Instinct 融资10亿美元、估值100亿美元，Meta 发布 Muse | Miro 曾获175亿美元估值后以13.6亿美元出售

If you are creating something you can control, you might have to stop development. If the federal government believes that someone in downtown San Francisco is developing technology with only a 10% chance of destroying the world within the next 10 years, they will deploy a special task force to eliminate everyone inside and shut down the facility. So, the first topic this week is “Flow of the Frontier.” We discuss what this means for infrastructure, energy, and the largest frontier model providers. Next, we take a look at Town, Instinct, and Mark Zuckerberg and Muse. What will happen next? Next is Bending Spoons. It is continuing its acquisition spree by acquiring Miro for $1.35 billion. Finally, Mistral attracted 3 billion euros in investment, the largest in European history. In addition to this, listen to more stories in the colorful discussions hosted by Lori and Jason. The market moved 0.1% at the news that there is a 10% chance of extinction. Capitalism took that risk and said, “It will be fine.” $2 billion is just an initial investment round today.

1. Should Frontier AI Labs Be Forced to Slow Down?

I was debating where to start, but honestly, I felt I had no choice but to talk about the most important thing first. Dario said, “We need to control the speed of cutting-edge technology,” and Sam Altman agreed. Elon Musk also agreed that it is important to establish an external regulatory body that can slow down and regulate the technological development of model providers. What do you think about this issue?

To be honest, this project received almost unanimous criticism. I am closer to thinking that such a hostile reaction is natural. What on earth is the problem you are trying to solve? Dario, as you would know if you read the memo, the problem you defined was quite clear: cyberattacks, economic problems, and the loss of control over artificial intelligence.

It might look complicated, but it was actually a fairly controlled problem. The person who quit the company and started this ridiculous project was someone named Coxton or something, and, on top of that, an internal employee at Anthropic chimed in, saying, “It seems there is about a 10% chance that humanity will go extinct within 10 years.” Isn’t this a completely different story? So I think we need to address that issue first. I think that claim makes no sense.

To be honest, that is exactly why I say this is complicated. Whenever Dario posts a comment, many people tweet, “If you’re going to blow up the world, you should stop.” That is correct. Actually, he didn’t say he would blow up the world. They didn’t mention P(doom), either. But I’ll throw in a few words first, just like everyone else.

If the federal government believed that someone in downtown San Francisco was developing technology to blow up the world, and there was even a 10% chance that it would do so within the next 10 years, do you think they would deploy a special forces unit to kill everyone inside and shut down the facility? What I mean is, regarding AI, it is like saying, “We are building a nuclear reactor. It is completely safe right now, but there is a 10% chance that a problem will occur within the next 5 years and destroy the world.”

Try changing it to this. If someone says this, how many seconds would it take for the entire U.S. military to step in and immediately shut down the system? This 10% probability seems like it foreshadows a disaster. The truth is, the U.S. government is asleep—which is unlikely—or they’re thinking, “This is just the nonsense of excited teenagers. We can intervene later if things get serious.”

2. Meta Enters the AI Assistant Race With Muse

That is what they think. So my first point is that it is not just individuals, but people who are obsessed with the idea that the world is going to end and that we are going to destroy it, who keep trying to do something. I have not given up hope that the U.S. government will take some action if it is a really serious problem. Right? So I think all of that is exaggerated nonsense. Right?

This is different from saying that Dario’s words are exaggerated nonsense. It’s just that it’s not realistic. Contrary to Dario’s argument, it is difficult to agree that cyber risks are real. Right? It seems a bit far-fetched for Dario to claim that, among the 3 risks, cyber risk and economic risk—the risk of everyone losing their jobs—are real. The third risk is the loss of agent control, which is difficult to assess. Okay, that is the problem he is trying to solve.

The solutions range from slim to impossible. An unlikely solution is to voluntarily install third-party monitoring personnel, but the next requirement is to make it mandatory. This is the first one. The second point is that all democratic governments must agree, and since Canada and Europe are on good terms, that should be easy. The third point is that we need to reach an agreement with China, and then it will go very well. It might go better than Canada, because we like China more than Canada. Russia must also be included, because we are on good terms with Russia. Great.

Now let’s look at the overall content. Let’s set aside the second and third pieces of nonsense and just think about the first one. The first proposal is to introduce a third-party monitoring body. If it is voluntary, do as you please, Dario. However, if the government decides that something needs to be done, that might be different from Dario’s thinking, and in that case, it is not your friends who are regulating it. It’s not something you can choose, either.

It will be established by law, and a regulatory body will be created, similar to a banking regulatory body. To be honest, I’m not sure if that’s a good idea. I’m not sure if this would really be effective for innovation, but if it actually happened, many people would say, “This makes no sense; it’s unnecessary,” or “If it is necessary, who are you to tell us what we need? We are the government.” I sympathize with both reactions to some extent. Sorry, I just vented a bit. But I think it’s really terrible.

3. Is Anthropic Preparing Its Biggest IPO Risk Factor?

I’m sorry, Harry. I just started. No, no, no. From my very tactical perspective, this was just a risk factor that occurred during the actual S-1 review process. From my tactical perspective, Anthropic is going public. That is partly because an employee who worked there for about 4–8 weeks said there was a 10% risk of destroying humanity, and many people agreed with that opinion.

However, he is merely identifying the risk factors in advance. That way, there won’t be any problems when the $2 trillion IPO actually happens. Honestly, I think he is clearly identifying the risk factors and that we are responding to them proactively. We will discuss this issue socially, and we will ensure that no one cares when we conduct roadshows in New York and other places. This is my opinion. I am not cynical at all. I believe it is something I naturally have to do as a CEO.

It seems that part was discussed less. Words have meaning.

I would like to reiterate that I will evaluate that part fairly. He didn’t say, “There is a 10% chance the world will explode,” did he?

Yes, that is correct. He responded to that remark. If the government were doing its job, it would convene a congressional committee, call in the relevant people, and say something like this:

“Dario, I’m talking about your company’s safety manager. It’s not about the person who quit; the important thing is that your safety manager said, ‘Several of our employees and I believe there is a 10% chance the world will explode within the next 5 years.’ You need to convene a congressional committee, summon that person, and ask him this: ‘As the head of this organization, do you agree with the safety manager’s statement that there is a 10% chance the world will explode within the next 5 years? Please answer yes or no.’”

That’s right, but it will happen. It’s just the beginning. You’re talking as if the federal government won’t raid Anthropic. I don’t think it will happen literally, but conceptually, the federal government will raid Anthropic.

How many texts did you receive last week from people outside the tech industry asking if AI is going to kill us? I even received text messages from relatives I hadn’t been in contact with for several years: “Is AI going to kill us, Jack?” Congressional committee hearings will continue endlessly for the next 24 months.

We have effectively woken up a sleeping giant. AI isn’t about some people living in San Francisco becoming billionaires. The public will believe that AI will kill us. I think that is also the reason why Trump ended the conversation so quickly. I don’t want to go on too long.

Actually, I think Trump said, “This is not a problem. We will do everything we can.” I don’t think he said that suddenly. I think it’s because of the thought that everyone will be talking only about AI for the next 2 years.

Let me make it clear that you are right on that point. Maybe I should have said it in the future tense. You’re right. Just as we just created a rock waterfall [?], you’re right. If I were an ambitious politician, I would do this. So maybe you’re right. There are 2 risk factors in S-1 documents. That is a really good point.

The first risk factor, just for your reference, is that there is a 10% chance that we will destroy the world. This part is worth discussing. That’s an interesting topic. However, the second risk factor is that I am hiring people who believe there is a 10% chance of me destroying the world. The reason I am doing this is because they have a very strong motivation to create excellent AI.

You could consult with our psychotherapist to find out why they are so motivated, right? As Harry just pointed out, because people at the company I run are spouting nonsense, it is highly likely that I will be under investigation for the next 2 years, which could lead to various bad things happening, including the government shutting down my company.

Jack, you're right. That is exactly the risk factor. I agree as well. Yes, that is a really strange danger: to think the company could go out of business because we said something stupid. Look, so much is happening here that it's hard to keep up.

I would like to add the 2 things I thought were the best. One is slightly political, and the other is nonpolitical. I think the best thing was what David Sacks said this week: “If the situation is that serious, Dario, it is your responsibility.”

Whether it is 10%, or whether your company poses a grave risk of causing the extinction of a portion of humanity, that is your responsibility. Your mission is to solve the problem. If we can't solve it, we have to shut down the company, right? And if it cannot be resolved, a new CEO must be brought in. If there is a company facing this kind of situation, it would be this company. That's your job, isn't it?

