Victor Riparbelli, CEO @Synthesia: OpenAI vs Anthropic vs X.ai - Who Wins and Why | E1246

I think we're definitely in a bubble. There’s a lot of money that’s going to go up in flames in AI products. What I see a lot in the enterprise is that buyers don’t really know what they want. The real signal is not that you sign a contract; the real signal is renewal.

Victor, dude, we met 6 or 7 years ago, when I was very, very young and very naive. I can’t believe it’s been that long, but thank you so much for joining me.

I’m glad to be here, man. I’m still—like we said before—I’m sorry for you that we’re meeting under these circumstances.

So am I. I said to you before, I would have done better if I had invested in every single company I’d ever met, because it would have been you and ElevenLabs. But I want to start today with some really exciting news that you have. What is the exciting news you have for us today?

We just raised a Series D, which is super exciting—a big milestone for us. We raised $100 million, led by NEA, with participation from all of our existing investors. We’re very excited to get into ’25 with a big war chest and, from our perspective, to reach escape velocity, shut down the category that we’re in, and win.

Dude, first, amazing news. Congratulations.

Thank you.

And thank you again for rubbing it in my face. My question to you is about the funding story of the business. Can you take me to the seed round? People didn’t get it. What happened, and how was that?

This was back in 2017, so a long time ago now. The very short story is that we were a bunch of people—myself, my co-founder Steffen, and my co-founders Professor Matthias Niessner and Lourdes Agapito—and we had this idea that generative AI, which back then wasn’t really a term most people thought about, would change how we create content.

The big shift was that, back in 2017, when most people thought about AI, it was about analyzing data and making decisions. That was the era of AI. But there were these early GANs, which were basically a new network that could produce new data instead of just analyzing existing data.

We thought this was going to be a world-changing technology. We thought it was going to change everything we knew about how we create content—from video and speech to audio and music. We were focusing on video, though. I went out to the world with a great PowerPoint deck and what we thought was a great vision, but understandably, most people thought we were pretty crazy.

We basically went out and said, “Look, in 10 years you’re going to be able to make a Hollywood film from your laptop, needing nothing else than your imagination.” That wasn’t a pitch that landed particularly well, especially not in Europe. We were based in London at the time.

Why do you think that was?

I think in Europe—it’s gotten better, but the VC industry is still dominated by people who used to work in private equity. That’s a very different mindset from that of a technologist. When you used to work in private equity, you’re more of a financier than a technologist, and this wasn’t a pitch you could understand in an Excel sheet.

We got turned down by basically everyone—I think 80 or 90 investors, something like that.

Including podcasters?

Including podcasters who weren’t in private equity and don’t have that as an excuse.

How much were you raising then?

The first round was $1 million.

And you ended up doing it at a $5 million post-money valuation with Mark Cuban?

Yes. Mark did 1 on 5—1 on 5.

With today’s valuation of $2.1 billion, that’s pretty good. He has something like 15% today?

Something in that range.

Then you went out and wanted to raise, like, $7 million?

No, that was the later round. Back in 2017, we did $1 million at $5 million, and that got us started. We went 12 to 18 months, as you do, and went back out to the market. We felt like we now had working technology. We were focused on AI dubbing at the time, not the avatar technology that we mostly have today.

Again, we learned a hard lesson. We went out and said, “We built this great technology, we think we’re a great team—let’s raise $8 million.” That completely failed. For 9 months, we were just dragging our feet. I made all the mistakes we could make as founders. I dragged out the funding process over 9 months, with different data points and different investors. It was a big shit show.

Eventually, we had to rewind because we were running out of money, and we ended up raising $3.1 million. That took us through to the Series A, which was when we had actually found product-market fit and had a sustainable business.

Those first 2 rounds were very much rounds we raised based on a story. It was a story that, back then, didn’t resonate that much because it was very hard for people to see what we had and how that could extrapolate into everything that became today.

Today, that would have been a $5 million to $10 million round at a $40 million to $50 million valuation, given the team and the vision. Would you have been as successful as you have been if you had raised that round instead of the smaller, leaner rounds that you raised?

We talk about this a lot, and I don’t think so. If we had raised $8 million, we could have done a lot more things, and we would have done more things. We would have built deepfake detection, which everybody wanted us to build.

Remember, this was in 2018. At that point, AI video, to everyone outside of academia, was just deepfakes. People thought that was going to be the big thing. Of course, it’s a problem and a real thing, but we always thought it was a subset of AI video. Most people were going to use these technologies to create awesome creative content.

If we had got that money, I think we would have built a team to do that, and that would have meant we lost our focus. I actually think operating under the constraints we had at the time focused us on our customers and selling the product. We were ferocious about charging people from day 1, even if it was £500.

I think that focus, and working under those constraints, definitely helped us get to the point where we found product-market fit.

