Kling AI reaches 12M MAU, China's aging tech worker crisis, drama at ThinkingMachine, TBPN x CAA | Diet TBPN

0:02Kling.ai, Kling.ai, the video generation, video model, has been on a tear. There's an article in the Wall Street Journal showing some pretty staggering numbers. They hit 12 million monthly active users, and they generated more than 20 million in revenue just last year. I wanted to dig in and understand where this came from, the history of Kuaishou, the Chinese company behind it. There's a new farmer filling up the trough for everyone. Kling has hit 12 million MAUs, 20 million revenue in the last month. Now, they seem to be on the up and to the right curve. It's a pretty massive ramp. The product launched 18 months ago, but Kling is not its own startup.

0:40It is a new project from Kwaisho Technology. This is a Chinese company and the founders, I couldn't find a single. John, I'm going to blow your mind right now. Guess who worked at Kwaisho? Who? He's been on the show. Okay. I share a name with him. Wait, Connor Hayes? No. No? Jordan Schneider. Jordan Schneider worked at Kwaisho? Yes. No way. He just responded to the Daily Newsletter. Does he want to hop in? said lol i worked at kwaisho what a comeback tell jordan jordan come get on the show you're welcome come hang out if you want to i'll send him a note it'd be great to hear more of uh more of the story from him he might be the only person we can get to go on the record about kwaisho because the founders i don't think they've ever done a podcast appearance um they haven't been uh you know certainly on like the victory lap podcast american circuit uh like many founders that get a company to this scale.

1:32Kweisho is pretty old, especially for the AI. Boom. It launched in 2011 as a mobile app for creating and sharing GIFs. This was before Vine and before Musical.ly. They were sort of like a precursor, all those, just by a year or two. So you could almost say that Jordan Schneider made his money in GIFs. I think so. I think he created TikTok, basically. So there's some similar threads. So Dom Hoffman, the founder of Vine, he worked at Yahoo before founding Vine, and Alex Zhu at Musical.ly worked at eBay and Microsoft, and I think SAP as well. And so both of those companies, they had like big tech experience, then they went and founded these companies.

2:07And the founders of Kuaishou, Su Hua and Sheng Yixiao also did the same. One of them worked at Google, one of them worked at Hewlett Packard, and then they jumped into the social media boom that was going on in the early aughts, like in, or the late, I guess the late aughts, early 2010s. Is there a word for the period between 2010, the teens? The 2010s? The 10s? The 10s? I don't know. But that era, post-Facebook, there was Foursquare, Twitter, Pinterest, Snap. There were just like a new one every year was popping up. And so they jump in on this, and the mobile internet was expanding really rapidly in China at that time.

2:45There was a big boom. There were a bunch of big winners that came out of that era. Kuaishou hit 100 million DAU by 2013, so basically 18 months after they launch. And they actually pivoted in that 18 month timeline from GIFs to videos. And so once they had the new product dialed, they ramped pretty quickly. Now, there was a little bit of a slowdown between 2013 and 2019 because they didn't hit 200 million DAU until 2019. So they were sort of saturating then. But they were on a fundraising tear the whole time. They raised$350 million from Tencent in 2017. They integrated with WeChat to accelerate distribution.

3:20And the company was worth around$18 billion by 2018, just seven years into the journey. So not bad. The Quiet Show IPO was a particularly crazy moment for the company. So they raised$5.4 billion at the IPO and they were massively oversubscribed. For that, so they say, hey, we wanna raise 5 billion and 165 billion of demand shows up from the market. So retail investors just went insane. 165 billion of demand, only 5 billion for sale. And so the stock trades up 192 % at the open. And all of a sudden the company is worth$180 billion and both of the founders are deca-billionaires. Pretty sick. That didn't last, though.

3:59There was a massive sell-off. And there were a couple speed bumps that Kwaisho seemingly hit shortly after going public. So China had a big crackdown in regulation on local tech companies. ByteDance became much more dominant as a competitor and had Kwaisho in their sights. And then the Kwaisho user growth just sort of slowed down. And so there were a couple misses on, you know, DAU, MAU growth. The metrics weren't looking as good. And so the investors said, hey, we're rotating to other things. They were pulling out. And so the stock traded down 80 % within six months. So not great. But today, the market cap's around$40 billion.

