20VC：为什么“追赶前沿”是胡扯｜Instinct 以100亿美元估值融资10亿美元，Meta推出 Muse｜Miro 在175亿美元估值后以13.5亿美元出售｜Mistral 融资30亿欧元，Sam Bankman-Fried 能否重获自由？

Now, I was thinking about where we should start. In all honesty, I thought it'd be egregious not to start with the most important thing, which was Dario coming out and saying that we need to pace the frontier, to which Sam Altman then agreed with him, and then Elon also agreed that it is important now to put in place some form of external regulatory body to slow down and regulate the capabilities of model providers moving forward. How do we think about this?

1. Dario Calls For Regulation

It's been fairly universally panned. I tend to be on the side of the people who are saying that the hostile reception is deserved. What is the problem you're trying to solve? Dario, if you read the note, Dario's definition of the problem was fairly well-defined. It was cyberattacks, it was economics, and then it was, “We lose control of the AI.” So even though that sounds like a lot, it's fairly controlled.

The guy who quit and started this crazy thing as a co-founder, whatever his name is, and then the employee within Anthropic chiming in and saying, “I think there's a 10% chance of human extinction in 10 years”—that's a very different thing. So let's deal with that first of all, because I'm just going to call such bullshit.

To be fair, and that's why I'm saying it's complex, every time Dario makes a comment, a lot of the tweets back are, “Hey, if you're going to blow up the world, you should stop.” And they're right. The truth is, he didn't actually say they were, to be fair. He didn't talk about p(doom). But I'm going to do the same thing as everyone else, and I'm just going to throw some rocks first.

If the feds really thought that there was someone in downtown San Francisco building a technology that was going to blow up the fucking world, that had a 10% chance of blowing up the world in the next 10 years, they would move in with a SWAT team, kill everyone in the place, and close it down. Replace the word AI with, “We're building a nuclear reactor. It's totally safe right now, but there's a 10% chance it goes wrong in the next 5 years and blows up the world.” Issue that statement: how many seconds after that before the entire weight of the U.S. military is shutting you down?

What it says is this: these 10% p(doom) people—the truth is, either the U.S. government is asleep, which I doubt, or it's looking at this going, “This is a bunch of excited teenagers. We'll step in later if it gets crazy.” My first point is, to be fair to Dario—not Dario, but the “Oh my God, the world's going to end. We're going to destroy the world, but we're going to keep doing it” people—I cling to the hope that if it was a real issue, the U.S. government would do something, right? So I think all that's overwrought bullshit, which is different from saying what Dario said is overwrought bullshit. It's just not realistic.

Dario's stuff, on the other hand, it's hard not to agree that the cyber risk is real. Of the 3 risks he raised, the cyber risk is real, the economic risk—that we'll all be unemployed—I think is a little bit bullshit, and then the 3rd one, that we lose control of the agent, is hard to assess.

So, okay, that's the problem he's trying to solve, and then the solutions range from the unlikely to the impossible. The unlikely is, yeah, we're going to install third-party monitoring agents voluntarily, but then the ask is that it be made a requirement. That's the first thing.

Then the second thing is that all the democratic governments have to agree, and that'll be easy because we get on so well with Canada and Europe right now. And then the third thing is that we have to agree with the Chinese, and that's going to go super great, though it might go better than with the Canadians because we like the Chinese more than the Canadians. We should put in the Russians. We get on great with them.

So, stepping back from the bullshit of sides 2 and 3, right? Just proposal 1, which is that we're going to introduce these kinds of third-party monitors. If it's voluntary, knock yourself out. You do whatever you want, Dario, right? If the government decides that they should do something, which is different from Dario thinking they should do something, then it's not going to be your friendly regulator. You're not going to get to choose that. It's going to be a law and regulators, just like bank regulators.

I'm not sure that's a great idea. I'm not sure it'll work really well for innovation. But if it does happen, a lot of the comments were either, “This is silly, we don't need it,” or, “If we do need it, well, who the fuck are you to tell us what we need? We're a government.” I'm kind of sympathetic to both those responses.

Sorry, that was a vent, but I just think it's so overwrought. Sorry, Jason, I just went on.

No. My very tactical view, and then I'll give you my—my very tactical view was this was just a risk factor in an S-1 done live. My tactical view is Anthropic's going public. He's just, in part because of the employee that worked there 4 weeks or 8 weeks who said there's a 10% risk of destroying humanity, which many agreed with, getting ahead of a risk factor so that when the $2 trillion IPO happens, it's a non-issue.

I honestly think he's enunciating a risk factor. We're ahead of it. We're going to debate it as a society, and so when we go on the roadshow to New York and everywhere else, no one cares. It's not even cynical. I think it's your job as CEO. That was less discussed.

I'm sorry. You can't say—I mean, words have meaning. If you read it—and again, I'm going to be fair to Dario—he didn't say there was a 10% chance of blowing up the world.

He responded to it.

Yes, you're right, he responded to it. I feel like if the government was doing its job, we would convene a congressional committee, we'd subpoena Dario, and say, “Your head of safety”—not the guy who quit, Jason, this is an important point, not the guy who quit, but “your head of safety said, ‘Me and many of my people think there's a 10% chance of blowing up the world.’”

Get a fucking congressional committee, subpoena the guy, and say, “Mr. Dario, as the head of this organization, do you believe that your safety commissioner is correct and that there's a 10% chance you're going to blow up the world in the next 5 years? Yes or no?”

Right, but it's going to happen. This just started. You're acting like the feds aren't going to raid Anthropic, and I don't think it's literally going to happen.

The feds are going to raid Anthropic. Conceptually speaking, this just happened last week. How many texts did you guys get from folks outside of tech asking if AI’s going to kill us in the last week? I got texts from people, relatives I haven’t heard from in years: “Is AI going to kill us, Jason?”

There will be congressional committees in hearings ad nauseam for the next 24 months. We have just awoken the sleeping giant here that AI isn’t about a bunch of folks in San Francisco becoming centimillionaires. The public is going to believe it’s going to kill us, and I think that’s why Trump cut it off so quickly. I don’t want to go too much into it, but I think that’s why Trump said, “It’s not an issue. We’re going full bore.” I don’t think that was out of nowhere. I think this is because there are going to be 2 years of congressional hearings. This is all they’re going to talk about.

To be clear, I think you are right. Maybe I should’ve said the future tense. We’ve kicked off a boulder cascade here, whereby, you’re right, if I’m an ambitious politician, I would do this. Maybe you are right: there are 2 risk factors in the S-1, which is what you’re saying. It’s a really good point.

The first risk factor is, by the way, there’s a 10% chance we’re going to blow up the world. We can discuss that one; that’s a fun one. But the other risk factor is that I employ people who believe we have a 10% chance of blowing up the world, and I do this because they’re actually very motivated to build great AI. For whatever reason, that motivates them. We can talk about that with our therapists.

Because I’m running this company where these people say this crazy shit, as Jason’s just pointed out, I’m probably going to spend the next 2 years being investigated. There are a series of bad things that could happen because of that, up to and including the government shutting me down. You’re right, Jason; that’s the risk factor. I agree with you. Yes, it’s going to be the weirdest risk factor ever: we say dumb shit, so we may get shut down.

There’s so much going on here. It’s hard to track, right? I would just add 2 things that I thought were the best things said. One may be a little political, and one I think is nonpolitical.

I think the best thing actually was David Sacks this week, who said, “If it’s that bad, Dario, it’s your effing job.” Whether it’s 10% or there’s any material risk that your company is going to exterminate even a subset of humanity, that is your job. Your job is to fix it. Shut down the company if you can’t, right? If you’re unable as CEO, bring in a new CEO. This is your job. If there’s any company that had these sorts of odds, it is your job, right? This is product liability 101. You can’t kill them. There are only so many people you’re allowed to kill with your product.

And Jason, it is me—first of all, you’re exactly right. I’ve got to hand it to Dave; I think it was spot-on, 100%. I think it was really odd to see Lina Khan, who is probably the antitrust regulator most of Silicon Valley hates the most, come out on exactly the same point.

When you get David Sacks and Lina Khan both on the same side, and both making excellent points, it’s a really fun issue. I agree with you.