This is a basic principle of product liability. You can't kill a person. There is a limit to the number of people you can kill with your product.

4. What Happens if AI Becomes Impossible to Control?

I agree. And Jason, first of all, you are absolutely right. I should give you a compliment: it was 100% accurate. It was really surprising that Lina Khan, the head of the most hated antitrust regulator in Silicon Valley, expressed the same opinion. It makes for a really interesting situation to see David Sacks and Lina Khan standing on the same side and both offering excellent opinions. I agree as well. You hit the nail on the head.

I was naive. If you build recursive self-improvement, as people say, it is relatively difficult to predict what the results will be. Although it started with good intentions, couldn't it go in a different direction, or be exploited by malicious actors and weaponized? In my opinion—and I am saying this very clearly—I agree.

Actually, out of the 3 risk factors Dario mentioned, this is the only one I find interesting. I think the remaining 2 economic risks were foolish remarks for various reasons.

It makes no sense to say that you will not develop technology because it will cause people to lose their jobs. Eventually, we will all end up working on a farm. Agricultural workers account for 73% of total jobs. That is a really foolish claim. They are just trying to sugarcoat it and flatter people.

Cybersecurity is clearly a serious problem, but the situation is already irreversible. We think we are finished because the communists have taken over cyberspace. The only valid point is that, on the other hand, there was no mention of P(doom). So, it is actually a valid point in his letter, and at the same time, although I do not have the answer, it is the point that we may not be able to control things related to recursive self-improvement.

It's that kind of story. You're right. Among the 3 questions he raised and the 3 answers, this was the part that made me think, “Hmm, I see.” I understand Jason's argument. I think Jason and David's argument hit the nail on the head.

What you are saying is that we should set aside the 2 issues—that others could misuse our technology—and think about them again later. What you are saying is that you are creating something out of control. If you are creating something you can control, you might need to stop creating it, right?

You are the CEO, and to quote Jason and David again, while a third party might be monitoring you, if there were people smarter than you who could solve problems you couldn't, you would have hired them long ago. You should probably consider yourself the smartest person in this field.

So, if you think this problem is solvable, continue developing, and if you think it cannot be solved, it would be best to stop. But please don't rub your hands together and lament, “I wish the world would stop me from destroying humanity.” You are the CEO.

Actually, it was quite interesting to see the dissenting opinions pouring out just now. However, it makes no sense to admit your mistakes and try to put on the brakes now, after you weaponized the Chinese economy to build an incredibly powerful open-source ecosystem. You have weaponized these technologies to the point where they pose a serious cyber threat to all our organizations, and now that you are trying to put the brakes on it, do you expect us to bear the cost?

First of all, I would like to refute the expression “weaponization.” You are basically saying, “If we hadn't done it, they wouldn't have had these models,” but I don't think that is true. We could examine how much open-source data has been leaked, but I cannot evaluate that. There may have been some leakage, but there are smart people among them as well.

It seems excessive to grant an individual godlike status. In fact, apart from what Anthropic did, there were likely many alternatives to LLMs in China. This technology is already known to the world.

Yes, that is correct. Your point is correct. Well, since the secret has already been revealed, it can't be helped. Because you were the first to expose it. To be honest, I'm sorry, but I'm not sure if you just did that. That's right. You were the best. It's just over.

Again, I really hate the atomic bomb analogy. It sounds like a Nazi analogy. It's too crude and simple, and in this case, it only incites the pride of everyone involved.

But Harry, unfortunately, that is a really appropriate analogy in this case. Because even if people like Oppenheimer in the United States invented the atomic bomb, there is no doubt that Russia already had an atomic bomb 3 to 4 years later—not 4 to 5 years later. Part of the reason is that they stole our technology.

I will tell you once again that there was a spy in Los Alamos from the UK—I mean Klaus Fuchs. They stole our secrets and also made nuclear bombs. But what can be done? They would have arrived there anyway.

I have been making that point clear for the past 3 years. He said that it would have been better if artificial intelligence had not emerged, but, “It's already happened, so I have no choice but to do it. It's too late,” he said. He didn't say this last week. I have been saying this for 3 years.

It means that artificial intelligence has become too powerful. He has consistently said this since before xAI had developed to a certain extent. I don't want to do this, but it is already too late. I do not think this level of progress is desirable.

There are many exaggerated expressions in the letters sent to the CEO of Anthropic, and there is that letter signed by scientists, which he does not say himself. There are many exaggerated expressions that he did not mention directly, and there are subtle differences compared to what he is worried about.

Ultimately, the really important thing is losing control, right? However, I think his previous boss will hold him back right at this point. Of course.

Just as Jason showed an overly arrogant attitude regarding the impact on jobs and employment in the previous episode, now, no matter what the outcome, the responsibility is sometimes unfairly shifted onto him because he maintained that stance in the past.

Yes, that's right. And as Jason pointed out, other people really don't sympathize with this opinion. So, we are developing technology, and it has been revealed that the unemployment rate for office workers will definitely rise by 20% to 30% because of that technology.

And the Vice President of Safety, whom I haven't fired yet, wrote on Twitter that, “There is about a 10% chance the world will explode,” right? Hmm, I wonder why we aren't popular.

Okay, let me give you a few hints. May I tell you what the best thing I've read on Twitter or elsewhere is? It seems simple, but it might actually be true.

Jay Kreps, the founder of Confluent, which was acquired by IBM for $12 billion, recently resigned, and the article he wrote is, in my opinion, the best one. I heard that many people are spreading fake news or making foolish speculations. It means that marketing makes up the majority. He said this:

“It goes without saying, but for most positive use cases of AI, there exists a corresponding dark version. If you have superhuman talent for coding, you will likely be superhuman at hacking. If you have superhuman talent for structural engineering, you are likely to be superhuman at finding structural flaws that could bring down a building, aren't you?

“Even if you have superhuman talent for drug design, you will likely be superhuman at designing new, undetectable poisons. If you can cure viruses, you can also create them. Some of these aren't that bad and are manageable. But others are terrifying. This is a fact.

“If there are no safeguards in every positive use case, a dark version can be created. And in fact, open-weight models have no safeguards. A dark version can be created. All other things being equal, your AI will be equally excellent in both its bright and dark versions. This is a fact.

“We need to solve these problems, but this is an undeniable fact. And this is even more rampant with too many open weights, anthropomorphism, and free interpretations. Let's be clear: there are safeguards, but the darker versions will escape into open spaces.”

Jason, I completely agree with you. You're absolutely right. When people invented books, when Gutenberg invented the printing press, the Catholic Church was furious because they hated the spread of information and wanted to control all knowledge.

The same goes for the Internet. There is a reason why every totalitarian regime hates the Internet.

Every technology has a positive side and a negative side. We will find ways to develop the positive aspects and manage the negative ones, just as we have always done. However, we will not sell chips to China for the time being. Of course, this would trigger a typical Jensen Huang reaction—completely different from his usual response. So, in terms of implementation, it is not realistic.

5. How “Pacing the Frontier” Could Hit AI Infrastructure

Will there be ripple effects across the entire infrastructure, including the public and private sectors?

There was a slight hit to semiconductor stocks on the first day. Meanwhile, fortunately, CrowdStrike and cyber-related stocks surged by 10%. Actually, the interesting thing is that when I looked at it on Monday—I watched it; this content was recorded on Tuesday and is scheduled to air on Thursday. In the world we live in now, Thursday might feel like the very distant future. However, the immediate reaction was that investment spending related to semiconductors and AI dropped slightly.

It was not a sharp decline. There was no distinct change, just enough to hint at some degree of economic slowdown. Cyber-related stocks rose significantly, which I think is because the actual problem seemed to lie there. I always compare the software index, WCLD, with the semiconductor index, SOX, and today was a very good day for WCLD relative to SOX. Software rose, and while semiconductors fell slightly, it was not a sharp drop.

The tweet I saw was quite funny. The point was that “even with news of a 10% chance of extinction, the market moved by only 0.1%.” In other words, capitalism essentially took the risk and said, “It’ll be fine.”

This might be a much faster version of the 1920s. There are no regulations, and everyone is just trying to get rich like in the 1920s, aren’t they? We do not know yet what our version of 1929 will look like in the age of AI, but everyone wants to be rich.

The amount of $2 billion is just a starting round today, right?

I saw a chart today showing that the field raised $20 billion and $30 billion, and I thought, “Wow, in this day and age, $2 billion is just the starting round.”

There is so much money, and since AI itself is a massive money-making tool, it is too easy to exploit its dark side. If you can make money that way, you can fall into danger too easily. Since every large language model program has a dark side, it is too easy to resort to shortcuts if you can attract a $1 billion investment after Demo Day. It is so easy.