4:34Like, it's still a very real business. This is all USD. I converted everything. So$40 billion market cap,$20 billion revenue,$11 billion gross profit, net profit$2.6 billion. So that's like a lot of cash flow, a lot of net income to work with. certainly enough to do some training runs, start a Neo Lab or a Video Lab or take some new bets. That's exactly what they did. 2024, they launched the first version of Kling and they did this just three months after OpenAI demoed Sora. So it was kind of like the Chinese answer to Sora. Since then, they've updated Kling 30 times and carved out a nice little place in the grid of trade-offs around model performance and cost.

5:11For cinematic footage, people often recommend VO3 still, but for Kling, it has some really strong characteristics in motion control and in physics simulations. There's a number of places where Kling's been outperforming there, and then there's a number of other sort of niche applications where Kling's really great. And at 10 cents per second, pricing has been very attractive relative to some of the other options in the market. They've certainly come out as a frontier quality model at a very affordable price. Interestingly, Khoi Sho claims that Kling is gross margin positive. And so now it's unclear.

5:49That doesn't include training. And we don't really, I mean, it's gross profit. So it doesn't include the hundred or so R &D people on the team. My question is like, what happens if they try and scale another order of magnitude? Will they be paying more for inference? Will they be able to get those chips? Because there's still this big debate over how many NVIDIA GPUs should be able to be sent over to China. It honestly gets me excited about MSL dropping a video model because a lot of the same sort of like precursor elements in terms of like the training data in the ecosystem when you think about just the raw data in Reels.

6:25It could wind up being a model that's a little bit more opinionated. Maybe it has some mid-journey sprinkled in there. Like I think that the MSL, the pressure on MSL has been intense and the initial Vibes launch was not loved. But they certainly have the compute. They have the team now. They have the data. They have all the key ingredients to a really successful video generation model launch. I can't wait to see Meta's next real launch here. They have so many advantages. They have the talent now. The pressure is immense, but they should be able to come out with something that's super competitive.

7:00Yeah. Year-to-date, Kwaisho is up 23%, so they've been on a run. And there's a lot of investor optimism about Kwaisho's potential to monetize AI. The company's AI-powered video and graphics generation capabilities will likely drive earnings over the next few years, S &P Global Ratings analysts wrote in a recent note. That could lower content creation and advertising costs and boost content creation on Kwaisho's short-form video platform. The company has the second largest independent short-form video platform after ByteDance's Douyin, which is basically the Chinese version of TikTok. And so, is this gonna be super important to the AI race in America?

7:39Probably not. We're still living in a world where you have OpenAI, Anthropic, Google, really battling it out. And then XAI is doing interesting stuff. MSL has interesting stuff. And then there's a couple of Neo labs that are maybe doing interesting stuff. The Chinese labs haven't been putting a ton of pressure, but I think this is interesting because we're actually getting firm data on how these products are monetizing. Of course, there is another interesting, just sort of like nuanced narrative that I hadn't really noticed while I was digging into Kuaishou. One is this China's aging tech workers issue, the curse of 35.

8:18They're firing unks over there. That's what's happening. They're firing the unks. You can't even be an unk. You truly can't. You truly can't. Discrimination against older employees, particularly apparent in sector where executives openly state a preference for youth. Different approach. We've had back and forth on, you know, in VC Twitter about, oh, should your team be really young and cracked or should you get the experts in there? What's, you know, obviously in America that you cannot discriminate based on age. But that hasn't stopped plenty of people from opining about what the correct mix is, whether or not you should follow those rules.

8:57But there's an article in the Financial Times that sort of has some charts here about how big tech groups in China have been downsizing in the past few years. We can go through that. But let's read through a little bit of this China's aging tech workforce. The first hint, Laobai34 received that his position at short video app Kwai Show might be at risk was when a 35-year-old colleague was sacked. I was both shocked and anxious. I realized that our situations were very similar and the same thing could soon happen to me, said Laobai, using his nickname to avoid repercussions from his former employer.

9:31Just months from his 35th birthday, the developer was dismissed, another victim of the group's reorganization known internally as Limestone. Wow, they're giving code names to age discrimination. That's insane. Kuaishou is pushing out junior workers in their mid-30s. Laobai was told his termination was part of the company's overall redundancy program. The so-called curse of 35 has long plagued workers across white-collar professions, with older staff widely perceived as being less willing to put up with long working hours because of responsibilities at home. As China's tax sector reels from Beijing's crackdown, This article is from 2024, by the way.

