news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

The CEO and several partners of Y Combinator said in their latest blog that there is indeed hype about AI at present, but the technical foundation of AI is more solid. In the long run, artificial intelligence technology will bring sustainable growth and value to enterprises, but it will still take some time to see its real impact.

Y Combinator (YC) is a world-renowned startup incubator and seed-stage investment company. It is known for its unique model and success in Silicon Valley and is known as the "university of startups."

YC mainly invests in very early-stage startups, even those with vague ideas. The company invests in hundreds of companies each year, but the amount of each investment is relatively small.

On the 23rd, YC CEO Garry Tan, together with three partners Jared Friedman, Harj Taggar and Diana Hu, shared their views on the recent AI boom in the latest podcast.

They believe that there is indeed hype in the current artificial intelligence, but unlike the cryptocurrency hype, the technical foundation of AI is more solid, and it will take some time for the development of AI to see its real impact.

The key points of the dialogue are summarized as follows:

○ The current AI boom is over-hyped, similar to the previous Internet bubble. Although AI has made great progress in technology, its value and business model in practical applications are still uncertain.

If you think about what's driving Mag 7's recent gains, it's essentially all AI hype.

○The competitive landscape of large language models has changed significantly over the past year, with OpenAI once the only major player to now have multiple competitive models, such as Claude 3.5 and Llama.

○ There is still a lot of uncertainty about the distribution of value in the AI ​​value chain. The development of new technologies takes time to verify and mature. Just as the popularity of smartphones has led to the rise of companies such as Doordash and Instacart, the development of AI will also take some time to see its real impact.

○ Compared with the cryptocurrency bubble, AI hype is also characterized by overvaluation and strong speculation. However, AI has a wider range of application scenarios and a more solid technical foundation.

○Despite hype and bubbles in the short term, AI technology will bring sustainable growth and value to businesses in the long run.

AI hype leverages Mag 7 revenue

Garry Tan：

Some of the things people are saying about AI right now is that it's a hype cycle and nobody's going to make money from it. You look at how much money is being poured into Nvidia and data centers. The numbers show that there's no way you can make money in this space.

It's like the dot-com bubble and the rise and fall of cryptocurrencies. You know, doomsayers, decelerationists, and so on. When I think about market manias, one of the memes that comes to my mind is a really hilarious cartoon. It's famous. I have a stock that can really do well, sell, do well, sell. And then the next frame is, this is crazy. I can't stand it. Bye, bye, bye.

It’s like a frenzy has taken over the market. We’re all too familiar with the Gartner hype cycle, or our own version of it, the life cycle of a startup, the ups and downs of false hope and long periods of sadness before finally arriving at the promised land.

Where are we now? You know, we saw a lot of people who were just starting their careers ask us at Startup School a few weeks ago, should I be working in AI right now? That’s the craziest question to me.

Diana Hu：

Where does this fear come from? All the founders looking for ideas look at this, is it real or just hype?

Garry Tan：

When you were just starting your career, you probably read about the hype cycle in the past, like the $200,000 parties that were thrown every Friday or Saturday night that were just free alcohol and crazy revelry in San Francisco in 1999, 1998, and you read about that and heard, oh, all those companies are dead, and you were a little worried, is this where we are now?

Jared Friedman：

Yeah, I have to say, it's been an astonishing experience for me because we live in Silicon Valley and with our friends and colleagues, people have basically been talking about AI and feeling very strongly that it's an incredible moment in history.

But a few months ago, when we went to Cambridge University, I met with university students who were working on startups.Only a few of them are researching AI, and very few are really thinking about AI. They are just going to do the same startups that we have seen college students start for 20 years. I was surprised at how big the gap is between these two worlds.

Harj Taggar：

I think what's unusual or novel about this current hype cycle is that I feel like the startup world always goes through periods like this where the idea is hot and you feel like everyone is like that meme you just talked about, people start to get into a particular idea type and then suddenly everyone is working on something like Uber for X or their social mobile native app. However, this time around, it's happening with both AI and the startup world.

But if you look at the public stock market, AI is having a huge impact there as well. For example, the stock market is up this year, but all of the gains have come from big tech companies.

Garry Tan ：

Like Mag 7

Harj Taggar：

Yeah, Mag 7. I believe that it's never been so constrained in history. If you think about what's driving all the gains in these seven big tech companies, it's essentially all AI hype. So I think like I've never seen these two things so synchronized before, the startup stuff and the YC batches are trending more and more towards 100% AI. And then the public market returns are also essentially 100% AI driven, which is why I think this is captured. Everyone has some imagination, but also fear that this is unsustainable and everything will suddenly collapse at some point.

