking a little bit about what will make content stand out in the age of AI.

13:16This is from Business Insider. Instagram's head says the aesthetic that helped the app become popular is dead and AI helped kill it. Instagram's top executive thinks AI has made the social media sites carefully curated grid a thing of the past. Unless you're 25, under 25 and use Instagram, you probably think of the app as a feed of square photos. The aesthetic is polished, lots of makeup, skin smoothing, high contrast photography, beautiful landscapes, Mosseri wrote. That feed is dead. People largely stopped sharing personal moments to the feed years ago. He said that users kept their friends updated on their personal lives through unpolished shoe shots and unflattering candids shared via direct message.

13:57Mosseri says the growing ubiquity of AI images meant creators would have to embrace this trend of rawness and shy away from curated grids and professional style photography in favor of a more raw aesthetic. Flattering imagery is cheap to produce and boring to consume. people want content that feels real. The social media feeds are starting to fill up with synthetic everything. So what's your reaction this, Ranjan? It doesn't seem like it's exactly a brand new thing, and I don't think it really is AI related. This one hit hard, and let me explain why. So I worked in direct-to-consumer fashion and apparel for a number of years at Adore Me, And in 2021, Mark Zuckerberg starts openly saying on earnings calls that reels are the future, more lo-fi content.

14:50Basically, TikTok is a threat and that they're starting to advertise the increase in percentage of reels on your feed. Now, we all know what happened. I vividly remember getting into an argument with our creative director. And as you can imagine, a traditional fashion creative director being told that highly produced, beautiful imagery is not the future. And it's going to be lo-fi, short form video content. We all know how it played out. But this one just I was already sending this to former co-workers because I was like, I mean, we pushed hard. But it was like that idea that Instagram was one thing and then it became something completely different.

15:31Yet everyone uses it more and more. Again, Mark Zuckerberg was right. Like I have to say, there's been a few of these really big strategic moves. I don't know if old timers will remember the shift from like HTML5 to native mobile apps was like a big bet they made. I honestly think quietly taking this incredibly profitable, engaging engine of photos that are beautiful and square and turning it into a TikTok clone and just doing it and doing it successfully. I think like this is kind of the final. It happened. Now, no one can argue that this is this is what Instagram is today. that's right and the ai the ai thing is interesting because on one hand when i read what most area said i was like oh yeah um maybe he's exaggerating a bit because there is ai slop out there and uh like maybe my nice pictures from my vacation will look better than the shrimp jesus but then you're right there is there is a lot of truth to it which is that um ai is so good at taking the average of averages and it takes the average of like you know when you think about you know say, you say, create me a beautiful picture.

16:45It takes all these beautiful pictures that have been posted on the internet and gives you the average of that. So if you put what you think is like a beautiful picture on your feed, um, it's just going to look like something that people have seen over and over again, maybe even before AI, but even more so afterwards. Uh, now I'm really rethinking my, uh, I was really cooking up a nice post post vacation. And I don't know what I'm I don't know. I got to put a picture of my feet. Well, not feet. That would be weird. Did you buy a lot of socks recently? Did you buy a lot of matching socks recently?

17:19Black Friday, 2024. That was one of my great moments. I'm still using those socks. I mean, I would hope so. I bought 24 of them. Feet pics, Alex. Feet pics. No. And we're all in the prison of financial mediocrity, which we'll be talking about. It's the only way to go. So I'm starting to regret taking the conversation in this direction. Let me try to bring it back. AI slop. Please do. What I think is interesting about this is like, and you actually made the point right there. When we, everyone says AI slop, you think shrimp Jesus, you think six fingered baby, you think like crappy content. But in reality, AI slop has become to me that average of beautiful, that it's like what everything expects it to be engaging, beautiful, highly produced.