2. Recursive Control Becomes Central

I’m sorry, I’m naive. If you build recursive self-improvement in the way that people talk about it, doesn’t it become relatively unknown what it becomes? I could have good intentions for it, but it could become something else or be used by malicious actors. It could be weaponized into something else.

Being really clear here, I agree. That is actually the only one of the 3 risks itemized in the Dario thing that I think is interesting. The other 2—economics—I think it was a foolish thing to say for a whole bunch of reasons. You can’t say, “I’m not going to invent technology because it puts people unemployed.” We’d still be back on our farms, with 73% of work in the farms. That was a dumb point, just trying to be nice and sucky-uppy.

The whole cyber thing is real, but that cat’s out of the bag. The commies have the cyber, so we’re done. The only thing that was a fair point—and, on the other hand, the P(doom) thing, he didn’t talk about—the only thing in his letter that was actually simultaneously, “Yeah, that’s a fair point, and I don’t have the answer,” is exactly that we could, quote, “lose control of the things—recursive self-improvement,” that kind of stuff.

You’re right. Of the 3 issues he raised, from the 3 responses he gave, that was the one where you go, “Hmm, okay. I can’t…” Jason’s point, I think Jason and David’s point, is spot-on at that point. What you’re saying is, forget the other 2, which are about other people doing things with our technology; we can come back to those. What you’re saying is you’re building something you can’t control.

If you’re building something that you can’t control, then maybe you should stop building it. You’re the CEO, and as Jason, again, I’m going to agree with Jason and David, the third party can keep an eye on you. But if there are people cleverer than you who can figure out a problem that you can’t figure out, you’d have hired them years ago. We should probably assume you’re the cleverest people doing this.

Either you think this is a manageable problem, in which case, keep building, or you think it’s not a manageable problem, in which case, you probably should stop. But stop wringing your hands and saying, “If only the rest of the world could stop me killing humanity.” You’re the CEO. The cacophony of anti was really good there.

But you’ve also taken it to a level where you’ve weaponized a Chinese economy to have an open-source ecosystem that’s incredibly strong. Now, to go mea culpa, let’s put the brakes on. You’ve weaponized them enough to be a serious cyber threat to all of our institutions, and now you want to put the brakes on.

First of all, I’m going to push back on “weaponized.” You’re basically saying, “They wouldn’t have had these models without us doing it.” I’m not sure that’s the case. You can get into how much of open-weight models has been distilled, and I can’t assess that. So there’s some of it.

But they have smart people, too. I think there’s too much attribution of godlike status to individuals. The truth is, independent of what Anthropic probably did, there were probably going to be a bunch of LLM alternatives in China. This technology—the cat’s out of the bag. It exists. Your main point is correct: it’s out of the bag, so you can’t…

But it’s out of the bag because you let it out of the fucking bag. Sorry to be blunt.

Well, I don’t know if he just did. You are the best at it. You are, yes. I hate the atom bomb metaphors because they’re like Nazi metaphors. They’re so crude and simplistic, and in this case they actually feed the ego of everyone involved.

But it’s a really good metaphor here, Harry, unfortunately. You can say, if these folks who were the Oppenheimers in the U.S. invented the bomb, there’s no doubt that the Russians had the bomb 4 or 5 years later—or 3 or 4 years later—and in part because they stole our stuff. They had spies in Los Alamos from the U.K.; let me remind you, Hans Fuchs. Then they stole our secrets and built a bomb, too. But what are you going to do, right? They probably would have got there anyway.

But Elon has been clear on that for 3 years. He said he wished that AI hadn’t happened. He said, “But since it did, I’m going to do it anyway because it’s too late.” He didn’t just say this last week. He’s been saying this for 3 years, that it’s gotten too strong. Before xAI even got anywhere, he was consistent: “I don’t want to be doing this, but it’s already been done. It’s too late.” I don’t think we should have this much progress, right?

In fairness to the CEO of Anthropic, some of the most hyperbolic stuff he’s not saying. But you have that pause letter that a bunch of scientists signed. There’s a lot of hyperbole that’s not coming from him, and he’s a little more nuanced on what he is worried about. The only one that’s really meaningful is this kind of losing control of it, right?

But I think this is where his prior messaging has come back to bite him on the ass. To Jason’s point, in previous episodes he was so dogmatic about what it’s going to do to jobs and employment. Now, when things do come out, it’s tied to him, sometimes unfairly, just because he’s had that stance and that labeling before.

Yes. To Jason’s point, which I think is spot-on, my God, is this unpopular in the rest of the world. It turns out, if we’re building a technology that’s definitely going to cause 20% to 30% white-collar unemployment, and my VP of safety, whom I haven’t fired, has said on Twitter there’s a 10% chance it blows up the world, I wonder why we’re not popular. Let me give you a few quick clues.

Can I just say what I thought the best thing I read on all of it was, on Twitter or anywhere? It was Jay Kreps, who was the founder of Confluent, acquired by IBM for, I don’t know, $12 billion, and just stepped down. This is the best thing I thought was written on all of it.

He said a lot of people have fake and stupid takes. It’s a lot of marketing. He said, “Here’s the captain-obvious point: most positive use cases for AI have a corresponding dark version. If you’re superhuman at coding, you’re superhuman at hacking. If you’re superhuman at structural engineering, you’re likely superhuman at finding structural flaws that knock buildings down. If you’re even superhuman at designing drugs, you’re superhuman at designing novel, undetectable poisons.”

If you cure viruses, you can create them. Some of these are not that bad. They’re manageable. Others are scary.

It is a fact. If you don’t get guardrails right, and there are no guardrails in open-weight models, for all intents and purposes, dark versions can be created. Your AI will be just as good, all things being equal, at the dark version as the light version. It’s a fact. We need to solve these issues, but it is a fact, and it’s not even—there are so many opinions and anthropomorphizing and 10-percenters, but the dark versions will escape into Hugging Face.

I do agree with what you said, Jason. Absolutely right. When people invented books, when Gutenberg invented printed books, the Catholic Church was pissed because they really liked having control over all knowledge, and they just didn’t like information dissemination. Ditto the internet. There’s a reason every totalitarian regime hates it.

I think it’s great framing: every technology has a positive side and a negative side. We’re going to do what we do every time. We’re going to roll out the positive and find a way to manage the negatives.

But in the interim, we’re not going to sell chips to China, which of course produced a Jensen response like you’ve never seen—the usual Jensen response. So, on the implementation side, not realistic.

Do we see ripple effects across the infrastructure layer, in public and private markets?

3. Markets Shrug Off Extinction

There was a 1-day minor hit on the semiconductor stocks that first day. Separately, pleasingly, CrowdStrike and all the cybersecurity stocks jumped 10%.

I looked at Monday, and we’re recording this on Tuesday. It’ll appear on Thursday, which is a whole lifetime away in the world we live in now. The instant response was a slight markdown on semiconductors and AI CapEx, but not a lot. So, no discernible slowdown. I mean, it was maybe a slight one, which implied some level of slowdown.

There was a significant markup on the cybersecurity stocks because it looks like that’s where the actual reality of the problem exists. Overall, I always look at WCLD, which is the software index, versus SOX, which is the semiconductor index, and it was a great day for WCLD relative to SOX. Software was up, semiconductors were slightly down, but not by much.

I loved a tweet that said, “Market moves 0.1% on news of a 10% probability of extinction.” In other words, capitalism took on board the risk and said, “It’ll be fine.”

I mean, this may be a much faster version of the 1920s. There’s no regulation. Everyone just wants to get rich as fuck in the 1920s, right? What our version of 1929 will be for AI, we don’t know yet.

Everyone just wants to get rich. $2 billion is just a Series A round today. I was literally thinking that today when I saw Shield AI raise at a $20 billion or $30 billion valuation, and they had a little chart. I’m like, “Wow, $2 billion is a Series A round now in today’s world.”

It’s just so much money, and AI is so much money, that it’s too easy to take advantage of the dark version if that’s how you make money. It’s too easy because every LLM has a dark version. It’s too easy to cut that corner if that’s how you can get raised at a billion after demo day. It’s too easy.