Competition in large language models is growing, and new players are emerging

Jared Friedman：

Will it suddenly collapse at some point?

Diana Hu：

There are a lot of different articles online saying that there is overinvestment in AI. Even in the public markets, a lot of experts are worried about, what are you going to do with all of these huge investments in AI chips? I mean, Nvidia became the most valuable company in the world. Who would have thought that? And then there's the underlying infrastructure. A lot of people are thinking, well, you invested in all this infrastructure, so something needs to happen to pay dividends on it, right? It's kind of like the early railroad analogy. So you lay the roads, and the trains will come?

Garry Tan：

I mean, I feel like there's a very extreme version of what's happening to all of us right now, like about a year ago,It seems that only a few base models will stand out, and there is a threat of becoming not just AGI, but ASI, artificial intelligence super intelligence.You know, there's this idea of, oh no, what if there's actually no opportunity for other people? It could destroy society.

You know, Claude 3.5 sonnet is very competitive. You actually have choices, and we're moving to another moment where we're thinking, well, how does value accrue to the underlying model rather than the hosting company? I think they're the biggest winners. And then we hope that the software companies themselves, both startups and established companies, will also benefit from these underlying models.

Harj Taggar：

It feels like because everything is moving so fast, it's easy to underestimate the importance of this. If we go back to early 2023, we start to see the first products of AI ideas, just a few months after Chat GPT was launched. The Chat GPT rapper meme is really popular, right? Everyone is talking about how these startups are going to be crushed because GPT and OpenAI are going to own everything. Fast forward a year and a half later, it's clear that's not going to be the case. There are multiple models. You just got the first real open source model from Facebook now, which we would never have predicted.

Diana Hu：

Yeah, that's pretty cool, right? Who would have thought that the best model would be open source since it's 6 months to a year behind OpenAI, right?

Harj Taggar：

I remember before this latest Llama release, all four of us were talking about it, like if it was just the gap between the pioneering model and open source, with every new release, open source could catch up in X months. If we could make X smaller and smaller, that would be really exciting.But it’s basically on the same level, and I didn’t expect any of us would see this a month ago.

Garry Tan：

We're here. I mean, there's that chart, the open source model is growing exponentially, whereas the pioneer model looks like it's on an S curve.

Diana Hu：

What you're saying about the difference between using four models in the current competition versus six months to a year ago is very different. I do remember the ballpark numbers for the previous batch. I would say about 90% of people were using the OpenAI model because that was the best and, simply put, it worked.

Now, we did an informal survey, Sonnet Claude 3.5, and there were actually a lot of people using it, but then it was only 1 or 2 companies, and now there are dozens using it. Llama is also a lot more. So I think we're seeing a decrease in the use of the OpenAI model as all of these have become competitive.

AI has huge potential in application

Jared Friedman：

I think you make a good point about the lack of clarity around value accumulation. Even if you believe that AI is going to be huge and it's going to create trillions of dollars of value, there's still a lot of uncertainty about who's going to get the lion's share of it. Is it the GPU manufacturers? Is it the hosting providers? Is it the model developers? Is it the application developers? Which parts get commoditized and which parts become very valuable?

It reminds me of web 1.0 and web 2.0, where you had the same phenomenon where a lot of people were very bullish on the whole space, but it wasn't clear which space you wanted to live in. Even if you go back to web 1.0, there was a huge hype about owning a browser. For a long time, people believed that the way to get super rich on the internet was to own an internet browser. Because that was the entry point to the internet, right? And Netscape was worth, I don't know, billions of dollars at the time. It turns out that was not the place to play. But that wasn't obvious for a few years.

Harj Tagga：

I think time is a big factor. Because if you think about it, in our own world, in our careers, some of the biggest companies that YC has funded are probably Doordash and Instacart, and they were driven by all of us having smartphones and wanting to do things on our phones. Uber is obviously another big company, but those were launched, or at least the first versions of those came out about 4 years or so after the iPhone was launched. These things take a while, and then you really know which ideas are going to catch on and where all the value is going to go.

Garry Tan：

I think a very important factor is that there are all these other parts of the value chain.Obviously, there’s the base model, the hosting providers, the chip makers, and then there’s the startups that we fund, which is the application layer.