18:12And he actually made the point, too, that you said flattering imagery is cheap to produce and boring to consume. He said that like it's the and this isn't just like AI, you know, like mid journey image diffusion models. This is Apple making every photo run through 10 layers of processing and like color correction. So even that kind of AI also has contributed to this. But so to me, it's really interesting because it's like AI slop is actually beautiful, boring, generic imagery that everyone has taken versus what is what's engaging is something that's just real, raw, unfiltered. Not going to talk about feed pics, but something along those lines.

19:01Well, there was a thing. There was a line in the story where he talked about untied shoes or something like that. That's shoe shots and unflattering candids. That's what I was going for. I apologize about bringing up feet. But OK, so then you get into the question of like, is AI generated content good for these feeds? Because one thing I'll admit, actually, I have a great Facebook story. And you love my Facebook story sometimes because I don't go on it often. And actually, as we get into our Meta Ray-Ban stories, from the holidays, I was home in Lexington, Mass., where I'm from. Over Thanksgiving, my drone got caught in the top of a tree at the elementary school park area where I grew up.

19:50So I kind of considered it gone. over Christmas break, someone from my high school I had not talked to in 20 years sent me a message that someone had posted in a Lexington Mass residence group, hey, found your drone. Does anyone recognize this person? And it's a photo of me. They'd taken the SD card out and pulled the photos. So someone randomly posting to this town Facebook group, someone from my high school I I haven't talked to in 20 years, sending me that message, and I got my drone back. So even my wife was like, you have to say something nice for Mark Zuckerberg in Facebook for this one.

20:30That is a beautiful story. It is, right? I was like, this is one of those, remember the promise of social media, what it was supposed to be? I thought that was pretty amazing. But it made me go on Facebook a number of times, which I don't normally do, while I was waiting for confirmation, organizing around this drone. And it reminded me what's actually on my Facebook feed as a non-heavy user. And my God, that shrimp Jesus version of AI Slop, it's there, it's alive, it's just raging in there right now. Yeah. Well, you know, you've definitely answered a question for me, which is what happens when you lose your drone in a tree.

21:14I thought maybe you call the fire department like you would for a stranded cat. But I'm glad you didn't do that. I'm glad you use social media to recover the thing. I was I was actually wondering about it, but like as per Reddit, fire departments are not very friendly about that. And if you're devoting resources and the cost to like there's actually lots of debates, the cost to the taxpayer versus the cost of a drone. So I was like, I'm not going to not going to do that. Just hope for the best. And it worked. Well, I mean, in the defense of people that do call the fire department, I mean, they do have a truck with a ladder on the back.

21:49So, you know, it's like sort of, well, what do you expect? But anyway, let's talk briefly before we go to break about our Meta Ray-Ban stories. So I'll tell you one. So you and I are both big fans of the Meta Ray-Ban stories. It turns out that they're not very good in the cold. I had brought my pair with me to Ecuador, where I was for a good chunk of the end of December, trying to climb up Cotopaxi, which is a fairly big volcano, 19 ,350 feet. And we were a couple hundred feet underneath the summit. And I was like, all right, time to break these things out. I'm going to film this all first person on the way up.

22:25And I break them out the case, put them on, turn them on. They were fully charged when we were on our way up. And the battery was totally empty. I know you had a similar experience here, too, with your Ray-Bans. I will say I was only at 2 ,743 feet in altitude at the summit of Mount Sunnipe in New Hampshire. But I was very excited. I was going to take my son. It was the first time at the top of the mountain. Had my Ray-Bans fully charged. Even checked the battery before getting on the lift. Get up top. start taking video, get about five seconds of video, and the battery died too. So I was like pretty unhappy.

23:09I prepared for that with the Ray-Ban. So if anyone's listening and knows about cold weather effects on Ray-Ban, I mean, I guess it should be pretty, I don't know, standard. But if there's any way to prevent this and still get these shots, let us know. some praise for meta some criticism for meta uh but all together you know hey look they're they're in the arena right they're they're doing things they're building ai they have social feeds they have the smart device uh that we both tend to like okay small flaw uh but you know what but that is that is the meta picture think about this this is the first time in a long time we have devoted the first half, at least, of the show to meta and AI.