In a similar situation, there are also cases where everyone says, “Oh my God, this could be evil, bad, and dangerous.” It is as if you are worried to ease your pangs of conscience. But, Harry, you are right. No one says, “I should step down from this position.”

When I look at the people who run these companies, especially companies like Anthropic and OpenAI, I strangely think they are not profit seekers. Of course, we might have to defend them to some extent. They are not profit maximizers. If you were a profit maximizer, you would have owned more than 2% like Dario, or 0% like Sam, right? No, they are not people with antisocial personality disorder. What we should be worried about is not those 3 people, but the others.

I agree. Actually, we have some pretty decent managers in our top management. Sam, Dario, and Elon are top-notch. Theoretically, there might be better people, but in reality, it is difficult to find 3 managers better than them. They have a legitimate reason for what they are doing. I mean it.

The real problem is that 10% to 15% of founders have antisocial personality disorder. They really are like that. In fact, this is especially true for successful founders. Antisocial personality disorder allows you to create something out of nothing—the ability to manipulate people, the ability to have that kind of look in your eyes. People do that.

Bad people who abuse AI, namely sociopaths, are exactly that kind of people. If you do not think that is true, you have probably heard the saying, “10% of the founders I have invested in so far are sociopaths.”

Of course. I will move on to the next part.

Let’s talk about the AI assistant competition that I briefly mentioned last week. However, I was really disappointed to miss the fact that Meta released Muse. Muse is a product of Meta that acts as an AI assistant in many ways. People like that kind of thing.

We did not miss it. You mean you missed it? From what I see, it was released after we recorded the broadcast.

You are right. I would like to give us an A grade. We recorded on Tuesday, you know. To quote what I just said, on Tuesday I mentioned that 20 Meta engineers were locked in a room and instructed to release something. I said that on Tuesday. Muse was released on Wednesday, and on Thursday, the product was shipped.

So, I think we did a good job. It is just that I did not know there were 500, not 20. That is the only minor detail I missed.

There are 500. I did not know that as soon as OpenClaw started gaining popularity, Mark Zuckerberg brought in a significant portion of the AI team and declared, “I will make OpenClaw for consumers.” From that night on, people worked day and night.

I was just logging bugs on Twitter for fun during the Muse weekend, but the engineering team was responding in real time from Saturday night through Sunday morning. 500 people are working on this—not 20. It has been a top priority since OpenClaw started, hasn’t it?

It is a bit strange that it came out later than the other agents, but it has been P1, or top priority, ever since OpenClaw.

Think about it. There was a time when people would lie on the street with Mac Minis, trying to figure out how to run agents. Now you can do it on Facebook, too.

So, shall we start with this agent? I think Jack’s analysis will be as interesting as mine.

Yes.

I am in the UK, so I cannot use it. I would like to hear your opinion.

Yes. Harry, I will connect you to a VPN.

Hey, how is it? What do you think about it? How good is it? How good do you think this is as a reaction to OpenClaw?

First of all, as software, it is really, really, really great. It works right away. This is exactly the definition of great software. I cannot believe how well it works. All the unseen difficult tasks have been handled, right?

I can accomplish most of the tasks I want—things like creating reservations and sending emails. I asked Muse to send me the stories for this show, and Muse sent them to me and Harry. I passed them on to Lori. I sent them at the perfect time.

I was asked to rebuild the entire SaaStr.com website, so I logged into WordPress and created a new one.

You did a pretty good job.

It gets most of the work done that you want.

I thought of one more thing, but let’s move on to the core question now. The really interesting thing is something we have not talked about, but something I know a little about. It is not cheap, right? You provide free virtual machines to everyone up to a certain level, right? Two CPUs, 2 GPUs, 8GB of RAM, and about 100GB of storage should be enough.

I understand that companies like Replit, Lovable, and Vercel cost about $3 to $4 per person to provide such VMs. Muse sits on that boundary line because we offer more than other companies. The developers are trying to improve this part every week. That is because it is a very important part of their business operations.

It costs $3 to $4 to build a paid website, not a free one. Wix was 2 cents before Base44, but now it is $3 to $4. So, they are trying to lower costs every day. Meta is lucky. The biggest advantage is that the infrastructure is already in place. Whether it is free or not is debatable, but it possesses a massive infrastructure.

Because it runs on its own LLM platform called Muse LLM, you can enjoy massive infrastructure and Muse LLM benefits that no other company possesses. That is precisely why it is fast, has excellent performance, and provides outstanding performance in every aspect, such as more VPUs and GPUs. I believe there are few companies that can compete with Muse in terms of infrastructure.

All VMs, infrastructure, and storage are provided, as well as its own LLM platform, Muse LLM. What I have learned is that Muse is very suitable for general uses, as we are discussing, rather than for cutting-edge fields such as new drug development. It is undoubtedly a good product.

But the question is, is that really important?

What do you mean? What is a killer app?

Is it an increase in reservations for restaurants like The Cheesecake Factory or TGI Fridays? We need to look at the visual aspects of Manus—that is, what the killer app is. I am talking about the app. Every platform traditionally needs a killer app, right? I am a bit skeptical. I was just curious.

I do not think there was a killer app in OpenClaw either. I am not sure if there are killer apps for Muse or instant-messaging services, because there was not one in OpenClaw. Is it a search mechanism for shopping? I mean super apps like WeChat. Mark Zuckerberg said that receiving transaction fees is his business model. Do you not agree with that?

That is good, but what example did you give? You want to book a carpool for your daughter? Is that really what a billionaire has to do with a regular app? He could not even think of a good use case for Muse for himself.

We are just making consumer apps. I would like to book my daughter’s ballet class. Of course, it is a good app, but what should I do to run it for 8 hours a day? Like Claude Code, Codex, or a really important app.

If we used it like a super app for 8 hours a day, it would be fine. But if it is just an app used occasionally, I am waiting to see what the killer app will be. I tried Manus, too, and it was really fun. The design is so pretty, and it captures all the beauty and gives me ideas.

There is an Ideas tab that also tells you what to do. I tried most of them, and they were all good. However, I’m not sure if it will become a killer app.

I don’t think it necessarily has to be a killer app. It might sound a bit silly, but I use Instinct in a similar way, like a power user. However, the problem is that Manus’s response speed is too slow. You have to wait a few minutes for a response, just like in the early days of ChatGPT. It’s a really serious problem.

But I don’t think that’s absolutely necessary, because they do everything for me. It manages everything for me, from reservations and travel to restaurants. They even do all the shopping for me. I really like it. It also handles all calendar invitations.

There might not be a killer app, but it seems a little better than everything else. I’m not sure, but wouldn’t a combination of small features be enough? I’m not sure.

I think that might be the case. I’m skeptical about this category itself, but I always come up with 3 venture questions: Could this be a category? Who is the winner? And can we receive a reward for the risk? Could this become a category? Do people want AI to act as a personal assistant and messenger?

6. Can Standalone AI Assistants Beat Meta and OpenAI?

I have used both Instinct and Muse, and I think there is potential. I think the term “killer app” is a bit ambiguous. There isn’t a single app that does everything—an app that completely changes a platform, like VisiCalc. Putting the concept of a “killer app” aside for now, it’s when people around you say, “Ah, this looks pretty good.” That should be enough.

Looking at it from a broader perspective, there were Chinese messaging systems like WeChat a long time ago, remember? They integrated all functions into a single super app. Facebook also made a similar attempt with Messenger, but it didn’t succeed. That’s because the UI of the pre-AI era—specifically, the UI for booking via chatbots—was really inconvenient. It was much easier to go to the United Airlines website, view all the flights at a glance, and book. Right?

If there is intelligence on the backend—or even if there isn’t, if sufficient intelligence is stored in the cloud—it will be possible to provide many more features on mobile devices. It becomes possible to understand more information about the user and process desired tasks much more easily. That’s why, as you mentioned, we can imagine people using such services. For example, just like a Facebook user using Muse.

You must be a happy little camper. Are you on your way?

So, do I think more things will happen in the future? Yes. The question that’s really interesting to me is the second one. Will the winning company here be an independent firm, or will Facebook win regardless of what product it releases through Meta AI and OpenAI? Do you think a company like Instagram can build an independent business here?

7. Would You Invest in Instinct at $10BN?

To be honest, less than 5 months ago, the pre-money valuation was $50 million, and since then it was $200 million, billions of dollars, and now there are rumors circulating that it will raise over $10 billion. Do you think you can raise between $1 billion and $10 billion?