The important thing to note is that you don't need $100 million to start an application layer company. You just need you, sometimes only you, and usually a co-founder. Then, if both of you know how to code, you can take these super powerful features that are basically available now, enter another market, and you can create a product that solves a real problem and get funding from people who are willing to pay you forever if you can solve their problem with the technology at hand. You can do all this from your desk or the computer you are watching this video on. You don't need any permissions, except a working internet connection and your laptop.

Jared Friedman：

That’s exactly the story of Instacart and Doordash, right? These are application layer companies that are powered by mobile phone technology, but they don’t have to make the phones themselves.

Garry Tan：

Every other part of it is like, yeah, maybe you need $50, $1 million to get your base model company off the ground. Maybe you need that much for fabs or hosting or all these other things. But even then, I guess that’s not entirely true. Like, it’s just a lot harder.

AI technology has a more solid foundation than the cryptocurrency bubble

Harj Taggar：

I also think if we go back to this question like are we in an AI hype cycle and try to like, if we define it more precisely, nobody is saying AI doesn't have any value. It clearly does have value. I think usually when people talk about cycles, what they're reacting to is they see things going up in price very quickly.Whether it's public market stocks like Nvidia, which you mentioned, or whether it's in the startup world, you see like, you know, companies reaching $1 billion valuations within six to twelve months of founding.

Diana Hu ：

This has happened with these very famous AI research teams. I used to work at Deep Mind or OpenAI. They leave and start, there's no product market fit, and six months later, they have this huge valuation.

Garry Tan：

I don't know, right. Yeah, it's kind of like Professor Coin during the crypto boom. If you had experience with distributed systems, and you know, you could just walk into the first crypto VC you met, and you'd walk out with a market cap of $1 billion to $5 billion, like, without a line of code, without even a white paper, without anything.

Harj Tagga：

Crypto is the perfect example. Like that was the last time. It's not even that long ago, it's been two years. That was the last time it felt like there was a hype cycle. It was specifically defined as, hey, to us, these companies are growing in value at what seems to be an unsustainably fast pace. It's all these crypto companies that are doing token launches or even just raising equity rounds and seeing valuations double and triple every three to six months. So I think there's definitely some scar tissue there.

On the other hand, it's also just the startup world. Like, I always remember people talking about Stripe because when it first launched, Stripe raised this massive round from Sequoia at a valuation of like $100 million, and I don't think they even released it publicly. It was all about belief in the idea of ​​the founders, in the market. So part of it was investors investing in the hope of making money.The way they make money is that whatever price they pay today is a profit on how much the company will grow in the future.

Diana Hu：

But there's a big distinction, maybe a subtle distinction, about how you value technology companies versus something that's more like asset speculation, right? Because in the cryptocurrency world, there's a lot of that. So we can talk about it versus real technology.

Garry Tan ：

I wouldn't say it's pure speculation. There's probably some irrational market valuations of billions of dollars that people are getting, but if you're a professor of distributed systems and you have a network somewhere, you need to actually create a cryptocurrency that people trust and build applications on, that's reasonable. That's actually a group of people who can do it. And then the money closes the door behind you. I think that's why you have things like Cognition Labs and Devin, you have things like Harvey, it's the market trying to figure out where the talent of the really smart people is.

Garry Tan：

I think this is actually on this YC channel. So it's not completely irrational, but it's laughable in a way.

Harj Taggar：

I think that's a fair view of the recent past for cryptocurrencies. I think during that time, you could divide it into speculative assets,This is essentially people who are focused on launching these tokens very quickly and trying to increase the price of the token. But there are also some legitimate and very fierce technical teams trying to develop new protocols and trying to turn existing services into decentralized versions. I think the investor mentality is pretty much what you said.

In fact, we should not be against smart technologists trying to solve very hard technical problems. There are many such problems in crypto. So let’s invest before product-market fit because, these people will attract the next smart people in the future. If there is such a thing, these people will figure it out.

Jared Friedman：

So, I think the cryptocurrency analogy is really good because when I talk to a lot of great students at Harvard and MIT, the consistent thing we hear is that a lot of them feel burned by the crypto hype cycle, whether it's personally or friends, or they're old enough to have gone through the crypto hype cycle of 2021, 2022. A lot of people are a little skeptical right now about AI and anything that's new and hot.

So YouTube is a founding investor in Coinbase and one of the most successful cryptocurrency investors in the world. How would you compare what's happening with cryptocurrency in 2021 to what's happening with AI right now? What are the similarities? What are the differences?