I don’t think that’s actually what happened. I actually think a different version of the same thing is this: everyone’s saying, “Oh my God, this is evil and bad and could be dangerous.” As it were, they’re purging their conscience by worrying.

But Jason, you’re exactly right. No one’s saying, “I’m going to push away from the table.” I don’t think the people running some of these companies, especially Anthropic and OpenAI, are profit maximizers. Oddly enough, you’ve got to defend them a little. If they were profit maximizers, they would own more than 2% in the case of Dario and 0% in the case of Sam.

It’s the other folks that we should be worried about, not the 3 of them. We actually have half-decent stewards at the top. Sam, Dario, and Elon are about as good as— they may be better in theory, but in practice, we can’t get 3 better stewards. They have the right reasons to do what they’re doing.

The real problem is that 10% or 15% of founders are sociopaths. They genuinely are. In fact, especially the successful ones. It lets you will something out of nothing into existence: the ability to manipulate people, the ability to have that look, and do it.

That’s where the bad people using AI will come from: the sociopaths. You really don’t think 10% of the founders you’ve ever invested in are sociopaths? Of course they are.

4. Muse Enters The Assistant Race

We’re going to move on, but fantastic. Next, we have the AI assistant race that we touched on last week. I was really annoyed because we actually missed Meta releasing Muse, which is Meta’s product that is an AI assistant in many ways.

What did you say, Howie? We didn’t miss it.

Well, it came out after we recorded the show.

Correct. I’m actually going to give us an A. I’m going to give us an A. We recorded on Tuesday because you used that clip of me. On Tuesday, I was like, “Somewhere, there are 20 Meta engineers locked in a room being told to ship something.”

We said that on Tuesday. On Wednesday, Muse shipped, and on Thursday, the pod shipped. So I think we nailed it, man.

You just didn’t know it was 500, not 20. That’s the only slight thing that was missed.

You didn’t realize that the minute OpenClaw took off, Zuck took a huge chunk of his AI team and said, “We’re building OpenClaw for consumers.” From that night on, people worked days and nights. I was logging bugs over the Muse weekend on Twitter for fun. The engineering team was responding in real time Saturday night, Saturday morning, Sunday morning. They’re working 9–9–6 on this thing.

This is not 20 people. This has been a top priority since OpenClaw launched, right? It’s interesting that it came out seemingly oddly after some of the other agents. But this has been a P1 since OpenClaw, right? People were lying in the streets with their Mac Minis trying to figure out how to run an agent. Now I can do it on Facebook.

Well, let’s actually just start with this, Warren. I think you’ll be as interested as I am by Jason’s analysis.

Yes.

I’m not able to use it, being in the UK, so I’d love your thoughts. Jason—

I’ll get you a VPN.

Yeah, thanks, dude. What did you think? How good is it? How good a response is it to Instincts? What do we think?

Well, first of all, as software, it’s very, very, very good. It instantly works. This is the definition of great software: you just can’t believe how well it works because all the hard work was done that you can’t see.

Most of the things you want it to do—create a reservation, send an email—just work. I asked Muse to send us a bunch of stories for this show. It sent them to me and Harry, and I forwarded them to Rory. It did it at the right time. It sent them to us, right? I had it rebuild the entire SaaStr.com website for me. I had it log into WordPress and redo it. It did a pretty good job of it. A lot of the things you want it to do just work.

I’ll add one more thought and then the Meta question. The really interesting thing is—we didn’t talk about it, but it is something I know a little about—it’s not cheap, okay? You’re giving everybody, up to a certain point, a free VM, which I think has 2 CPUs, 2 GPUs, 8 GB of RAM, 100 GB of storage, or something.

I do know that, for Replit, Lovable, Vercel, and others, it costs about $3 to $4 per person to deliver that, and Muse is at the edge of it. It gives you more than they do, okay? Those companies are working on it every week because it’s such a huge part of their COGS, right? Every time you spool up a website—not a free one, but a paid one—they’ve got a $3 to $4 nut.

Wix, before Base44, has a 2-cent nut to serve that website. Now it’s $3 to $4, so they’re very incentivized every day to work that down, right?

Meta is lucky. Not only does it already have the infrastructure, but it has tons of infrastructure. We could argue whether it’s free, but it has tons of infrastructure. This is running on its own LLM, Muse LLM. So it has a massive infrastructure and LLM benefit that no one else has. That’s why it’s fast. That’s why it works well. That’s why you get more VPUs, more GPUs, more everything.

So I think from an infrastructure perspective, almost no one can compete. You get all the VMs, all the infrastructure, all the storage, and they have their own LLM, Muse. My learning is that for these lay usages—not frontier drug discovery—Muse LLM is really good for this.

But the question is, does it matter?

What do you mean, does it matter? Sorry.

Yeah.

5. Muse Needs A Killer App

What’s the killer app? Another reservation at a Cheesecake Factory or TGI Fridays? We need to see the VisiCalc of Muse. We need to see what the killer app is.

Every horizontal platform traditionally needs some sort of killer app, right? I’m just skeptical. I’m just wondering. I don’t think there was a killer app for OpenClaw. I don’t know if there’s a killer app for Muse or Manus, because there wasn’t one for OpenClaw.

Is it not your discovery mechanism for shopping? Like a WeChat, a super app. Zuck has spoken about taking portions of transactions as being the business model as well.

The model’s good, but what are his examples? Scheduling his daughter’s carpool? Is that really what a trillionaire needs to do?

He's making up consumer-y things: “I had to schedule ballet lessons for my daughter.” Great, but you need to run this thing 8 hours a day, I think, like Claude Code or Codex, to really matter. If we're in it 8 hours a day, like a super app, it's cool. If it's a random task, I'm just waiting to see what the killer app is.

Muse is really fun. It's so beautiful. The beauty is that it gives you all the ideas. It has an Idea tab, and it tells you all the things to do. I've done most of them, and they're great. I just don't know if it's killer.

I don't know if it needs to be, and I know that sounds stupid.

No, but I use Instinct in a similar power-user way. I think latency's a real problem with Instinct, by the way, which might not be a problem with Meta. I wait minutes—

Meta's faster.

—for responses.

Yes.

Minutes. It's like the first days of ChatGPT. It's a real problem.

But I don't know if it needs to be. It does all of my bookings, travel, and restaurants. It does all of my shopping. It's very good. It does all of my calendar invites. I know there's no killer app, but it's just incrementally better than everything else. I don't know, but it's lots of little bits—not good enough.

I have a feeling it might be. I think, 1, I'm skeptical of the category on a standalone basis, but I always have my 3 venture questions: Is this a category? Who's the winner? And are we getting paid for the risk?

Is this a category question? Is there a role for AI in a personal-assistant, messenger-type thing? Having used both Instinct and Muse, I can see it. I don't know if I love the words “killer app,” because a single thing that does everything, that ignites a platform like VisiCalc—let's leave the killer-app concept out there, right? At the margin, you kind of go, “Yeah, I'd use this.”

You do have the example from long ago of WeChat, Chinese messaging systems becoming very much a super app. Facebook tried to do that with Messenger. It didn't take off because the truth is, pre-AI, the UI of trying to book things on a chatbot is a pain in the butt. It's actually a lot easier just to go to the United website, see all the flights at the same time, and book it.

It is possible that, with intelligence at the back end, as this thing moves from non-intelligence to intelligence, you actually have an interaction on a mobile device where enough of the intelligence is in the cloud that it can do a lot more for you, it knows a lot more about you, and it just becomes an easy place to get shit done. Therefore, to your point, Howie, at the margin, you can imagine people using this.

If you're a Facebook user, you get Muse, you're a happy little camper, and you chug along. Do I think there's more that's going to happen? Yes. To me, the interesting question—the 2nd one—is, do you think it's a standalone company that wins here, or do you think it's going to be Facebook with Muse and OpenAI with whatever product they come up with?

Do you think the Instincts of this world can build a standalone business, given that—just to put it out there—they raised at something like 50 pre less than 5 months ago, then 400 pre, then a couple billion pre? They're now, rumor has it, raising at north of $10 billion.