23:59So I think if nothing else, they're back in the arena, as you said, 2026. Maybe there's going to be something big. I think so. I think there's a chance for them. Absolutely. All right. So, I mean, there better be something big, given the amount of money that they're spending. But we'll talk about it through the year. All right. On the other side of this break, we're going to talk about NVIDIA's deal to license Grok, not the Elon Musk grok, the chip grok, licensed its AI technology, bring over their CEO. We'll also discuss, we are going to have a double grok, I guess, second segment. We'll discuss Elon Musk grok going bad.

24:34And then finally, the prison of financial mediocrity, which Ranjan and I are both excited to speak about. We'll do that right after this. And we're back here on Big Technology Podcast, Friday edition. All right, briefly, Ranjan, interesting story here. NVIDIA licenses, this is from the Wall Street Journal, NVIDIA licenses Grok's AI technology as demand for cutting edge chips grows. NVIDIA has forged a licensing deal with the chip startup Grok for its AI inference technology, a sign of growing demand for cutting edge AI chip. Under the deal, Grok's CEO and founder, Jonathan Ross, will be joining NVIDIA along with its president and some of the startup staff.

25:12Actually, it's 90 % of the startup staff. I guess this is still not an acquisition. Grok's language processing unit chips are built for inference, the everyday process that occurs when consumers or businesses ask trained AI models to provide answers, make predictions, or draw conclusions on new data. Ranjan, what do you think the significance is of this? It's an acquisition, right? That's what it is. Do you think this is NVIDIA recognizing that Grok, basically because it does provide inference services for AI models, could over time be a very big threat and they'd much rather have it on the inside than otherwise?

25:52Yeah, I think, and I'm glad you said acquihirization because I was trying to remember what that phrase was. But I think to start when When I was reading about this deal, like now the first thing I ever do in any of these news items is try to figure out, is it actually an acquisition? And this one was one of those funky ones, whereas Manus was a straightforward old school acquisition. They actually acquired the company. Here we have lots of quirks around who is joining, how, what is licensed, what is not, what's intellectual property, what's not. I think, as you said, this is it as it's talent and technology.

26:34It's like the entire chip. It's the language processing unit. It's if NVIDIA, they just have to stay ahead of whatever competition Google and their TPUs are coming. Like, so I think it makes sense from that standpoint. Is it like, do you think it was just to avoid antitrust that they structured in this way or? Absolutely. There's no other explanation. Yeah. I mean, it's so funny how everyone just says that, like everyone, even people who are very strongly like for antitrust enforcement, even people who are hate Lena Khan, everyone just casually is like, oh, yeah, then everyone's just doing that to end around antitrust.

27:17antitrust. But yeah, I think it makes sense to me from like a strategic standpoint in terms of the actual chip technology and having some kind of differentiated innovation versus like what you're typically working on. So. I mean, just listen to the details of the deal. So Grok was valued last at$6.9 billion in a$750 million September round. This was a$20 billion acquisition. This is from Axios. Most Grok shareholders will receive per share distributions tied to the$20 billion valuation. 85 % will be paid upfront, 10 % paid mid-year, and the remainder at the end of 2026. Around 90 % of Grok employees are said to be joining NVIDIA, and they will be paid cash for all their vested shares.

28:06Their unvested shares will be paid out at the$20 billion valuation. That's an acquisition. And everything, it's effectively everything. But in and they calling it to call it a non-exclusive licensing agreement is it's an it's an I feel like it's an insult to the intelligence of the people reading and sharing about the lawyers who came up with that one. And I mean, and again, I'm saying that in jest because I think it's actually like in reality, what also happened is here is a promising alternative technology infrastructure to NVIDIA's core business and they've taken it out. Like, does that mean they're going to now use this and compete against Google or maybe or they just took out a promising technology and for 20 billion?