Garry Tan ：

I think that Coinbase is not a cryptocurrency per se. It's the enabling technology that's needed to enable it. I think they're still on that journey. I mean, Brian Armstrong and Yield, you just started talking this week about, you know, we're literally working with or with every financial institution in the world to integrate blockchain technology into the actual norm of finance. I think that's the promise that got me excited when I first met Brian, when I was an anti-fraud engineer from Airbnb. Yeah, like.

Harj Taggar：

It's essentially a marketplace. I think it doesn't look much different than any other startup, where it's like, oh, hey, there's this thing where people want to buy and sell from each other. There's no great marketplace to do that. So why don't we build a trusted marketplace?

Garry Tan：

It was fringe and niche. I don't think people thought at the time that it would go that route.

Harj Taggar ：

The unknown with Coinbase was how big the market was, but it very clearly had product-market fit with people who cared about it early on. Like there were definitely people who wanted to buy and sell Bitcoin, and there needed to be better ways to do it. So there was never any doubt that Coinbase had utility in the sense that: Hey, there's a group of people who want to do something, and Coinbase is making it easier for them to do it.

I think the difference with AI right now is the sniff test with cryptocurrencies. So when you look at products, especially a lot of web 3 products, for most people, never really pass the sniff test, like I don't understand why I would use it.But when you look at AI products, obviously, being able to summarize a 50-page PDF market analysis report and actually extract three key points is obviously something that someone is willing to pay for.

Many AI startups are overvalued

Garry Tan：

I mean, there's a company in my YC group right now where they can do accounts receivable. They went from a team of 12 to one person doing accounts receivable and the other 11 people can do everything else in finance. It's like as tangible as you can get. Like you're actually doing something better than humans delivering butter, and we're replacing them with robots delivering butter.

It's actually kind of like a Rick and Morty joke, but on the other hand, it's actually pretty serious. Like how good the software is at doing that thing, you know, it's not actually the best thing a human can do, which is coordinating emails and bank records to make the required payments. It's not like the kind of work that inspires you, you know, opens up your creativity, right?

It's one of those things that can be painful sometimes, you know, we can give people other things to do. That's the kind of thing that we're seeing right now.

Harj Taggar ：

We actually saw that, right?

Diana Hu：

I mean, one of the cool stats that we mentioned earlier in an earlier episode is that if we totaled up all the revenue that companies had when they applied to YC, the total revenue was $6 million. By the end of the batch, about 3 or 4 months later, if we add up all the revenue that they had grown to, it was $20 million, which is huge in just three or four months. It managed to outperform a lot of the still-ambitious advice that we give, like, try to grow your company 20% a month.

That's still exponential growth. If you use that benchmark, 20% growth per month from six months to $6 million for three to four months is only about $12 million. So, we're seeing real revenue that's actually accumulating.That value that you talked about when companies find a good idea, their customers recognize it. I mean, customers are discerning, they pay for it, and there is a value aspect.

Garry Tan：

And then I think the tricky part is it can't just be the first renewal or the second renewal, like you actually have to get through all the renewals. Like the only way the EV company is is the discounted cash flows into the future. So that means retention has to be every customer that you get, you better have that customer forever. Only then can you build a business of any size.

Harj Taggar ：

This also seems to get to the heart of the matter, are we in a hype cycle debate? Because I feel like all of us at YC are seeing examples of companies on the ground floor that are growing ARR much faster than what we've seen in a batch of companies.

Even within six or twelve months of a batch of companies going public, revenue growth is impressive. Gustav has been talking about Leia, which just completed its Series A round, and they are doing legal AI to automate legal work.

Diana Hu：

There's a company that I worked with last year that's probably going to be $10 million by the end of this year. And that's only 12 months after they found this.

Garry Tan：

Idea. Man, after 10x, you can IPO.

Harj Taggar：

I think it's very different than when Jared and I moved to San Francisco, like 2007, right? Because back then everybody was chasing something. You obviously had this ridiculous thing too. What everybody was chasing was page views. Yeah, page views, and then active users, just, I know registered accounts, whatever that was.

Garry Tan：

But yes, a lot of indicators. Yes, you say things that make you feel good. But.

Jared Friedman：

Well, I want to jump to a point that Hash mentioned earlier, which is, if you define a hype cycle as a situation where an asset is massively overvalued. That might be the case right now. Nvidia is the most valuable company in the world right now. I don't know if it should be like that, maybe Nvidia is overvalued, the TPU will start working, like, you know, all of these things could change. There are startups that have raised over a billion dollars in valuations.