The most popular segment on today’s broadcast was Jack’s IC. Jack, ding!

Instinct has raised $1 billion in investment. Welcome to the partnership meeting. It’s a $10 billion valuation. Speaking of company valuation, are you asking if we plan to invest $200 million in this $10 billion investment round?

Yes, I will invest, and I will be honest with my team members. However, this is a risky investment. I was deeply impressed by the achievements Muse has made, and it could be a problem if they have to compete head-to-head with Meta. I would not recommend this investment because it cannot compete with Meta’s balance sheet, servers, Instagram accounts, and LLM.

However, through several reference calls and more than 12 calls, I learned a few things about Manus. First of all, Meta cannot move toward being cross-platform. It will not work across all platforms or services and is focusing only on its own platform. In other words, Meta is merely a part of the way we communicate. How many people between the ages of 18 and 55 use Facebook all day long? It will be a very small minority. It’s practically an app used by the grandmother generation.

Although WhatsApp and Instagram are gaining popularity, their reach is significantly limited due to the lack of compatibility between the various services. Secondly, how long can Meta pour its passion into this project? This project actually generates almost no profit.

Do you remember Workplace? Workplace had many advantages over Slack for people who use Facebook a lot. It might have been better. There were many useful use cases, and it worked well. The architecture was excellent, and in fact, it was the product with the highest NPS across the entire Facebook and Meta platform, but it failed to maintain that enthusiasm.

From what I’ve found out, this product is going to collapse if Alex leaves the company. If this were all Meta had to do, I wouldn’t guarantee its success, but I don’t think Meta will continue to put in the same level of effort as the Instinct team. The Instinct team is a really great team. As I’ve seen through my investment experience, it’s truly excellent. These kids come and play World of Warcraft in real time during pitching. Both have skills that place them in the top 10 and top 15 of League of Legends.

I recommend leading this round, but you must manage your reserve funds carefully, as the next round’s valuation may reach realistic IPO limits.

Then, would you like to participate in this round?

I won’t do it.

You told me to do it, but I will absolutely not participate in this round. Great, great.

The reason I’m not participating in this round is as follows. You may call me old-fashioned. Most of the pitching content is correct. However, infrastructure costs are too high, and existing companies have a major advantage.

8. Why Meta Is the Biggest Threat to AI Assistant Startups

This is just like what we’ve been talking about since we started this show: What if Claude, Anthropic, and OpenAI actually built an app? If you look at the entire history of this show, they didn’t build anything other than Codex and Claude Code, and effectively only made a half-baked app. Claude Desktop can’t be called a complete app either.

But Meta is making an app. They have an LLM, cost advantages, speed advantages, and more computing resources and GPUs than anyone else. And they are making an app. This is the threat that all venture capitalists were worried about. We all received a free pass because LLMs didn’t make apps in the early days of AI. But Meta is making an app. So I don’t want to compete with them.

I think this will be a very important priority for the next 24 months. As Harry already said, Instinct is slow. That’s a sign that other companies’ computing costs are high, so they have to raise $1 billion. Could venture capital provide $5 to $10 per user per month? Of course. But what happens if you try to generate excessive profits? It could be just like Poolside.

Of course, it’s good, but imagine there are as many as 10 million users paying $10 a month. Now I have to repay a huge sum of $1.2 billion annually, but I’m having difficulty attracting the next investment. I’m not a successful company like Databricks, you know. It’s concerning when an existing company attempts to develop an app with all the necessary capabilities.

No one can be wrong, right? That’s what I’m saying. I could be wrong, too. But was it Socrates, or who was it again? Among the ancient Greek philosophers, there was someone who was able to switch to a different argument in the middle of presenting one side of the argument. Jack can definitely do that too, right?

This is exactly why I think I need to invest $200 million. However, I don’t want to make this kind of investment transaction.

I was really impressed by your mental ability to handle both sides. It’s enough to give me goosebumps. It’s just like a human LLM. You can persuade me of the kind of person I want to become.

Basically, what you said is that this is the core of Meta, and it feels like something they would want to do. Despite your opinion on cross-platform compatibility, Muse is a standalone app, so it isn’t a cross-platform issue. However, if they do integrate this, we can only expect that they will make it available for use in Messenger or Instagram.

I think if there is just 1 thing Meta needs to do in the field of AI, it would be this. It’s hard to imagine that they would invest over $100 billion in AI and say they would create personalized AI, yet not devote all their efforts to this area. So, I agree with your opinion, Harry. This is something that must be done.

Honestly, I don’t think the comparison with Slack is appropriate. Facebook’s business products were like toys. They were not a core issue for the company. However, Manus is the core problem. So you’re right. They were completely captivated.

9. Could Instinct Become a $50BN Acquisition?

The interesting thing is that similar results can be obtained in the case of the Poolside analogy and Instinct. In the case of Poolside, “We ran out of capital to play the game any longer, but we got a good result because a company with a much larger market capitalization wanted the assets accumulated by Poolside.” NVIDIA wanted access to models and talent.

Manus is the same. It has built a massive user base, and OpenAI says, “That’s interesting. You can acquire it by saying that.” I think it’s highly likely that the founder of Instinct worked at Sierra or is a huge fan of Bret Taylor. Bret Taylor is active in various fields, including serving as the chairman of OpenAI.

Even if Instinct achieves significant results, its IPO potential might not be great in terms of cash flow. However, we cannot rule out the possibility of a very attractive sale. While $10 billion might be a somewhat exaggerated figure, if it delivers meaningful, differentiated performance, it would be an attractive investment target for someone looking to build a business in this sector.

Everyone I have spoken to has named Instinct founder Noah Shinn as one of the most outstanding talents. Given the current situation, where top talent in strategically attractive sectors is actively pursuing multi-trillion-dollar M&As, I believe there is a sufficient possibility of a $50 billion to $60 billion sale, though that might sound a bit exaggerated. It is possible. It is just my opinion; we all have different experiences.

The way I learned about venture investing is that, to succeed, you should not make investments that require a 100% M&A outcome. That is how it was. It is just too unpredictable. I have been through that situation myself, so I know how fickle it is.

For example, they might say they are Clem’s best friend, then Clem might set you up with Jensen, only to quit the very next day. You really cannot predict it. So, you have to see something certain—where selling at an incredibly high valuation is the only feasible exit strategy.

Of course, that is also a form of gambling.

It is indeed gambling.

Yeah, I agree. It is a really big gamble. Coming back to that topic, I have thought about it myself, too, although I do not make those kinds of investments myself. Looking at this, as you know, I always ask myself: Is there anything to learn from this?

There is room for debate. I am not making an argument yet, but I acknowledge that the expected value of these investments can be quite high. Of course, the volatility is high. In other words, it is a risky way to make money. When constructing a portfolio of 30 investments, is it okay to include about 3 of these?

It probably would not work if it were $10 billion, but we will have to wait and see. It was not a reckless investment. I remember Kleiner asking for $500 million, but Mamoon is always smart. Raising an investment of $500 million is not reckless, even if you assume there is a 2-in-10 chance of achieving a positive multiple of 20x.

If you do not get that positive multiple, you will end up creating a company with poor cash flow. You might not succeed. As you know, the expected return is high and the portfolio is solid, so it is worth the risk. I would do it the way I run my own business.

Actually, we are in a bull market, and as Howie said, frankly speaking, there are many companies with significant market capitalization that need to move freely and quickly, with free M&A opportunities. It is not a crazy thing to actually make money. Even if the situation changes now, it will not be a big problem, since we need to secure $1.5 billion in preferred stock. Because we have a universally recognized, top-tier team, the downside risk is relatively limited.

This is a typical Silicon Valley case. If you look at the return profile for $50 million—I think it was around April—it is an excellent profile. It is even more interesting when it was $50 million, around May or June. It was a good investment opportunity because they were a bit ahead.

The problem is that they raised the investment in just 2 or 3 months. I think it was $2 billion, is that correct?

I think it was $2.5 billion. Was it $2.5 billion?

Once they surpass $1 billion, that is the point where a 4x return is generated. Now they are trying to raise another $10 billion, but they are in a situation where they really need $50 billion. I should remind you once again: the largest M&A in history was Cursor being acquired for $60 million from $4 billion, was it not?

Without massive revenue, it would be difficult to achieve even a profit of around $40 million.

What I mean is that, in April, the risk-adjusted return was incredibly attractive, but its appeal faded quite quickly by the month. That kind of problem comes with this type of investment.

However, the interesting thing is that the consumer market response to this product has been truly explosive. I uploaded an Instagram Reels video, and I received over 1,000 DMs asking for invitation codes. I have never received 1,000 DMs after one of my videos before. That is really surprising.