They are raising $1 billion at a $10 billion valuation, and the most popular segment of the show that we got was Jason's IC. So, Jason, we have a billion-dollar round for Instinct. Welcome to the partnership meeting.

Thank you.

It's a $10 billion valuation. Will we be putting in $200 million into this billion-dollar round at $10 billion?

6. Instinct Faces Meta Competition

Well, we will, and I'll be honest with the team: This is a risky one. I am impressed with what Muse has done, and if we had to compete head-on with Meta, and that was the only thing, we'd be in trouble. I would not recommend this investment because we can't compete with their balance sheet. We can't compete with their servers. We can't compete with their GM. They can't compete with the LLM.

However, after having done several reference calls and over a dozen synthetic ones on Claude, I've learned a couple of things. First of all, Meta cannot go cross-platform. It does own WhatsApp, which is a real threat, but it won't work with all carriers. It won't work across all services. It's highly focused on its own platforms.

That is only a subset of how we communicate. How many folks between the ages of 18 and 55 are on Facebook all day? Very few, right? It's really your grandmother's application. Now, WhatsApp is popular, and Instagram is popular, but the fact that it is not going to be interoperable across all these different services means it's got a fairly limited reach.

2. How long will Meta maintain the energy here, right? When this produces essentially trivial to no revenue, anyone remember Workplace? Workplace, in its own way, was probably better than Slack for folks that lived in Facebook. It was probably better. It had a lot of neat use cases. It worked well. It was architected, and it was actually the highest NPS of any product in the entire Facebook-Meta platform, but they couldn't maintain the energy.

My reference checks say when Alex leaves Scale, this product will fall apart. If this was the only thing that Meta had to do, I wouldn't bet against them, but we're just not going to see the commitment to do the kind of things the Instinct team is going to do.

Let me tell you, this is 1 of the greatest teams I've seen in my history of investing. They're great. These kids come and play World of Warcraft in real time during the pitch. They're both top 15 in League of Legends. I recommend leading the round, but being cautious with reserves because the next-round valuation may hit realistic IPO limits.

So you would do this round?

I wouldn't. You asked me to do it. There's no effing way I would do this round.

Okay, good. Okay.

No effing way.

We've got to use that whole clip.

Here's why I wouldn't do the round, okay? You can call me a fuddy-duddy for it, okay? I believe most of that for the pitch. I just believe the infra costs here are so high, and the incumbents—

This is like if Claude—what if Anthropic and OpenAI actually built apps? In the entire history of this show, they've only built—outside of Codex and Claude Code—they've really only built half an app. Claude Desktop, which isn't even really a full app.

Here's Meta building the app. They have the LLM. They have the cost advantages. They have the speed advantages. They have more available compute and GPUs than anybody else in the world, and they're building the app. This is the threat that every VC worried about, and we all got a hall pass since the start of AI because the LLMs didn't build any apps. This is the 1 they're building.

I just don't want to compete against this because I do think, for the next 24 months, it's a big priority. Harry already said Instinct is slow. That's a sign. That's a sign of compute costs.

Can they subsidize with venture capital at $5 to $10 per user per month? Sure. But if they have to over-monetize it, what if they become Poolside? They could become the next Poolside. It's great, but literally, I've got 10 million users at $10 a month. Now I've got a $1.2 billion nut a year to pay off. I'm struggling to raise the next round. I'm not Databricks.

I worry when the incumbent has infinite capabilities here and wants to build the app. That's why I would say no, but I might be wrong, right? I might be wrong. I wouldn't bang my fist on the table at the meeting.

I'm trying to remember, was it Socrates or who was it? 1 of the ancient Greek philosophers where you could literally say, “Take 1 side of the argument,” and then halfway through you could say, “Now take the other side of the argument,” and Jason clearly can do that here, right?

That was a perfect, “This is why you should write—”

Harry put me on the spot. I don't want to do this to you.

He did put you on the spot, but I'm just impressed with the mental facility with which you can do both sides. It's terrifying. It's like a human LLM. You can be convincing on whatever you want me to believe.

Pulling them both together, I actually think—and I want to take the Poolside analogy on that—what you basically said is this is core to Meta, and it feels like something that they would want to do. Despite your comments on cross-platform, today Muse is a standalone app, so it's not a cross-platform issue. But to the extent that they fold it into Messenger or make it accessible on iMessage, et cetera, I've just got to assume they're going to do that.

It seems to me that if there was 1 thing that Meta should do within the world of AI, it would be this. It's hard to imagine spending $100 billion-plus on AI, saying that we're going to build the personal AI, and then not putting all your effort into this. So I agree with you, Jason. I think this has to be an all-in Meta bet, in a way that, frankly, I reject the comparison with Slack.

The Facebook for Work product was like a toy. It was not a core issue to them. This is a core issue. So I think you're right. They go for it, hook, line, and sinker.

Now, the interesting thing about the Poolside analogy and Instinct is you could have the same outcome here, which is Poolside said, “We ran out of capital to keep playing,” but we had an excellent outcome because there was a company with an even bigger market cap that wanted the assets Poolside had assembled.

NVIDIA wanted access to the model, access to the talent. I mean, let's put it out there: the same thing could happen here, which is that Instinct executes, builds a huge user base, and OpenAI steps up and says...

It's interesting. I think that the founder of Instinct either worked at Sierra, or Bret Taylor's a big fan of him. And Bret Taylor's obviously, among the many other things that man does, the chairman of OpenAI. It may well be that Instinct builds a lot of traction, doesn't have a cash-flow-positive IPO potential, but has a very attractive upside exit. I don't discount that. Maybe $10 billion is a little lofty, but you've got to believe that if it got meaningful, differentiated traction, it would be interesting to someone who wants to build a business in this space.

The only thing I will say is that Noah Shin, the founder of Instinct—every single person I've spoken to cites him as one of the most generational talents. In a world where generational talents in strategically attractive segments are very attractive to multi-trillion-dollar companies, you can see that being a very legitimate upside scenario for $50 to $60 billion, as crazy as it sounds.

It's possible. Listen, we all have different experiences. The way I was raised to invest in venture was: don't take bets that 100% require an M&A outcome to be successful, right? They're just too unpredictable. I've been on the other side and know how capricious it can be. You could be Clem's best friend. He brings you in to meet with Jensen, and then he quits the next day, right? You literally just can't predict it.

You need some real something to steal to make a bet where an arbitrary, super-high-value outcome is the only plausible exit. I'm not saying it's not a bet. It is definitely a bet, right? It's just a big one.

And just to come back on that, because I've been thinking about that. You're right; it's not what I do either. But you always question yourself: is there anything you can learn? There is an argument—I'm not yet making it, but I'm just acknowledging—that the expected value of a number of those bets could be strongly positive. Admittedly, the variance is high. In other words, it's a risky way to make money.

You should ask yourself, as you're building a portfolio of 30 bets, is it okay to have 3 of those bets in your portfolio? Probably not at $10 billion, but it wasn't reckless. I think Kleiner did a round at $500 million. Moonshot was very shrewd. It's not reckless to do a round at $500 million, even if you believe the payout profile is a 2-in-10 chance of a 20x positive multiple. And if you don't get the positive multiple, you're going to build a company that just can't cash-flow and doesn't make it.

For sure. If you've got the right portfolio and you can take the risks, yeah.

It's not the way I will run my business, but I'm in a bull market, or as Harry said, there's a whole bunch of upside acquirers with, frankly, free market caps that are fairly untethered themselves and have a massive need to move quickly. It's actually not a crazy way to make money.

And you don't even need much. You've got a billion and a half pref to reach. So your downside is relatively capped on an incredible team that everyone acknowledges is industry-leading.

The classic Silicon Valley thing is to examine that risk-return profile at $50 million pre, which is as far back as April, I think. That's a wonderful profile. At $500 million pre, which I think was the May or June round, that's interesting. Yeah, that's a good bet because they're a little further along.

The interesting thing is, in the space of 2 or 3 more months, you've had a round. Was the other round $2 billion, Harry, from memory? I can't—

$2.5 billion. Yeah.

$2.5 billion. Now you're getting into the—hmm—it's a 4x if you get out at $1 billion, and now you're raising at $10 billion. Now you really need that $50 billion or, you know, you need a...