10:10And economic slowdown. Tens of thousands of jobs have been cut over the past several months, and older workers are seen as particularly vulnerable. Technology companies have made no secret of favoring young and unmarried workers. Ageism in the tech sector is a big problem. There's a perception that older workers don't keep up with the latest technological developments. They don't have the energy to keep up the hard work and that they're too expensive. thankfully our very own john coogan who is you're in our you're very much keeping up with the latest technological developments otherwise we wouldn't be talking about this because you didn't just read about cling you studied and you wrote about it yes um while china's labor law prohibits employers from discriminating on the grounds of attributes such as ethnicity gender and religion it does not explicitly refer to age whoa between 20 and 30 most people are full of energy, you're more willing to march forward and sacrifice yourself for the company.

11:04But once you become a parent, your body starts aging. How are you going to keep up with the 996 schedule? ByteDance, which owns the video app TikTok and e-commerce giant Pinduoduo, have some of the youngest recruits among Chinese tech companies, data suggests the average age of their workers is just 27, according to the latest figures. The average age of the worker in China is 38, according to the All-China Federation of Trade Unions. This trend has become only more entrenched with progressive waves of layoffs. And this is the chart that I wanted to show you. So big tech groups in China have been downsizing in the past few years.

11:37So this is 2021, 2022, and 2023 for Alibaba, Baidu, Kuaisho, and Tencent. And they're all downsizing across those three years. It's worth noting, I pulled up MAG-7 employee counts, and every single company in the MAG-7 has increased headcount over that same period that we just showed. Yeah, yeah. There have been like periods of layoffs, but even when Mark Zuckerberg comes out and says, hey, we're laying off a thousand people in the metaverse team and the reality labs team, it's like, well, he's adding tons of people to the AI teams and tons of people to the Instagram teams and threads teams and Facebook core and marketing and salespeople and events people.

12:18Like the whole organization is just growing so fast that one niche layoff is not really going to cause a multi-year trend. Back to the timeline. The messy drama that dealt a blow to one of AI's hottest startups. This is an exclusive in the Wall Street Journal. After a relationship with a colleague, a Thinking Machines co-founder had his role changed. Months later, he was fired after a contentious meeting. Mira Morati's meeting with her co-founder is going off the rails. Marati, the chief executive of AI Startup Thinking Machines Lab, had shown up for work on Monday last week expecting to have a one-on-one with Barrett Zoff, her chief technology officer, according to people familiar with the matter.

13:03Last summer, she had learned that Zoff was in relationship with a colleague. In the months since, she had expressed repeated concerns about his lack of productivity, according to the people. She was invited instead to an impromptu meeting with Zof, another co-founder and a third employee. The three told her they agree. They disagreed with the direction of the company and that they were considering leaving. They asked Zof to be given charge of all technical decision-making, according to the people. Maradi responded that Zof was already CTO and asked why he hadn't been doing his job for months. So they're clearly beefing.

13:37That's a very funny, it's a funny request. Like what was he doing as CTO if he didn't have full control? But I mean, obviously CEO can override all sorts of stuff and there's other people around the table and titles only mean so much. There's soft power all over the place. So two days later, Zof was fired. Within hours, all three had signed offers to rejoin OpenAI, the AI lab that they had ditched a year ago to join Marati's fledgling startup. The departures are a sign of how the heated AI race that is consuming hundreds of billions of dollars in transforming the economy is as much a battle for talent as technology.

14:13For all the high-tech advancements AI startups are sprinting to develop, They are ultimately at the mercy of the humans powering. So do you think this was a three-hour talent acquisition process? Do you think OpenAI just looked at, checked the timeline and said, hey, we got three people that are on the market. Maybe we should hire them. It seemed like they were talking before. They entered the portal and then they immediately got the cap. The trade portal, for sure. That's exactly what happened. No, I think very obvious that the conversations had been ongoing. It's very likely, I would assume, they had already agreed to terms prior to the news going out.

14:5220 people from OpenAI left during the Thinking Machines Foundation. Addressing Zof's departure to Thinking Machines employees, Maradi said there had been multiple issues with his performance, trust, and conduct, according to an internal message viewed by the Wall Street Journal. Zof said she fired him after he exposed an intent to take a job elsewhere. Thinking Machines terminated my employment only after it learned I would be leaving the company. Full stop. At no time did Thinking Machines lab cite me on any performance reasons or unethical reasons. John, I'm breaking up with you. And you're like, no.