It is amazing.

No, and I should add one more interesting fact. Speaking of you, Alex Kurland went to Menlo Ventures, did he not?

Yes. He was a board member with me. We have known each other since the very early days. He is a mentor I have known since he first stepped into the industry.

He wrote a short presentation on how Menlo Ventures views this issue. It is nothing groundbreaking, but he mentioned that Menlo Ventures is aiming for 100 technology companies worth more than $25 billion. That is what they are using. That is how Menlo Ventures models the world, and the reason he joined Menlo Ventures is precisely to find some of those 100.

It does not have to apply to all 100 $25 billion companies. Exits are included too, right?

That is the model. We can say, “Wow, Cursor is at $60 billion, but Cognition has gone up by $48 billion,” right? If this is the world model you are presenting, there are currently 81 technology companies worth $25 billion, up from 23 10 years ago, and this trend is expected to continue in the AI era.

I wrote that the standard for a new decacorn—a top-10 company—is $25 billion, so this Instinct investment round seems reasonable. If all successful exits are over $25 billion, you could make at least a 3x return on your Instinct investment, right? It was interesting to find out that their model was exactly that.

We are aiming for an exit of over $25 billion from every investment, and we expect there will be at least 100 more such cases in the future. However, considering companies with GDPs and market capitalizations amounting to trillions of dollars, I would have to ask Dr. Rory O’Driscoll, who is sitting next to me, how all these calculations fit together.

I do not have a master’s in math, so I cannot calculate it in my head. However, I think there must have been considerable deliberation in this calculation process. This is one of the 125 billion-dollar exit cases.

Taking it a step further, let us look at Miro, one of the leading companies in the AI field. Miro raised $17.5 billion in investment in 2021 and is famous for its interactive whiteboard solutions for teams. For those who are unfamiliar, the company was sold to Bending Spoons, a giant Italian investment firm, for $1.3 billion. That is a significant difference from $17.5 billion.

10. Miro Sells for $1.35BN After a $17.5BN Valuation

We were unaware of this case, but my smart partner, Paul, analyzed it. Miro generated revenue, and the founders and employees made money as well. However, recent investors have not seen particularly large returns. In fact, they only made a return of about 1x.

How do you view this exit by Miro, the darling of the SaaS ecosystem?

I think it was inevitable, and I do not see it as bad. The reason I say it was inevitable is that there is something interesting: all exit cases are listed in our Salesforce reports. They are unicorn companies, right? There are various ways to do it. One of them is literally using investment funds. It is viewed as a standard, so you just scan the data to see what is happening.

When you rank them, something immediately stands out. The last investment round was $17 billion in 2021. If you scan that, you can see that the level is steadily rising, and you can see other companies investing $10 billion or $20 billion above and below it. The number of employees is increasing simultaneously. Good heavens, the workforce is growing explosively, yet that company is stuck at $17 billion.

That was the highest valuation at the time, but it has completely stagnated. Looking at that, you can see that you are almost inevitably going to get into trouble because you are only making personal investments. Evernote, a productivity tool, was capable of maximizing productivity in a world detached from reality, but in terms of corporate valuation, it was far ahead of its time. This is especially true when considering the actual market size or growth potential. Ultimately, this outcome was inevitable. It was only natural that something like this would happen someday.

I am reminded of a funny picture posted on Twitter by Andrew Reed of Sequoia Capital. It depicts the Grim Reaper holding a scythe and knocking on every door: he knocks on Evernote’s door, then Airtable’s, and now he is knocking on Miro’s door. Instead of a scythe, he is holding a bent spoon. It implies that death comes to everyone, even in the SaaS industry.

That was a truly accurate description. The pricing was so wrong that a liquidation process was unavoidable, and ultimately, it was a good thing for everyone. I think I mentioned late-stage investors a bit bluntly, so let me reiterate: the biggest advantage of a late-stage business is exactly this. If you lose this deal, you get a 1x return, and then you can die rich.

ICONIQ seems to have invested a massive amount of money in that deal, did they not?

So, that was a bad deal. Getting a 1x return is a bad deal. When making venture investments, especially when investing in late-stage companies, assuming a worst-case scenario, if you consider it advantageous to earn a 1x return, a simple calculation shows that the overall profit ultimately turns out to be positive. It is a good outcome for everyone.

It was unavoidable, and it has now become a liquid asset. An interesting point, which I learned for the first time today, is that, if I may speculate, a portion of the consideration was carried forward. In other words, some people said they would accept Bending Spoons stock. It is an interesting choice.

I am not sure if it was a mandatory condition or a method to maximize profits. They could increase profits through the rollover. I believe either way is possible, right?

11. The Bending Spoons Playbook: Raise Prices, Cut Costs

Bending Spoons needed funds. As you know, they do not have unlimited capital. There are limits to the investments they can make, right? However, it could have been a rollover. There is a possibility that it was rolled over to a company that is already on a growth trajectory.

In my opinion, a rollover is an interesting thing. Basically, “We couldn’t invest enough to make this company cash-flow positive, so at 3x—2.7x—I am selling at that price and selling to a company that is trading at 14x, because they are strong enough to make the necessary investments.”

It is like saying, “…” Right? It is actually happening. To some extent, it might sound like nonsense, but it is true. It is a venture syndicate. I have served on a board where the company was stagnant and we had to ruthlessly increase efficiency.

Venture investing is not in our DNA. It is not our way, either. It is usually a syndicate composed of 5 different people. These assets—this is how we will do it. There are times when it is better for a single owner to manage it, right?

Did you hear what the CEO of Bending Spoons said? He made a very controversial remark: “We don’t care much about the title of founder. I don’t want to know what you did 10 years ago.”

That’s what I said. In other words, “We don’t want to know what you did when you started your business 10 years ago. I want to know what you are doing right now.” That is what it means.

Ultimately, it means, “We do not have the capacity to solve this problem at 3x revenue, so we will sell it to you for 12x revenue. Then, you will buy your stock back at 12x revenue because you will be cool-headed enough to cut unnecessary costs, raise prices, and tolerate significant customer churn to move forward.” Right?

It is interesting, but not entirely accurate. It is also an interesting commentary on how institutions can determine outcomes. Not everything follows rational economic principles, and perhaps there is some truth to it.

Bending Spoons would be better at making a business cash-flow positive than a venture capital fund. If you are a customer of these companies, prepare for a 40% price hike.

Well, that is all. Customer churn is obvious. The customer churn rates of the companies they acquired are massive. In fact, if you look at the usage graphs, the churn rate is truly serious. They are not revitalizing these companies; they are raising prices and cutting costs to the extreme.

Yes, that is correct. However, let me explain it a little differently. What they are actually discovering is the difference between marginal propensity to pay and the amount of investment.

Venture capital firms across the industry have invested excessively in sales and marketing, creating customers who are forced to buy products. What they are saying is, “We do not need customers who are forced to buy products. We want customers who dislike us even if the price doubles, but do not leave because they still need this product.”

That is the point. Do you understand? It is a completely different perspective. Of course, customer churn will occur initially after a price hike, but their view is that the remaining customers truly need the product.

It is just like when you have to sell everything. If you triple the price, the customer churn rate reaches 30% to 40%, doesn’t it? It is very simple to calculate, right?

12. Are We Entering the Era of SaaS Capitulation?

That is correct. Let me share my thoughts on Miro. I almost wanted to skip over it. You know the game of musical chairs that kids play? Don’t you remove a chair every time you go around?

Yes, that is correct.

It seems like there were only 1 or 2 seats left even before the AI era. Miro looked at Bending Spoons.

In an interview this week, they said they seriously review 1,000 companies and invest in only about 5 to 10 a year. Right? Besides, their financial power is not unlimited, either. Most other private equity funds are out, and Thoma Bravo is no exception.

So, out of 1,000 unicorn companies, there are only 1 or 2 seats left. It seems the same. We could discuss why Iconiq only received 1x. If there had been many options, the situation would have been different.

When the musical-chairs game ends, there are usually only about 2 seats left. Miro took 1, right? Just like Airtable, they only received 1 offer.

If they sell for around 2.x, you know for sure there are no other offers because anyone can pay around 2.4x or 2.5x. It does not make a big difference for companies like Salesforce, Thoma Bravo, or Francisco Partners.

So I acquired 2 deals, and honestly, I want to move on to something else now. Personally, I think I have finished this game. I am not going to run around the chairs anymore.

Well, that is fine. This is the last one, guys. Who is next? I do not think there will be anyone. Of course, there will be more deals ahead, but I think we have entered the era of surrender.