A reminder: the largest M&A outcome ever was Cursor doing $4 billion, getting $60 billion. I don't know if you get a $40 billion outcome without a shit ton of revenue. My point is, the risk-return profile was wildly attractive in April and decreasingly unattractive fairly quickly by the time you come to September. So that is the problem with those kinds of bets.

You know, one interesting thing, though, is just how wild the consumer product-market fit for this product is. I did an Instagram Reel on it, and I had over 1,000 DMs asking for invite codes. I've never had 1,000 DMs on the back of a Reel. That's pretty wild.

It's awesome. No.

Can I just add one interesting thing?

Yeah.

Just on your thing: Alex Kerlen just left to go to Menlo Ventures, right, from there. He was on the board of Owner with me. I've known him since the very beginning—OG SaaStr, when he started in the industry.

He wrote a little presentation about how Menlo thought about this. And, not that this is so profound, but he said they're targeting $25 billion-plus tech exits. They're targeting 100 of them. That's how Menlo's modeling the world. And that's what he went to join: to find some of these 100. Not all 100—they don't have to be in all 100 of the $25 billion-plus exits, right? That's the model.

We can say, wow, Cursor was at $60 billion, but Cognition just raised at $48 billion. So if this is your world model—that there are a total of 81 $25 billion-plus tech companies now, but that's up from 23 10 years ago, and that trend is going to continue in the age of AI—you know, I wrote that the new decacorn is $25 billion. These Instinct rounds make sense. If the good exits are all north of $25 billion, then at least I can make 3x on Instinct, right?

It was just interesting to see that's their model: we're targeting $25 billion-plus exits in any investment that we do, and we see there to be another 100 of these. Now, how all that math works out with GDP and the market caps of trillion-dollar companies, I need to defer to Dr. Rory O'Driscoll next to me because I can't make it work in my head without an LLM. But I assume there was some thought behind the Menlo math here of $125 billion exits.

7. Miro Faces Valuation Collapse

We're going to go a layer up and just take a little excursion out of deliberate AI, which is Miro, one of the hailed names from 2021 that raised at $17.5 billion for its interactive whiteboards for teams, for people who don't know. It sold to Bending Spoons, the Italian juggernaut that buys everything, for $1.3 billion. As I said, it's a long way down from $17.5 billion.

We did an—well, Paul, my partner, who is very intelligent, did an analysis of it. Accel made money. Founders and employees made money. Later-stage investors, not really. How do we see this exit, guys, for a darling of the SaaS ecosystem?

Inevitable and not bad, right? Inevitable because I have a little report in our Salesforce that literally lists every unicorn, and I can do it a bunch of different ways, one of them being literally by post-money. I just eyeball down and see what's going on, and you rank them by... It jumped out at you like a sore thumb.

The last round was in '21 at $17 billion, and when you eyeball, you start to see the logos, you start to see the levels, you start to see all the Cognitions just above them and just below them. New rounds at $10 billion, $20 billion. You see we track headcount growth at the same time. Oh my God, headcount's exploding. And then you have this thing stuck at $17 billion.

It was really the largest utterly stale valuation from that period. So you look and you go, "Ah, almost inevitably you're high and dry," because you're a productivity tool in a world that just doesn't work that way. It's obviously way ahead on valuation, I mean, of where its actual market size or traction can be.

So this was just inevitable—it was inevitable that at some point it would get done. Andrew Reed from Sequoia had a really cute tweet. It's like a little graphic of—you know that picture of Death with the sickle knocking on every door? It knocks on the Evernote door, and then it knocks on the Airtable door. And in this case, it's knocking on the Miro door, and instead of a sickle, it has a Bending Spoon. Death comes for us all in SaaS land, right? And it was exactly right.

It was just an inevitable cleanup operation because it was so far wrong in terms of pricing, and it was a good outcome for everyone. And you kind of snidely said something about the late-stage guys, but I'm going to say something. The great thing about the late-stage business is this: if your losers give you 1x, then you'll die rich.

So ICONIQ, I think, had a ton of money in that deal. That's a bad deal. A bad deal is when they get a 1x. You know, if you're playing the venture game and, because you're playing late and your preference gives you 1x on everything in the worst-case outcome, by definition—just using simple math—the overall distribution is net positive. So it's a good outcome for everybody. It needed to happen. It's now part of an illiquid asset.

I think—interesting. I didn't know this until today, so I'm winging it a little. I did see that some portion of the consideration rolled. In other words, some people said, "I'll take stock in Bending Spoons for that," which is an interesting choice.

Yeah.

The cynic in me says it's interesting if you do roll.

It's basically saying, “We couldn't do what it takes to turn this company into a cash-flow-positive machine, so I'm selling it at 2.7 times to guys who are trading at 14 times,” because they are tough enough to do what it takes. Because that's really what's happening. To some extent, even though that sounds like bullshit, it's true. I've been on boards where the company just flattens out and needs to get ruthlessly efficient. Venture syndicates—it's just not our DNA. It's not how we roll.

It's a syndicate of 5 different people. Oftentimes, these assets are better owned by a single owner who says, “Look, this is the way it's going to be. This is what we're going to do.” I don't know if you saw the CEO of Bending Spoons. He made a wonderfully controversial take where he said something to the tune of, “We don't get all excited about the title ‘founder.’ We don't want to know what you did.” His basic comment is, “We don't want to know what you did 10 years ago when you founded this. We want to know what you're doing now.”

What it's basically saying is, “We can't do what it takes to fix this thing at 3x revenue, so we'll sell it to you. Take your stock at 12 times revenue, because you'll be hard-nosed enough to cut extraneous costs, raise prices, accept a fair amount of churn, and plow through.” It's interesting, but also not great. It's an interesting comment on how institutions can determine outcomes. It's not all rational economics, and it probably makes sense.

Bending Spoons will probably do a better job than a venture syndicate at making that thing cash-flow-positive. If you're a customer of any of these companies, just be ready for the 40% price increase.

The churn on their acquisitions is brutal. When you actually study the graphs on usage, the churn is absolutely brutal. They are not revitalizing these. They're increasing prices and cutting costs to the extreme.

What they're really finding out is the marginal propensity to pay versus how much the VC industry probably, almost certainly, over-invested in sales and marketing and sold people who had to be sold into the product. What they're saying is, “I don't want the customers who had to be sold into the product. I want customers who hate us so much for doubling our prices but still need this product and won't go away.”

It's a different worldview. You're going to get initial churn when they put through those price increases, but the perspective is that the people who stay really need the product. This is what happens when you have to pay full boat—

Yeah, you triple the prices and you have 30% to 40% churn, and the math's pretty straightforward, right? I'll tell you my thought on the Miro one. I almost want to move on. What's the game, musical chairs, with kids where you take out a chair each time you go around?

Yeah.

I feel like there's only 1 or 2 chairs left from the pre-AI era. Bending Spoons said in one of the interviews this week that they look at 1,000 targets seriously and do 5 to 10 a year. Even they do not have an unlimited balance sheet. Everyone—PE is sitting out. Thoma Bravo's mostly sitting out. I feel like there are 1 or 2 seats left for 1,000 unicorns.

We can talk about why ICONIQ got 1x back and whether there were a lot of options. I felt like it's different. There's 2 chairs left at the end of musical chairs. Miro grabbed 1. Just like Airtable, they only got 1 offer. When you sell for a 2-point-something-x, you know for sure there was no other offer, because anyone can pay 2.4x or 2.5x. It's not much more for Salesforce, Thoma Bravo, or Francisco Partners to outbid.

I almost want to move on because I can tell you personally, I've ended the game of musical chairs for me. I'm not running around the chairs anymore. Whatever, it's fine.

Who's next, then?

Look, of course there will be more deals, but I think we've entered the era of capitulation. If there is a seat left in musical chairs, grab it. Grab it. Otherwise, the game's just ended.

These companies are going to go into 0%-growth mode. They're going to go into 0% to 5% growth mode, and no one may buy them. If this is the best Miro and Airtable can do, what if you're not Miro? These are not bad companies. Miro's $600 million in ARR, still growing high single digits, and cash-flow-positive. It's a pretty good asset, right?