15:24You're fired. I'm breaking up with you. It does feel like a little bit of that going on. But, yeah, there's a war for narrative here between Barrett and Mira, clearly. So he says Thinking Machines didn't cite any performance reasons or unethical conduct as part of the reason for the termination. and any suggestion otherwise is false and defamatory, Zof said in a statement to the journal. The exits, coupled with fellow co-founder Andrew Tullock's decampment to Meta last fall, leave Thinking Machines with just three of its original six founders. Marati's issues with Zof started over the summer when she began to suspect he was having a relationship with a colleague with whom he had lobbied to bring over from OpenAI, according to people familiar with the situation.

16:08In responses to questions from the journal, Zoff said that many people at Thinking Machines wanted to hire the woman, including Marati. At the time, Marati was in the process of raising one of the largest seed rounds in Silicon Valley history. The company ultimately raised$2 billion at a$12 billion valuation when she confronted Zoff about the possibility of an undisclosed relationship with a female employee who was junior to him at the company but did not report to him. He initially denied it, according to people familiar with the situation. By June, however, both Zof and the woman had told Maradi about the relationship, which had begun when they were colleagues at OpenAI.

16:42The woman then left the company and returned to OpenAI, which is sort of a wrinkle in this. Zof told his boss that he had been manipulated by the woman into a relationship, according to people familiar with the matter. Shortly after that conversation, he took a break from work. Okay, I got to, I generally think that the - Mira, she told me it was cuffing season. I didn't know what it meant. So I just said, okay, I got cuffed. Is that what happened? Yeah. I mean, I think just saying like, I was manipulated into becoming the significant other, you know, you got to take a little responsibility. Everyone is assuming that this is a romantic relationship.

17:22They never reported that this is a romantic relationship. They just said that he had a relationship with someone else and it was undisclosed. And I think that more companies need to be clear about the rules around disclosures of relationships. If you and Tyler start an e-sports team, for example, like you would have a relationship, you would be teammates on playing Call of Duty. And if you didn't disclose that to me, I would be, I would feel left out. I'd be like, why don't I, why am I not on the team? If I find out that Tyler and Scott are going off and drinking a bunch of athletic beers, athletic brews every night, you know, without me, that's a relationship to bros.

18:00And it wasn't disclosed. And it wasn't disclosed. And you might be very angry. I might be angry. That they didn't disclose. So I think relationship disclosures need to go beyond romantic relationship disclosures. If you're just broing down with people, that needs to be disclosed. That actually is kind of a real thing. Sometimes I'm catching up Monday after the weekend. Somebody's like, oh, yeah, I was hanging with so-and-so on Saturday. I'm like, wait. Where was my invite? You guys hang out on Saturday? What? You didn't disclose that relationship? You guys went to Korean barbecue? Yeah. Oh, so you have a relationship where you go to Korean barbecue together.

18:33Okay. And you just didn't think to tell the rest of the company that. A K-BBQ. Yeah. I was going to say, I think there's, you know, I like to imagine there's some kind of like Shakespearean tragedy here, right? Yes. Where it's like the Montagues and the Capulets, that's open AI and thing machine. Oh, yes. And it's these forbidden lovers. Yes. They can't be together. But now they are. They're united because everyone is in open AI. Even though, I mean, if you go to the journal and you say, I've been manipulated by that woman, who's now my colleague because they both work at OpenAI now, right? So that's odd.

19:03So the woman left the company, Thinking Machines, and returned to OpenAI. But they were both at OpenAI. They both went to Thinking Machines. And then they both went back to OpenAI. I still think this is a company that would have a pretty meaningful valuation in an acqui-hire context, right? Like if Amazon were to come and buy them, how much would Amazon pay to get John Schulman and Miramirati at Amazon? Should be able to clear the pref stack for sure. Yeah, by a long shot, you would imagine. I wanted to look at this photo of Davos. It's AI generated, but it captures a good vibe. A whole bunch of private planes parked there.

19:41That's not real, but I like that Kevin Kwok contextualizes it anyway, saying this looks like Hoth Echo Base. The Demis and Dario are apparently just bro-ing out in Davos. I love it. Demis Hussabas is CEO of Google DeepMind at Davos. Quote, when told that Dario Amadei was here earlier today, his face lit up. Two minutes later, he was talking about a CERN-like collaboration for AI once the labs are close to AGI. They're just stoked on each other. I'm on pretty good terms with pretty much all of the other leaders at the leading labs. I love Ilya, and we're good friends. You love it. Just some positivity hanging out at Davos.