If there is a seat left, grab it. Grab it. Otherwise, the game is over. These companies are going to enter zero-sum mode. They will show growth rates of around 0% to 5%, and nobody, nobody, is going to buy these companies.

If Bending Spoons is like this, then this is the best Miro and Airtable can do. What if you are not Miro? These are not bad companies. Miro has $600 million in annual revenue, is still showing high-single-digit growth, and has positive cash flow. It is a pretty decent asset, right?

A company worse than Miro or Airtable will not even get the last 1 or 2 seats. Nobody wants to buy something like this. I do not mean to sound mean. It is just like I am saying, “Guys, do whatever you want. Here are the keys to my house. I moved to a different city. Throw parties, smash cars, do whatever you want.”

Because, guys, we need to talk about what we should do next, right?

Jeff Dean’s company surpassed $50 billion after raising just $10 million in investment. Citrini and Dylan Patel said he sold the company to Sam & Ellis for $100 million, and OpenAI temporarily suspended Pro membership registration.

I wonder if SBF is being released?

Oh, well. Shouldn’t he receive the Midas List award?

What is that?

Should he receive the Midas List award? Maybe it was when he was selected for Forbes 30 Under 30. I think he was on Forbes 30 Under 30 at the time of the Anthropic acquisition.

I will be the voice of humanity.

Yes, he has been serving a considerable amount of time, which is truly unfortunate and amounts to a waste of a life. However, I do not want to blame him yet. I think the Supreme Court will hear this case and overturn it by a narrow margin.

That is why he does not want to hire a lawyer.

No, I watched a video of his lawyer on YouTube, and he was really excellent. They say he has handled more than 50 cases in court. He possesses tremendous skills, like a Supreme Court lawyer.

His point is that, although I am not a generous person, this is not an Eighth Amendment issue. With $12 billion, you cannot find anyone who can pay it back, can you?

According to the bankruptcy court’s terms, all creditors received full repayment along with interest. Of course, whether they made much more money is debatable, but there are clearly constitutional issues.

I think he will eventually win. I believe the Supreme Court will take on this case. The Supreme Court is not obligated to handle every case, but I think he will win.

Even if he is sentenced to 30 years and fails to repay the $11 billion, he might eventually be released. Having the Supreme Court hear this case is far from a release, but I do think he might get a chance to stand trial.

Could he be released? Honestly, back in 2023, he looked like the greatest fraudster in human history, didn’t he? Right? Isn’t it strange that, as far as we know, he did not profit materially?

It is controversial. While his transfer of assets between Alameda Research and FTX was certainly a terrible thing, the claim that it was permitted under the terms of service is an interesting argument, right? It is an interesting argument.

He did not pursue personal gain. Realistically, by a generous standard, the victims have been compensated for their damages.

Should he really spend most of the rest of his life in prison and pay $11 billion in damages after his release?

In an era of intelligent AI capable of killing us all, that amount seems excessive. It might look like a lot of money today, but back then, he lost $10 million over the weekend as Silicon Valley Bank went bankrupt.

I do not know how you felt, but at the time, we thought it was his own fault, didn’t we?

Yes. Just so you know, this was not on the list, so I was not prepared for it in advance. My wife is a lawyer, and she gets angry when I practice law without a license.

But Harry, this was truly unexpected, so I never thought we would be talking about this. Right?

I was surprised it was not on the list. Putting the fine aside, the amount imposed on him is not the most important thing, but there was embezzlement. It was white-collar crime. He deserves to be punished.

Actually, 30 years seems a bit excessive. In my opinion—what was it again? It was about a former Goldman Sachs employee who mismanaged funds at MF Global. It happened 10 or 15 years ago, and I think he was acquitted. I do not remember clearly.

I believe white-collar crime should be punished. It is truly terrible to live in a world where someone goes to jail for stealing $20, but another goes unpunished for stealing $10 billion simply because they are white and upper-class. However, I am not sure if a 30-year sentence is a commensurate punishment.

The interesting thing is that the issue the Supreme Court is dealing with is not that. It seems to be an appeal regarding the facts and circumstances of the case, not an appeal regarding sentencing guidelines.

We will have to wait and see. Also, I do not think the fact that he embezzled money and was a brilliant investor justifies his act of embezzlement. By definition, that would mean that if it works, anyone can take the money.

For example, “Harry, I stole your money, but I bought call options and the stock price went up, so I made money. Here, I’ll give you your money back. It’ll be okay, right?” You could say that, but naturally, they would be upset. At the time the money was taken, they didn’t know what the outcome would be, and there was a 50% chance the stock price would drop.

Naturally, they would want such a person punished, so that this behavior is never repeated and no one else can copy it. The fact that he was the best stock investor of our generation across Anthropic and Solana is irrelevant. What matters is whether he embezzled money and whether that broke the rules.

I haven’t heard any claims of a violation of the Terms of Service yet, Harry. That would be an interesting story. In any case, embezzling money is a rule violation, so he deserves to be punished. If he accepted the money, that wouldn’t be a rule violation, so he would probably have to resign.

The proceedings will go through, and that’s not my problem. Now, shall we talk about what happened to Matt Mullenweg? He was away for a day and came back, but what happened to poor Matt? Really, who hasn’t read the articles of incorporation? What on earth happened?

Right? Did you forget to bring the certificate of incorporation from Delaware? Anyway, Matt Mullenweg is the founder and CEO of Automattic. It seems he was removed from the board but returned and overturned their decision. Now that he has returned as CEO, the authority of a founder has been exercised once again.

He has returned as CEO. That is correct. The company is Automattic, which has been managing the open-source project known as WordPress. As you know, WordPress is one of the most widely used blog and website platforms. It is a very successful product, and Matt is the CEO of the company that manages it.

It seems reasonable to say that there have been operational issues over the past few years. He has been in major conflict with WP Engine because WP Engine is a company that hosts WordPress sites, and Automattic wants a portion of those profits. Automattic has been pressuring WP Engine, but his conduct has not helped the open-source project.

It was as if he was saying, “I will use my company’s influence to prevent others from enjoying the benefits of the open-source ecosystem.” That is why this action directly contradicts the spirit of open source. That is why the situation seems to have been bad for quite a long time.

In fact, the real problem is that the world is moving ahead of that product. It is a bit unfortunate. The world is turning its back on it, and as Jack knows, most of the features that Lovable can do can be implemented with 10 other tools, like WordPress or Replit.

Increasingly, as Henry Kissinger said, the reason academic politics and debates become so fierce is that the stakes are too low. In fact, automation is no longer important. We need to create something new, but as Jack pointed out, we are stuck at the tail end of technology trends.

To survive in the new world, we need to approach things in a completely different way, but instead, we are just arguing internally. This is an example that illustrates the overall situation. The funny thing is that, as with all lawsuits, getting bogged down in routine details makes you forget the big picture.

The big picture is that this company needs to lead WordPress in the future. The point is how to leverage what is happening in artificial intelligence to become a company that plays a significant role within the next 5 years. Otherwise, you cannot survive. That is an essential task for the company.

However, that essential task has devolved into a war of nerves between the board and the CEO. It seems the CEO won in the end. Congratulations—you have taken the poisoned chalice. You have secured the ability to maintain your shrinking empire.

I must also point out that the board is not filled with unscrupulous venture capitalists. I have served on a board as an unscrupulous venture capitalist myself, and the experience of having to replace a founder was truly terrible. However, I understand that Salesforce invested in this company, and there are a few excellent independent directors.

There is no need to get into this predicament. I will not get entangled in this. Now, let’s dig into the tactical details. I think Jack probably hit the nail on the head.

There is a board, and the members likely hold a majority. They tell the CEO, “We are independent directors who hold a majority on the board. However, I think I need to replace you.”

That’s what he probably said. My guess is that the founder CEO scoured the articles of incorporation and said, “You are members of the board and can replace me, but actually, I can replace the board.”

“I will vote in favor of replacing the board of directors. You all leave the board. Look, the new board is me, my pet dog, and my ventriloquist dummy.” After much deliberation, I decided that I would become an excellent CEO.

That’s what he would have said. Well, this is how it turned out. All the independent directors would have resigned immediately at that point, because there’s no need to waste your life and get entangled in lawsuits.

Do you remember the one where the Fortnite island keeps getting smaller? You are fighting to maintain control of the ever-shrinking island. Congratulations, Matt. You’ve become the leader. Now that you’re in charge, why don’t you try changing the situation a bit?

That is what I would like to say. Yes, that is exactly what I felt. In a way, it is sad that something like this is happening. People are just human, and they are bound to act as they please.