Anyone worse than Miro or Airtable is not going to get 1 of the last 1 or 2 chairs. No one wants to buy these things. I don't mean to be grouchy. I just mean I've given up. Kids, go do whatever you want. Here's the keys to the house. I've moved to another city. Have parties, crash the cars. Do whatever you want.

Boys, moving swiftly on, what would we like to do next? Jeff Dean's company hitting $50 billion after just raising at $10 billion. We've got Citrini selling his company to Dylan Patel and Sammy Analysis for $100 million. We've got OpenAI pausing Pro signups.

Is SBF getting out?

Well, he's got to accept his Midas List, doesn't he?

He might have even been on the 30 Under 30 or something when he did the Anthropic deal. I think it was 30 Under 30.

I'm going to be the voice of humanity. He is doing significant time, which really sucks, and that's a life wasted. I'm not going to pile on the guy yet, right?

I think the Supreme Court's going to take the case and overturn it narrowly. I watched his lawyer on YouTube. He was pretty good. He's done 50 Supreme Court cases. He's a badass Supreme Court lawyer.

Basically, his point was, this is an Eighth Amendment issue. You can't fine somebody $12 billion—this is unconstitutional—who can't repay it, when, at least according to the terms of the bankruptcy court, everyone was repaid in full with interest. We could argue whether they would have made much more money, right? But there are some constitutionality issues.

I think he's going to get his day. I think the Supreme Court's going to take it. They don't have to take any case, and I think he's going to get it. Whatever he's in jail for—30 years—and $11 billion, I think Sam may be freed eventually. Again, the Supreme Court hearing the case is a far cry from being freed, but I think he may have a day in court.

Jason, do you think he should be freed?

He seemed like the biggest scammer of all mankind when the—what was this? All the way back in 2023? It is weird he didn't materially enrich himself, as far as we know, right? It is weird that the argument that it was allowed by the terms of use is an interesting argument. It is an interesting argument. He did not self-enrich.

The reality is, if you use a generous version of “made whole,” folks were made whole. Should he go to jail for most of the rest of his life and have to pay $11 billion when he gets out? It seems like a lot in the era of sentient AI that could kill us all. It does seem like a lot today.

But at the time, Silicon Valley Bank failed. I lost $10 million over the weekend. I don't know about you guys. It seemed like just desserts at the time, right?

For what it's worth, I hadn't prepped on this because it wasn't on the list, and I don't like to practice law without a license. My wife, who is a lawyer, gets mad when I do that. But it was so low down, Harry, I didn't think we'd get to it. You never do, right?

Oh, there we go.

Right. Well, hurry up.

It wasn't on the list. Hurry up.

Genuine comment here. Separating the fine, which, no matter the issue, doesn't matter, there was a misallocation of funds. It was white-collar crime, and it should be punished.

I actually think 30 years was probably disproportionate. I'm trying to remember the ex-Goldman Sachs guy who misallocated brokerage funds in MF Global about 10 or 15 years ago. I think he walked entirely free. I can't remember everything.

I think white-collar crime should be punished, and it's a shitty world where someone steals $20 and they go to prison, and someone else steals $10 billion and they don't because they're white and middle class. But I'm also not sure if 30 years is the commensurate thing.

The interesting thing is that that's not the issue at hand in the Supreme Court. I don't think they're appealing the sentencing, or the sentencing-guidelines issue. They're appealing the facts and circumstances of the case. So we'll see.

I don't think it matters that he took the money and was a brilliant investor. That doesn't excuse him for taking the money. By definition, if it worked, then anyone could take money, provided it worked. “Hey, I stole your money, Harry, but I bought put-call options and the stock went up, so we made money. Here's your money back. You shouldn't mind.”

Well, of course you're going to mind, because when he took the money from you, you didn't know what the outcome would be, and there was a 50% chance it would go down. So you're going to want that guy punished, because you're going to want that behavior stopped. You're not going to want the next guy to think they can do it.

It doesn't matter that he was the most brilliant—genuinely, the most brilliant—equity investor of our generation between Anthropic and Cursor.

It just matters whether he took money and whether that was against the rules. I haven't heard the terms-of-service argument, Jason; that will be interesting. If he took money and it was against the rules, then he should be punished. If he took money and it wasn't against the rules, then he probably should walk, and the process will take place. It's not my problem.

8. Mullenweg Retakes Automattic Control

What happened to Matt Mullenweg? He was out for a day and back. What happened to poor Matt? Did someone not read the bylaws, Rory? What happened? Did someone forget to pull the certificate of incorporation from Delaware?

So, Matt Mullenweg is the founder and CEO of WordPress, and he was ousted by the board and then came back and overrode them, it would seem. Now he is back as CEO, and the founder's power has prevailed here.

Correct. Automattic is a company that has been the steward of the open-source project WordPress, which is one of the most commonly used blogging and website platforms out there. It's a very successful product. Matt is the CEO of the company that manages the product.

I think it's fair to say that the stewardship in the last few years has been troubled. He's been in a big argument with WP Engine. The argument is that WP Engine is a hosting company that hosts WordPress sites, and I think Automattic wants some of that revenue, so they've been pushing WP Engine.

But the way in which he's behaved has been unhelpful to the open-source project, because it's kind of like, "I will use the leverage of my company to try and, frankly, prevent other people from benefiting from the open-source ecosystem," which seems to be antithetical to the idea of it. So I think it's been a troubled situation for a while.

I think the truth is, the real issue is that the world is passing that product by. It's kind of a sad little thing, because the world is passing that product by, and Jason will be able to tell you that you can build most of what you have in WordPress with Lovable or Replit or any one of 10 things. And increasingly, they are.

As I've quoted before, the Henry Kissinger thing about academic politics: the fights are so vicious because the stakes are so small. The truth is, Automattic doesn't matter a damn anymore. It should try and build something new, but it's kind of on the tail end of tech trends.

To Jason's point about Matt, it's on the tail end of tech trends. They should be doing things totally differently to try and survive in the brave new world. Instead, they're arguing internally. So that's kind of the zoom-out comment. Within that context, that's kind of the big picture.

The funny thing is, like all litigation, when you get caught up in the detail of the day-to-day, you forget the big picture. The big picture is that this company needs to point WordPress in a forward direction and think about how to take advantage of what's going on in AI and become a relevant player in the next 5 years. Otherwise, it won't be. That's the corporate imperative.

Instead, the corporate imperative has turned into a pissing match between the board and the CEO, where now it looks like the CEO has won. Congratulations, you've won the poisoned chalice. You get to keep your diminishing empire.

It's worth pointing out that this is not a board full of evil VCs. I've been on boards as an evil VC where you've had to replace a founder. It sucks to no end. This is actually a board of—I think Salesforce is an investor—and they have some really good independent board members. They don't need this grief.

I'm willing to bet what happened—now, down in the tactical weeds—I think Jason probably nailed it. You probably have a board, and they probably have a majority. They probably said to the CEO, "We're a majority of the board. We're independent directors, and we think we should replace you."

My guess is the founder-CEO went deep into the bylaws and said, "You are the board, and you can vote to replace me, but I can actually also vote to replace the board. I'm hereby voting to replace the board. You're all off the board. Oh, look, the new board is me, my pet dog, and my ventriloquist dummy, and, after due consideration, we've decided I would be a great CEO."

There you go. All the independent board members at that point promptly resign, because there's no point in wasting your life and getting into a whole bunch of litigation about every dime. Remember, Jason, your island—was it the Fortnite island that's continually getting smaller?

Yeah.

You're fighting to maintain control of an island that's getting smaller and smaller. So congratulations, Matt, you're in charge. I feel like going, "If you are in charge, well, how about you turn this thing around?"

That's my takeaway of this. It's kind of sad in a way. These things happen. People behave how they're going to behave. I think a lot of other people were involved in that company and put a lot of effort in. One of the guys, Tony, who was at True Ventures, did a lot in it early on.

It's kind of a blah situation. It's been pending for a long time, and you kind of just go, "Gosh, I wish there surely should be more of a win-win here."