In hindsight, some of these deals will probably look really stupid. The good news for our business is that we don't care. Like, you know, if we were a public market investor, you'd probably be really concerned about whether Nvidia is overvalued. But given the space we play in, it actually doesn't matter.

Garry Tan:

I think the best example is the founder of Amazon.

Jared Friedman:

Yes, it's really good. Even if some things are a little overrated, that's true.

Garry Tan:

I think it's actually free money for the entire ecosystem. For example, Nvidia being overvalued means that if they need to raise money, they can raise it very cheaply. This is equivalent to more capital,It helps them drive future development faster. And then basically everybody else in the ecosystem benefits.

Harj Taggar:

The big difference between us and founders and public companies is that public companies have to deliver quarter after quarter. If they miss an earnings report, the stock price plummets and employees start to get upset. It's a huge distraction. But for YC, we invest in people's ideas, and we don't expect to know in 10-plus years whether something works, right? It's the same for founders, from a 10-year perspective, it doesn't matter if it's overvalued today. You only care about the direction, is it more valuable 10 years from now than it is today?

Garry Tan:

I think these businesses can actually be healthier, but they’re not always. I think professorcoins that raised hundreds of millions of dollars and never launched, or some of the AI ​​darlings today that are probably the biggest VC funds in the world, have $100 million or $200 million or $500 million on their balance sheet but absolutely no revenue.

It's like looking up at Mount Everest and saying, how am I ever going to get up there? That's in stark contrast to a lot of the YC companies we see, which may have only raised $1 or $2 million from Demo Day, but they start hitting milestones of $5 or $10 million. In fact, because they're profitable from day one, or they're profitable relatively quickly, their bank accounts are growing. They'll look down and say, I raised a seed round, I don't have a board, I don't have to sell my company. So why don't I just structure the company the way I want to?

I think maybe 10 years ago, there were no examples like this in the YC world. For example, Weebly only raised a seed round and then sold to Square for a huge amount of money. They were able to do that because they were profitable in the social software era. Even today, Zapier is probably one of the most dominant pure software companies that has only raised a seed round. They are generating hundreds of millions of dollars in revenue. It's kind of a crazy time for them.

Jared Friedman:

And they are making huge bets on artificial intelligence.

Garry Tan:

So it's a very interesting time. Do you really need a giant round? In fact, a giant round can be like a boulder around your neck, and you can never get over it.

Diana Hu:

What’s fascinating is that all of these companies that we mentioned are profitable or breakeven, they don’t need to raise money, and they have revenues rolling in and growing exponentially, and they don’t need to do that. So you probably don’t hear a lot about them because they’re not going public and trying to raise huge investment rounds because they’re really playing the long game.

AIGC has great commercial potential

Diana Hu:

If I think the categories of things that we can start to see now that we have more data, I think in every area of ​​augmented reality, AI, we're seeing early signs that things are working. Even for some of the things that people think are overhyped, like using generative AI to generate images, I think it's hard to imagine, is this just for toys or art or entertainment? But there is one company, Photo Room, that is actually making real revenue. I think they recently announced that they were valued at $500 million, and they are generating images for e-commerce.

So when you get into a vertical, they come up with something.They made generative AI very profitable because it was very hard for brands to take product images and product placement, all of that stuff. And they did it very cheaply.You don't need a team of photographers or editors, right? That's one example that I can think of. Another category that we talk about a lot is this kind of AI agent workflow, right? You're replacing a team of operations people. You mentioned, yes, like Green Light.

Harj Taggar:

Green Light is great. Another example is Permit Flow, which is a company that uses artificial intelligence to fill out the permit process. If we try to play devil's advocate and argue against these things, there are two ways of attacking them. First, you say, these things are being automated, but they haven't completely gotten the machines out of human control. So in order to unlock these valuations and get to such high values, they have to get the machines out of human control.

The second view is that businesses will never trust this stuff. So you won't sign six-figure or million-dollar contracts with large 14,500 companies to rely entirely on artificial intelligence.

Jared Friedman:

Oh, and I think the second harsh point you made is that they're going to be commoditized. They're like GPT wrappers, and all the real value is going to be attributed to the underlying model, and there's going to be 100 licensing processes. So how are they going to capture any real value?

Harj Taggar:

What about that view? I just think, I think that's really the biggest criticism. That was the main attack about a year ago. Maybe I'm just saying that because I believe it so much, but all the things we talked about, like multiple models, open source, licensing processes to be the winner in your space, from what I understand, it's all about how to sell, getting the details of the user interface right, having all these little details to make the product perfect.