Many people were involved in that company and put in a lot of effort. One true venture investor, like Tony, invested heavily in the early stages. It feels like a kind of blood-feud situation. Seeing how long this hasn’t been resolved, I think it would have been nice if the outcome had been satisfactory to everyone.

Automattic would have become a great company if it hadn’t received venture capital. Matt could have dedicated himself to what he truly wanted to do: commercializing the open-source products he created. Besides, he was very young. Imagine generating $500 million in annual revenue and spinning off $200 million. It’s like a much bigger Basecamp.

As you know, those aren’t exactly the kinds of companies that would receive venture capital, but they own 22 cars like Lamborghinis and Paganis, and villas in Italy while generating $60 million in revenue and growing it by 30% annually.

I have absolutely no qualms about spinning off $30 million or $40 million in cash. I don’t care at all. I don’t know the full details of the financing, but couldn’t it be that raising this much money wasn’t really necessary?

If it had been a similar company, it might have done quite well. It’s easy to say, but perhaps that was the problem, and it might be the reason Matt is frustrated. I would have regretted it too if that had happened. As you know, companies sometimes receive excessive investment.

I might be thinking right now, “I could have just left 37signals. I could have run it myself. Why was this necessary?” $800 million, plus people running around doing nothing all day.

I would have been better off running WordPress and Automattic with about 80 people, like DHH, making $100 million a year. If that is my life goal, what does growth matter? If a company is making hundreds of millions of dollars in annual profit and achieving a 5% growth rate, and I am happy and doing good for the world, then I don’t care about things like venture capital.

Of course, I agree that if you don’t intend to receive venture capital investment, you shouldn’t. I don’t think there are only 2 aspects: venture capital investment and a lifestyle business. I believe there are other dimensions, like an open-source business.

So, looking at those 2 aspects—venture capital investment and a lifestyle business—it is quite clear that if you want a lifestyle, running a lifestyle business is the right choice. There is another dimension, however: open-source community projects and managing them solely for the sake of the companies that control them.

I cannot be certain that those companies have managed the projects excellently.

Right? Yes, exactly as you said. In my opinion—and while that is merely an early-stage dynamic—as Jack said, if you started a project directly without receiving venture investment and can exert influence over licensing and copyright, such as open-source rights based on source code or copyright, then it is your company, and you can do whatever you want.

Actually, I agree that this is the United States. Whether you created something, managed it poorly, or wanted 100% ownership, whether you like it or not, that is allowed. Go team! That is the meaning of ownership.

Do you know what else is difficult? I know you want to wrap things up, but these are things I have only just begun to learn as I build my investment career. To run an open-source company that receives venture capital, you have to be ruthless. You have to be ruthless. Really.

Look at what Matt did. Matt said this: “I am focusing on the platform. This is the field I am passionate about. I am not very interested in hosting, because hosting is a product business.”

“WP Engine will let other companies handle about $500 million, $100 million, $80 million, $100 million, and $300 million cases on their own. I am interested in e-commerce fields like WooCommerce, but I don’t want to go toe-to-toe with Shopify or Tobi Lütke.”

So, in the end, he is unable to take a significant portion of the profits within the ecosystem.

I don't think Matt Mullenweg will get angry just because WooCommerce's revenue isn't large. WooCommerce is actually a much larger platform in terms of scale than Shopify, but if you look at WP Engine, this company treated our community much worse after being acquired by a private equity firm.

Objectively speaking, that's true, but honestly, $500 million isn't a waste. I wish I had invested another $500 million in this terrible hosting service that I could have done better. Actually, we're using their professional products. It is a great product, though. However, I think this product is doing only a tiny fraction of what it's supposed to do mathematically. So I want that $500 million. That's the thought that comes to mind.

13. Why Venture Is More Extreme Than Ever

I think being too generous in the open-source industry is a disadvantage. It means that being too generous leads to a loss. You need to stay calm and think about the few agendas we discussed here. On the one hand, Mira, which lagged behind in technology trends, was acquired at 2.7 times its value, and there was a minor dispute surrounding Automattic, whose valuation in the open-source sector had stagnated or declined.

At the forefront, there's a company like Instinct. Four weeks ago, it received a $2.5 billion investment from Lilly, and now it's receiving a $10 billion investment. The company spun off by Jeff Dean also received a $10 million investment a few weeks ago, and now Tencent is receiving $50 billion—that is, $10 billion—in investment. Now, $1 million is no longer treated as a number. One million dollars is an accounting unit used only in the home-design industry. These companies are receiving investments of $10 billion to $50 billion. They're big companies.

This, while stating the obvious, demonstrates that, unlike private equity, venture investment isn't about valuation, and there are no safe assets. It's like sitting in the front row of a train, where you experience new things and everything becomes possible, or sitting in the back row, where your life goes awry.

I met with one of the industry's top CIOs recently, and I asked him, “You've been doing this for 30 years. Have you ever seen times like this?” He said, “I've never seen a time like this before. I've never seen anything this crazy.” This is unprecedented. I really think so.

From my perspective, I've been investing since 1993 or 1994, and that was a really hectic time. There was widespread anxiety among the millennial generation that the world would soon end because of the Y2K problem, just like now. It seems like there's always worry about the end of the world.

To be honest, New Year's Eve in San Francisco on December 31, 1999, was a really crazy party. Half were rich, so they were drunk, and the other half worried all night long, hoping the Y2K problem wouldn't bring the world down. At the time, that was a real concern. Ironically, you see, compared to today's AI, it was nothing.

I think AI's ability to stimulate people's imagination is much more amazing. The internet was amazing, but AI allows us to handle software like humans do. You might get a little too excited, but it's nothing more than that.

Secondly, instead of the existential worry that computers might stop working because of the Y2K problem, we now have to worry that the world might end. Everything is happening in exactly the same way, but on a much larger scale. That scale has grown tenfold because the money has grown tenfold.

As you know, I live in Europe.

Roelof often reminds me of this, but that's not actually true. You live in the UK, and the UK intentionally didn't join the European Union.

But I understand what you mean. It's okay.

14. Mistral Raises €3BN: Europe’s AI Sovereignty Bet

I'm sorry. That was a bit harsh.

I wasn't a Brexit supporter, but you're right. I agree. It's really unfortunate.

Mistral has secured €3 billion in investment. It is the largest technology investment round in European history. They say they will achieve €1 billion in revenue by the end of the year. For a company that has faced a lot of criticism, especially in Europe, this is a very meaningful sign of progress and hope.

In other words, it's as if there are still horses left to enter the race.

What should we aim to gain from this round?

I think we should focus more on AI sovereignty than on becoming a competitive, cutting-edge research institute. I don't think the fact that they “have horses” is a stumbling block. It's nonsense to say that Mistral is now competing with OpenAI or Anthropic in the cutting-edge model race.

I think what you're really trying to say is that, considering the behavior of U.S. companies and the political interactions between the U.S. and Europe, Europe determined that an element of European sovereignty is necessary for such an important technology. Despite being absurdly inefficient from any rational perspective, Europe decided that it must support Mistral so that it can secure sufficient business opportunities as a European AI competitor.

Europe has maintained this tradition for a long time. Airbus is one example. They said, “We cannot entrust airplane manufacturing solely to the U.S. Otherwise, we'll just become a dependent state.” That's what they said. “We'll make it happen.”

France and Germany said, “We'll make airplanes, and we will too.” It took 10 to 15 years, but they developed competitive technology. It's probably the same here. They simply said, “We can't handle it.”

That is what they're saying. It happened recently, and if I recall correctly, the U.S. government told Anthropic to block all countries' access to Fable. I believe it was Fable, one of the most recent models, and the order was to block it because of security risks. It wasn't just about blocking Russia; it was about leaving the UK, France, and Germany untouched. It was to block every single country.

If you're in Europe, on the day that happened, you must have thought, “We can't rely on these people anymore.” On that day, you made Mistral a competitive company in Europe. Will Mistral grow as big as us? No, it won't. It took Airbus 30 years to overtake Boeing, you know.

In the short term, its market capitalization won't be as large as OpenAI's or Anthropic's, but it will certainly succeed, and it will succeed in Europe. So we should be grateful to politics. If you're a Mistral shareholder, you should be very grateful for the current political strategy, since it has generated billions of dollars.

Technically speaking, Samsung led this investment round, right?

It has generated cash. Just think about the scale of the annual capital inflow right now: $3 billion. I'm not sure whether I should add a margin. Should I add a margin? Of course I would, but if it's an investment led by Samsung, it's hard to view it as a truly reliable investment.