Well, Automattic would have been a great company if it hadn't raised venture capital, because Matt could do what he cared about, which was having a commercial arm of an open-source product he created when he was very young. Imagine it's doing $500 million a year, spinning off $200 million. It's like a bigger Basecamp.

Those guys aren't venture-backable, and they don't give a rat's ass about 22 Lamborghinis and Pagani's and villas in Italy. They're fine growing 30% at $60 million, spinning off $30 million or $40 million in cash. It doesn't bother them at all, does it?

I don't know the full funding history, but one could imagine it didn't really need to raise all this money. A version of it might have done just fine. It's easy to say, but that might be the one. That might be why Matt's frustrated. It would haunt me as well.

Companies do get overfunded, boys. I might be sitting here like, "This could've been 37signals. I could be running this. I didn't need this $800 million and a bunch of people running around doing nothing all day long."

I could have run WordPress and Automattic both side by side with 80 people, like DHH, and be making $100 million a year. Who the fuck cares if I'm growing if this is my mission in life? If you're making 9 figures a year out of your company and your growth is 5%, and you're happy and you're doing a good deal for the world, screw you, VCs.

Yeah. First of all, I do agree that you shouldn't take VC if you're not signed up for that program. I'm also not sure that that's one dimension: VC versus a lifestyle business.

There is another dimension here, which is the open-source business. On the first dimension—VC versus a lifestyle business—it's pretty clear you should do the lifestyle if you want the lifestyle. There is another dimension, which is an open-source community project versus managing just the, quote-unquote, "company that controls the open-source product." I'm not sure that on that dimension the company has been an amazing steward of the project, right?

Mm-hmm.

So I hear you. I think that's just an initial dynamic. But you could also say, to your point, Jason, that if you didn't take any venture money, if you initiated the project, if you have whatever open-source rights you have under the copyright, whatever leverage you have in terms of the licensing and the copyright, then it's your company. Do what you like.

Actually, I do agree with you: this is America. If you built the thing and you want to mismanage it, mismanage it, and you own 100% of it, you're allowed to mismanage it. Go team. That's what ownership means.

You know what's tough, too? I know you want to wrap it up. I'm just learning. I'm learning later in my investing career: you have to be effing ruthless to do a venture-backed open-source company. Ruthless. You really do.

Because look what Matt did. Matt said, "Listen, I'm focused on the platform. This is what I'm passionate about. I'm not so into hosting, okay? That's a commodity business. I'm going to let WP Engine do $500 million—you know, $100 million, $80 million, $100 million, $300 million—and let these other folks do it."

And he's like, "I'm kind of into this e-commerce thing, like WooCommerce, but I don't want to do what Shopify and Tobi did. That's too extreme." So you end up not owning that much of the revenue in your ecosystem.

I don't think he's mad about the WooCommerce thing not being huge from a revenue perspective, because WooCommerce actually used to be massive in terms of scale versus Shopify. But I think he's looking at WP Engine and saying, "Not only did these guys treat my community worse after the PE buyout"—which I do think is objectively true—"but I kind of want that $500 million now, guys. I wouldn't have minded having that extra $500 million for this crappy commodity hosting that I could do a better job at."

In fact, we're on their own product. It's a great product. But my guess is it's doing a fraction of what it has to be, mathematically, right? It's like, "I want that."

So if you're too kind in open source, I think you lose. If you're too kind, you lose. You've got to be ruthless.

If you think about a couple of different agenda items that we've talked about here, on the one hand, the trailing edge of tech trends: you have Miro getting bought for 2.7 times, and you have a nasty little spat over a flat-to-declining asset in open-source land at Automattic.

And then on the front edge of the thing, you have companies like Instinct, raising literally 4 weeks ago at $2.5 billion, now raising at $10 billion. You mentioned the Jeff Dean–kind of spinout, raising at $10 billion, now raising at $50 billion. The big-picture point—it’s really Captain Obvious here—but it just shows venture, unlike PE, is not about valuation, and there aren’t any safe assets. You’re either in the head of the train, in the new new thing, where everything is possible, or you’re in the tail end of the train and life is shitty.

I was with one of the biggest CIOs the other day, and I said, “You’ve been doing this for 30 years. Have you ever seen a time like this? I’m cognizant that I’m on the younger end of the spectrum. I’m not that young anymore, but I haven’t seen all the cycles.” And he said, “I’ve never seen a time like this. This is more nuts than it’s ever been.”

This is unprecedented.

I think that’s true. I was investing only since 1993 or 1994, so I did live through the dot-com investing thing, and it was pretty crazy, because you also had the whole millennium thing about Y2K and the world was going to end. Just like now, we always have to have a world-ending thing.

Let’s be honest: New Year’s Eve, December 1999, was a pretty wild freaking party in San Francisco, right? Half the people were getting toasted drunk because they were rich, and the other half of the people were staying up, making sure that the Y2K thing didn’t bring down the world, which was the actual worry at the time, bizarrely enough.

But it’s nothing compared to this. I think the ability of AI to just excite the imagination is much higher. The internet was awesome, but with AI, you can start talking about AI—the software is human. You can get carried away.

And then the second thing is, instead of existentially worrying that the computers won’t work because of Y2K, we now get to worry about the whole world ending. It’s the same thing on a way more magnified scale, and the money is 10× bigger.

Can I ask, with the money being 10× bigger, I sit in Europe, as you know, Rory, and you remind me frequently—

Well, actually, Harry, you don’t. You actually sit in England, which deliberately chose not to be part of Europe. But I know what you mean. It’s okay.

Sorry, that was mean. Listen, I wasn’t pro-Brexit, but you’re technically right. I agree.

Yeah.

Very sad. Mistral raises €3 billion. It’s Europe’s largest-ever tech round. They’re going to hit $1 billion in revenue by the end of the year. For a company that’s had, I think, a lot of criticism, definitely in Europe, this was a very meaningful sign of progression and hope that we actually do have a horse still, so to speak, in the race. What should we take from this round?

9. Europe Bets On AI Sovereignty

I think it’s less about being a competitive frontier lab and more about AI sovereignty. I don’t think you should take the fact that you have a horse into it. If you’re saying Mistral is now competing with OpenAI and Anthropic in the frontier-model race, you and I know that would be bullshit.

I think what you’re really saying is that Europe has decided, based on the antics they see from the companies at the frontier in the U.S., coupled with the dynamics of the political interaction between the United States of America and Europe, that a technology as important as this has to have a sovereign European component.

Even though it’s obviously ludicrously inefficient from any kind of rational perspective, they’ve decided, “We just have to make this thing happen,” and give Mistral enough business to make sure that it’s a viable European AI competitor.

And it’s clearly Mistral. There’s a long tradition of Europe doing this. Airbus is an example of that. They said, “We can’t just be relying on the Americans to make planes, because otherwise we’re a vassal state.” And they said, “We’re going to make it happen. France and Germany are going to make planes, and we’re just going to do it.”

It took 10 or 15 years, but they built a viable competitor. It’s probably the same thing here. They’re simply saying, “We can’t afford to have…” We’ve seen instances where—I’m doing this from memory—I think the U.S. government said to Anthropic, “Thou shalt cut off all other countries from Fable.” I think it was Claude, or one of the most recent models. “Cut them off because it’s a security risk.”

We didn’t say, “Cut off Russia, but leave the United Kingdom, France, and Germany.” We said, “Cut off everybody.” If you’re in Europe, the day that happened, you said to yourself, “We can’t rely on these guys anymore.” And the day that happened, you made Mistral a viable European competitor.

Is it going to be as big as the U.S. companies? No. I don’t think it’s going to be anywhere near as big a market cap as OpenAI or Anthropic, but it’s going to be a winner, and it’s going to be the European winner.

So thank you very much, politics. If you’re a Mistral shareholder, you should be very grateful for the current political tactics. It’s just made you a couple of billion bucks.

Also, just to be technical, the round’s led by Samsung, right? They’re generating, I think, $200 billion of free cash flow a year right now. So $3 billion—I don’t know if I’d take the markup. I guess I would take the markup, but you could argue it should have an asterisk and a dagger next to it, because is it really real if it’s led by Samsung?

Well, the last round was led by ASML.

Yeah, I don’t know that those count.