Jared Friedman:

I agree with you and I'm considering the others' points. Yes, it seems more likely that the opposite is true, that the value is absorbed quickly by the licensing process.

Diana Hu:

Even on a technical level it's not just a wrapper, I know they put a lot of work into really fine tuning it for specific areas.These are areas where private data has a lot of power, like licenses, banking data, which can't be done and copied. They actually did some really clever things, not just a wrapper. They did a lot of really well-thought-out work.

Harj Taggar:

That's a great point. I think Mark Zuckerberg said this, maybe a week or two ago, he said that even if all model development progress stopped today, there would still be five years of innovation, like application layers of point solutions built on top of the model.

Diana Hu:

Even on the compiler side, right? Even with the Github compiler, the most notable example is obviously the fastest growing product at Github. It was recently reported to be accounting for 40% of revenue growth. Rumor has it that they're making hundreds of millions of dollars in revenue. That's still to be verified, but it's somewhere around that, which is pretty impressive given the time it was released. Right?

Harj Taggar:

Another anecdote that proves this point is that in order to really be as valuable as everyone says, you need to completely take the machine out of human control. Based on the anecdotes I've heard from some startups, they're building it into the workflow so that the AI ​​can do the work, but you can review it through the UI or have a human do the review process. But increasingly, customers don't even use this feature anymore. So they just...

Diana Hu:

I've heard that, too. There's another company that's basically replaced a lot of call centers, and they were handling hundreds of thousands of calls. Same thing, they fired up the entire center. This large enterprise is using this startup, and it's 20 times, 100 times faster, and cheaper. Same thing.

Harj Taggar:

Going back to your point, even though it's not perfect right now and there won't be another 10x frontier model in the short term,It is very possible that all the work like fine tuning, collecting data, private data repositories, just to squeeze out and get to the right level of quality, big enterprises will start spending millions or even tens of millions of dollars on all these solutions.

Diana Hu:

There's actually a fourth category that I think is working. While there are a lot of tools out there that can be built, evaluated, fine-tuned, etc., several of our companies are actually having success fine-tuning, especially for enterprises, which are using their own private data, and even tools are valuable if you want to skip specific apps.

Jared Friedman:

I also think there are a lot of places where LMS can be applied that people haven't even thought of yet. Diane and I have a company on our team that just two weeks ago was at this Industry Review session, and they had an epiphany moment where they realized that there was an entire industry that was a perfect application for LMS, and no one in technology even knew or knew it existed. So no one had tried it. They just, you know, there was a multi-billion dollar opportunity in an obscure corner of the world like racing that was a perfect fit for this technology. And as you said, I think there are still years to unlock the current technology.

Garry Tan:

I think ultimately that's what's playing out here. This has played out in every hype cycle, boom-bust cycle, not just in technology but in the world in general, where there's a frenzy that starts out like the world is going to change, and then it's hard to really understand what's going on, and a lot of it is hearsay or media coverage.

This is what X said, you know, on Product Hunt, there's kind of, like, there's a fog of war around it. And what that leads to is a popularity contest. It's basically short-term, all businesses are voting, influenced by the voting machines, right? Like it hasn't really happened yet. It's happening so fast that humans are meaning-making machines, and we need time to understand.

So in the short term, we get scammed by fast-talking scammers. We get scammed by fancy credentials or, you know, I worked at a company like this. They're the fast-talking Stanford dropouts of the world who have used scams to rip off millions of dollars from very smart investors, right? You see a lot of this, and it's just the fluke and the craziness of the voting machines, right? It's not like people are trying to put their money into scams. It's more just due to pure social effects that we can't make sense of the world fast enough. So basically, the mob is wrong.

On the other hand, in the long run, the value of every company is its future discounted cash flows.You know, you need your customers to actually solve a problem. People pay them, and then, you know, your customers are with you forever. That's why when you look at Google or Yuan or any of the Big Seven, those are the most valuable companies in the world, because people think, you know, these companies are probably going to make money forever, right? And there's security in the process.

Garry Tan:

That's what people believe now, and then ultimately the public markets, you know, they're also completely crazy voting machines in themselves, but they eventually become weighing machines. Eventually you have to make money. Eventually, you have to have customers. At that point, you need to create something that actually has weight and weight and actually works. That's all we're going to have to say today. I'll see you next time.