20:22Demis said he thinks entry-level jobs and internships might fall away due to AI. Sees consistency as the biggest problem with current agents. Advice for college graduates, get incredibly proficient with these new tools, AI agents. Yeah, we were debating this earlier. Like the role of a junior engineer is going away, but is it possible that it just changes? Like what makes a senior engineer that much better with an AI tool that they don't need anyone else who can potentially use AI tools as well? How much of this is something where, I mean, to go back to like the Chinese young hiring boom, there's an element where I'm sure there's companies where you have older software engineers who are more resistant to adopt AI technology and you'd actually be better off with a young junior person.

21:19Here's a tough question for you, Tyler. What are you fantastic at that AI is bad at? Walking around being physically embodied. Embodied. That's the last thing. So maybe that's like for China, it's like maybe the, you know, Silicon Valley has kind of found like pederasty from first principles. Maybe, you know, it's just a similar thing over there. Right. There's actually a lot of value. I think like so generally, like I'm kind of I've always gone between like, is our jobs actually going to go away? Yeah. Or are they just going to get like more fake? Are they just going to be no junior jobs in like five years?

21:55Or are they just going to be like, you know, you're in the office because like it's like people like having other people in the office and stuff like this. I think I'm probably more in the camp of like, you just don't need to hire new people. First of all, it's a pleasure to be on the gloom and doom panel. We're not gloom. The 1920s were an extraordinary period. As I said at the beginning, it's not preordained. I just love the end note of the 1920s, which was, of course, the Great Depression, Andrew. So let's take a step back and talk about where we are right here, right now. The area of recklessness is the spending of governments around the world, who are all, with little exception, all spending well beyond their means.

22:42That's the recklessness of this moment in history. This is not a parallel to the 1920s in terms of the recklessness of the private capital markets. It's a story of the recklessness of government spending. Within the private sector, there's a huge question as to where AI will take us. And I was carefully taking notes and listening to what Larry has to say or to what Madame Lagarde has to say, because this is one of the big issues of our moment. Will AI create the productivity acceleration that is honestly hoped for in Washington and and in the halls of government around the world as a way to overcome the profit and spending that we're currently engaged in.

23:30Like the world needs a savior. And the hope is that AI is the savior that we need for productivity. And the challenge with this is it may or may not be. We just don't know yet. Now there's a tremendous amount of hype around AI. And in some sense, the large AI companies need to create that hype to raise the tens or actually hundreds of billions of dollars of investment that are going into the field. You wouldn't be able to raise hundreds of billions of dollars. We'll spend, and Larry can probably correct me on this, but roughly$600 billion this year in CapEx for data centers in the United States.

24:13I think it's going to be larger. But does that mean that it's getting hyped up too much, or it's just the hype is required as a sales mechanism? Go ahead. Larry's backing you up, Tyler. All right, let's play the next video, which was the one I was originally talking about, which is talking about job loss. What AI has done is it has re-empowered the head of technology in every business in the United States. and it has pushed budgetary resources into the hands of the chief technology officer. So what you're seeing across American business is actually the impact of American businesses spending more on digitization writ large, of which AI is just one component.

25:03You know, as Dario Amadei says, you know, half of all entry-level white-collar jobs will, you know, be gone in the next five years. Is it that kind of change? Look, you know, put yourself in his shoes. How much money does the AI community need to raise over the next five years? So he's wrong about this, right? Demis doesn't need to raise. Demis that are spending the United States this year over half a trillion dollars. Google's doing this with cash flow. Like over$500 billion. Like that's true for every other lab lead except for Demis. Like you would expect Demis to be like, no. Yeah, but the question was about Dario.

25:33And the question is, where will AI land in productivity gains at the end of the day? In certain areas, we know it's going to be profound, whether it's call centers, whether it's helping to improve the productivity of software engineers. But in a number of white collar jobs, there was a recent Harvard paper on this. They called it AI work slop, that it looks good. But if you sort of peel back the onion, the substance isn't there. This is, I mean, this was the, probably the story that got reported on the least out of Davos over the last 24 hours, but is arguably the most important, the most expensive car in the world.