But the money counts. I agree. Look, the valuation—

Yeah, the money counts, but I don’t know if the valuation is connected to anything real. We shouldn’t act as if it’s the same as an objective valuation done by financial parties. It’s just not right.

Agreed, but there is some intrinsic value to building a European competitor. Let’s ask the brutal question: if the American winners are worth $1 trillion and European GDP is roughly 70% or 80% of American GDP, I don’t think that implies, to be clear, $800 billion, but you might get to $30 billion, $40 billion, $50 billion. It’s not a crazy end state.

But I agree, Jason. It feels like you’re probably right. I don’t think they did it on the basis of market comps and comparables. I think they did it on the basis of state strategy.

Boys, is there anything else I’ve missed that you think we should cover?

I’m just trying to find interesting private stuff. Jason, any comment on Adobe’s new CEO taking over? I saw the results. Any thoughts, because you’re the Adobe expert here?

10. Adobe Capitulates On Growth

I think it’s great. You take 2 non-founder folks slowly leading different business units into no growth, and you pick from the lesser of 2 okay leaders.

I think it’s just Miro at scale, but with so much scale that it survives, right? You know what it is? It’s a reminder that Miro didn’t have scale. When the 3 of us met, if I told you about this $600 million company in collaboration, we would have thought that was scale in the old days, right? It’s not scale. Adobe has scale.

Even if Adobe isn’t really making the right moves in AI, can’t really afford it, and is just moving the deck chairs around for its C-suite team, Shantanu Narayen said about a year ago that he was going to retire, and it took him a year to decide which of the 2 internal candidates to promote. I mean, it’s pretty bad.

I think it’s a nothingburger. You know what it is, Rory? It’s a sign that nothing’s going to change. It’s a sign of capitulation. It’s a sign that we’re going to keep bulldozing our way through the world the same way instead of really changing: super-high margin, bleed our core products, add some image-generation AI, and call it a day, right?

I broadly agree with you, and it’s interesting. They did the, “We’re not leaning into ARR growth; we’re leaning into getting free usage,” and that was the play they did in the ’90s.

That’s their fake AI metric of the week.

And that’s where I was going. The reason I asked you is that this is a playbook that might have worked 2 or 3 years ago. But to your point, what you’re basically saying is that their net-new ARR went down significantly. In other words, the growth of new ARR didn’t happen.

So their AI ARR metric went up, but overall ARR went down. And you know what that means: you’re just channeling some of it into the good thing. But yes, you at least have the program that you need to win, which is a start. You’re not winning, because of the Jason rule, which I go back to: if it ain’t growing, you ain’t winning. If people ain’t paying for it, you ain’t winning.

They have scale. They’re not going away. $25 billion in revenue, whatever. There was no news here.

What’s interesting is that, overall, it’s been a pretty good few weeks for software in general. You’ve seen big jumps. I’d say huge jumps in the cyber stocks, big jumps overall in software, and the entire SaaS apocalypse has been unwound.

But I think, Jason, to your point, what’s been happening is there’s been a real distillation of, “These kinds of businesses aren’t going to be winners. They should stay low.” And then these kinds of businesses are doing super well. Adobe is more in the first category than the second.

So, boys, their market cap today is $105 billion.

Yeah, about $100 billion.

$25 billion, yeah, exactly.

In 3 years' time, what is their market cap?

$120 billion, $130 billion.

Same as today. It'll be the same as today.

Yeah, plus or minus. It's not going away. Basically, again, from memory, it's a sub-10-times cash flow multiple, Harry. So unless the ARR evaporates—not just doesn't grow, but evaporates—you can get down on the cash flow multiple.

If Jason's view of the world is correct—if there are 10 or $125 billion-plus outcomes, which, by the way, I don't think there will be—the point is, your relative significance will go down. I mean, I couldn't tell you HP's market cap now to save my life because it doesn't matter, and that's the same thing. You'll do fine. You'll be cash-flow positive. Maybe if you find the right leader and the right product, you'll reignite growth. Your trajectory is not to blow up, but your trajectory is to trade at an 8- or 9-times cash flow multiple.

Well, that's the way it was for a decade. Adobe stock didn't move for a decade. It traded on cash flow. Shantanu was amazing at that, then cloud worked better than anyone expected. Either they're going to run that playbook for another decade, or they'll have some magical AI thing that even Canva hasn't figured out.

I'm not betting on it today, but I was there when cloud happened. They didn't expect it. They could have a killer AI app. We're early in the AI thing, so I don't know, man.

Would you rather invest in Canva or Adobe today?

I'll tell you why you just can't invest in Canva. I want Canva to win very badly, right? I love them. You just can't be slow-growing. ServiceTitan got destroyed for lowering guidance. It fell 30% last week, okay? I think Canva fell 30% or 40% this year when they had to lower guidance, right? Maybe more, right? They're just not public.

But you've got to grow. At least Adobe is going to get to this stable plane, to Rory's prior point. It's going to get to a stable plane. It's not impressive, but it's going to grow 6%, 8%, or 9% a year with lots of free cash flow, right?

I'm going to give you the numbers. Canva's numbers, right? They are growing. They're growing 20%. They were growing 30%. They're growing at 20% now, and I know that because in Australia you have to file your revenue with the whatchamacallems.

But that's a big deceleration.

Agreed. So it's not that they're not growing; they're decelerating.

So are they entering DCF world, or are they still a growth stock? Are they growth or value?

Okay, I'm going to agree with you now, and the only reason I interrupted you, Jason, is that I didn't agree when you said they're not growing. I want to be precise. What you said now is really the insightful point. Seriously, I want to pause on this because it actually comes to a lot of different things.

Going from a growth story to a value story is really tough, right? You go from a revenue multiple to an EBITDA multiple. I think it was Gokul Hariharan. I saw a really good post on Twitter recently where it basically said that above 30% growth, you can use a revenue multiple; below 30% growth, you have to use an EBITDA multiple. It was a very insightful comment.

It basically said if you're growing fast, everything is forgiven, and you'll be valued on a revenue multiple. If you're growing slowly, nothing is forgiven, and you'll be valued on an EBITDA multiple, a low revenue multiple, right? It's just a different world.

I remember I did a post way back that Box, as a public company, went through that transition and came out the other side, but it takes 3 or 4 years. Because when you go from 6 or 7 times revenues to 20 times cash flow, you've got to get that cash flow to 30% just to hold the stock flat. It's brutal. It takes forever.

This goes back to the point: I think Canva is still growing very nicely, but it is decelerating. Even though I like them and I want them to win, they're founder-led, whereas Adobe, as you say, is at this point exploiting me every time I buy their product. I loathe their licensing system so much.

The hard truth is this: when you're going through that transition, it's hard to get to a public offering. So when you ask about the 2 stocks, Harry, what you're basically saying is, would you like to own something at 8 times cash flow with 13% growth where at least you're liquid, or would you prefer to own something at 20% growth, admittedly decelerating?

The question is, can they get through the liquidity window, right? Price clears all markets. I think the interesting question is how you think about relative valuation. I just think having to go from the growth-valuation world to the value-based world sucks, and doing it while private is hard because it takes a long time to get through that nut.

During that time, you're standing still, right? I mean, it sucks doing it as a public company, too. It just sucks doing it in general because you get a different investor base and you have different dynamics, but it's just a hard row to hoe. It's the risk of being private for a long time.

Stripe avoided that risk because they reaccelerated, and I really hope Canva can find a way to reaccelerate, too, because I want the founders to win. I want good guys to win. But if, instead of Stripe reaccelerating to 40%, they'd decelerated down to 20% and 15%, then they'd—you know, I mean, they still have huge cash flow, so no matter what, they're fine.

But it's just interesting. Slowing growth and being valued on a cash flow basis is a profound adjustment in the valuation metric. Oddly enough, as I think about it, Stripe is the only guy who could weather that storm because apparently it kicks off so much cash.

Rory, we've got to go let Jason be a superstar at Dreamforce.

Oh, you are? I'm sorry, Jason. You're off to be a superstar.

Yeah. He's off to be a superstar with Mr. Marc Benioff. Yeah, he's the star attraction.