The last investment round was led by ASML, too. I'm not sure if that counts. The scale of capital is important. Of course, corporate valuation is important, but I'm not sure it exactly matches the value objectively assessed by financial experts. Intrinsic value clearly exists, after all.

To put it bluntly, if the value of U.S. companies is $1 trillion and Europe's GDP is about 70% to 80% of the U.S.'s, I don't think that means Europe needs to invest $800 million. It could be $1 billion, but $30 million, $40 million, or $50 million would be fine. It isn't such an absurd final goal. I don't think it was based on market comparisons or lists of comparable companies. It was decided from a strategic perspective.

15. Adobe’s New CEO

I think so. Everyone, please let me know if there are any other stories I might have missed. Jack, what are your thoughts on Adobe? A new CEO has taken office, you know. I've seen the results, but since you're an Adobe expert, what do you think?

I think it's very good. In a situation where growth had stalled, with 2 non-founders leading different business units, they're essentially picking the less capable of the 2 decent leaders. What I mean is, it feels like they just expanded the mess.

But it survived despite being this large, didn't it? It shows that the moat is still there today. The funny thing is, even if it were just one of the 3, when we met, as I mentioned, if a $600 million company were collaborating, people in the past would have considered that size significant, right? But that's not the case now.

Adobe is huge. It seems like Adobe isn't taking proper steps in the AI field, is short on funding, and is just moving personnel around the C-suite. CEO Shantanu Narayen announced his retirement a year ago, but deciding which of the 2 internal candidates to promote took a whole year.

Honestly, the situation is pretty bad. I think it's a trivial matter. Do you know what I mean? It's just a sign that nothing is going to change. It's a sign of surrender. It's a sign that we're just going to keep pushing the world in the same way rather than truly changing.

We're sacrificing the core product for high margins, adding AI-powered image-processing features, and then wrapping it up, right? I completely agree. That's interesting.

They focused on securing free users rather than ARR growth, which is a strategy from the 1990s. It was a strategy of putting forward fake AI metrics. The point I was trying to make is that, as I asked you recently, it might have worked 2 or 3 years ago, but it's a problem now. That's because net new ARR has decreased significantly. In other words, new ARR growth hasn't occurred. AI-based ARR metrics have risen, but overall ARR has fallen. This is investing only a part in the good parts.

It means that it has been done. However, achieving success through the program is just the beginning. Due to Jason Lemkin's Law, which I want to mention again, it is not a complete success. If you don’t grow, you don’t win. If people don’t pay, you don’t win, understand?

Yes, the scale has grown, and it won’t disappear—$25 billion in revenue, something like that.

Hmm, but yes, there hasn’t been much news here. The interesting point is that the software industry as a whole has had a pretty good time over the past few weeks. Cybersecurity-related stocks have risen significantly, and the cloud industry as a whole has also climbed sharply. The crisis in the SaaS industry has also been resolved to some extent.

But, Jack, the point is that it has become clear which businesses will not succeed, which businesses should remain undervalued, and which businesses are doing incredibly well. Adobe is closer to the first category. Everyone, today Adobe’s market capitalization is $105 billion—about $100 billion, with $25 billion in revenue. Starting with cash flow, what will the market capitalization be in 3 years?

You ask what the market capitalization will be in 3 years? Around $120 billion or $130 billion.

Wow, that’s surprising. It’s exactly the same as today.

It will be exactly the same as today in a month, too. As you know, it won’t disappear. Basically, as far as I remember, the cash-flow multiple will be less than 10, Harry. So, as long as ARR doesn’t vanish entirely, you could probably get a new cash-flow multiple. But that wouldn’t be the case.

If Jason’s worldview and Menlo’s worldview are correct, and there are 10 or 100 companies worth more than $25 billion, their relative importance will decrease. Honestly, I don’t really know what HP’s market capitalization is right now. It’s an important issue, isn’t it? It’s the same thing.

They say, “It’ll be fine, and the cash flow will be good. If you find the right leader and product, growth will come back.” It won’t go bankrupt, but the stock price will trade at 8 or 9 times its cash flow. It has been that way for decades.

Adobe’s stock price barely fluctuated for 10 years. It traded based on cash flow. Shantanu Narayen was truly remarkable in that regard. Then, the cloud worked much better than expected. Adobe will either stick to that strategy for the next 10 years or develop magical AI technology that even Canva hasn’t caught up to yet.

I can’t guarantee it right now, but I was there when the cloud era arrived. Adobe didn’t foresee that success. Adobe developed killer AI applications, while we are still in the early stages of the AI field.

So, which would you invest in today, Canva or Adobe? Let me tell you why you shouldn’t invest in Canva. It failed miserably, didn’t it?

I like them. You can’t help but grow.

ServiceTitan completely failed after lowering its earnings forecasts. It dropped 30% last week. I think Canva also dropped about 30% to 40% this year when it lowered its earnings forecasts.

Right? It might have dropped even further.

After all, it isn’t a publicly traded company. But it has to grow. At least Adobe will get on a stable trajectory, just as Rory said before. It will get on a stable trajectory. It won’t be an impressive growth rate, but it will grow by 6%, 8%, or 9% annually, generating abundant free cash flow.

Let me show you Canva’s earnings. They are growing by 20%. At one point, they grew by 30%, but now they are growing by 20%. I know because, in Australia, they have to report their revenue to the tax authorities.

So what? That is a significant slowdown in growth.

16. What Happens When Growth Companies Become Value Stocks?

I agree. Growth hasn’t stopped, but the growth rate is slowing. That is why I am debating whether to put it into a DCF analysis or still view it as a growth stock. Is it a growth stock or a value stock?

Yes, I agree now. The reason I interrupted is that I did not agree with the expression “growth has stopped.” To be more precise, I believe that growth hasn’t stopped; I meant that the growth rate is slowing. That is a really important point.

I need to think about this part a bit more. Various factors interact in a complex way. Switching from growth stocks to value stocks is a really difficult task because the valuation method shifts from a revenue multiple to an EBITDA multiple.

I’m reminded of a post by someone named Gokul on Twitter. The point was that you should use a revenue multiple if growth is over 30% and an EBITDA multiple if growth is less than 30%, which was excellent. That is an insightful comment.

The point is, if you grow quickly, all shortcomings are forgiven and you are evaluated based on a revenue multiple. But if you grow slowly, nothing is forgiven, and you are evaluated based on an EBITDA multiple or, like us, a low revenue multiple. It is a completely different world.

I’m reminded of a post I wrote in the past. Box, as a listed company, also went through and overcame such a transition period. However, it takes 3 to 4 years. If it goes from 6 or 7 times revenue to 20 times cash flow, then, to maintain the stock price, cash flow needs to rise by 30%. It is a really tough process. It took a tremendous amount of time.

Getting back to the main point, Canva is still growing well, but I think the growth rate is slowing down. I like Canva and hope they succeed, but Canva is a founder-led company. On the other hand, as you mentioned, Adobe feels like it is exploiting me every time I buy a product. I am using their licensing system too much. That is the harsh reality.

If I, and you too, have to go through such a transitional period, it will be a really tough time leading up to an IPO. Harry, your question about the 2 stocks is basically asking whether you would hold a stock that has at least secured liquidity—showing a 13% growth rate at 8 times cash flow—or a stock with a 20% growth rate but with the potential for slowing growth. The key is whether you can overcome the hurdle of securing liquidity.

Ultimately, price dictates every market, and there is also the concept of relative valuation. What are your thoughts on relative valuation?

I think the question of whether to do so is an interesting one. However, transitioning from growth-oriented valuation to value-oriented valuation is a truly difficult task. It is not easy even for a privately held company, but it takes a long time to overcome that hurdle. During that time, nothing can be done.

The same applies to publicly traded companies. It is extremely difficult to make the transition simply because a company is publicly traded. Investor interest wanes, and you face various difficulties. While the situation is dynamic, it is a truly rugged path.

Remaining private for a long time carries such risks. Stripe avoided that risk by halting its growth trajectory. I sincerely hope Canva can also halt its growth trajectory because I want the founders and good people to succeed.

However, what if Stripe had lowered its growth rate to 20% or 15% instead of halting it at 40%? Of course, it would have been fine since they still possess massive cash flow. Slowing down the growth rate and having the company valued based on cash flow brings about a massive change in corporate valuation metrics.

When you think about Stripe, it seems like the only company that was able to weather that storm. That is because it generated so much cash. You did a great job.

Now, I guess I should hand the reins over to Jack. He is the superstar of Dreamforce. Sorry, Jack, but you have to become a superstar. You are going to become a superstar alongside Marc Benioff, aren’t you? Right, he is the biggest star.