I’m a big fan of working under constraints. I really do think it forces a lot of discipline that can easily get lost the more money you have, because you have more options.

You’re also an angel investor now. How do you feel about the $5 million on $25 million and $10 million on $50 million rounds with pedigreed founders coming from massive-name companies? How do you feel about them, and have you seen a trend in terms of how they perform?

Too much money too early is not healthy. Maybe some people are good at having that discipline—perhaps second-time founders—but it’s very tempting to spend on things you shouldn’t be spending money on, especially when things aren’t working.

Once you clearly have product-market fit, a lot of decisions become easier because they’re more obvious. But before product-market fit, it’s dangerous to develop 2, 3, or 4 things at a time, or to have 15 people working on your team when you’re still at the idea stage. A lot of people make the mistake of raising the money and then using it too quickly.

One thing you cannot use money to buy your way out of is product-market fit. Learning about your market and your customers just takes time. Having a team of 20 people instead of 2 people trying to learn that can actually slow you down.

You have to own that as a founder. I don’t think we could have learned the same amount about our customers faster if we had more money. It really took us 2 years to get into the minds of customers and understand video from first principles.

People try to hire a product manager to help fix the product-market-fit problem because they’re a product person. Or they hire a salesperson because they think they have product-market fit and just need someone who’s better at selling it.

Unless you have product-market fit, you shouldn’t have product managers and salespeople. That is your job as a founder.

Do you think that journey is ever done? I don’t believe in product-market fit being this thing where you say, “We have product-market fit.” You have product-market fit with creators, say, and then suddenly you need to move into SMBs. You don’t have product-market fit with them; you have it with independent creators, and then you need it with enterprise.

I see product-market fit as this ever-moving chapter book. Can you ever afford to move away from customers, and doesn’t it become more impossible with bigger teams?

I totally agree with you. A successful company is a long series of product-market fits, but you need that initial spark, because once you have it, you have something to build on.

The bigger you get and the bigger company you’re building, you need new products and new product-market fits, essentially. I do think that’s one of the things you should always be focusing on as a founder: What’s the next market? What’s the next product you’re targeting?

In our case, I’d say we already have a bunch of product-market fits. You can hire super-smart people who can run with those, make them great lines of business, and make great products. But pushing toward those next product-market fits, and figuring out where the company needs to be in 2 or 3 years, remains the founders’ job.

You talked about the importance of capital constraints and cockroach mode. You hadn’t touched your Series A when you raised your B, you hadn’t touched your B when you raised your C, and now you haven’t touched your C when you’ve raised your D. Why raise it?

When you have the money, you do spend it. For us, it’s been more a matter of always being obsessed with actually building a business. That sounds a bit wacky, but it means having great unit economics, making sure we’re a business that generates money and revenue, and that we’re always in control of our own destiny rather than tied to a VC parachute.

We’ve always spent conservatively, but it’s also very clear that if you know where to spend the capital, it’s an amazing asset. Building great go-to-market, for example, costs money. There’s a cash-flow element: You have to hire a bunch of really great people, train them, and wait 9 months before they ramp up. Then the investment is worth it.

You want to have capital. You want to have a really healthy balance sheet so that you can chase any opportunity that comes ahead of you. We want to build a really, really big company. I think there’s easily a $50 billion to $100 billion company to be built in this space, and we want to win that. You’re not going to win that by bootstrapping all the way. I think that’s a myth.

How does Synthesia become a $50 billion to $100 billion company? Can you paint that picture for me?

I think we’re in the early stages of a shift in how we communicate. If you think of most communication today, it’s text-based—emails and texts. We read things, and text is a great technology. It’s an amazing technology that has built the world up to where it is today.

But it’s actually a pretty bad way of compressing information. You lose a lot of context when you transform your thoughts into something written down in a document. As humans, we’re much better at consuming visual content. We like to hear things, see things, and feel things in the physical world. We can’t do that yet, but it’s very clear that higher-fidelity content, like video and audio, is a better way of training, informing, and entertaining people.

The reason we use so much text today is that text is the only scalable way we have of storing and sharing information. But that’s changing now. We don’t need cameras and microphones to capture things in the physical world around us, and we can get video and audio creation to be as scalable as text.

Once that happens, there isn’t really any reason for us to use text anymore. This sounds a bit crazy, but I actually think that maybe our kids’ kids will be one of the last generations to read and write as the default way of communication. I think we’ll increasingly consume everything through video and audio.

Look at all the trends. You look at TikTok—though I don’t know if you’re on TikTok.

Every night, from 12:00 to 12:30, I watch it, and then I send my team notes on what I think works and doesn’t work in transitions, fades, and music.

Yes, I do exactly the same thing. TikTok is a great example of how video truly can be the default way we consume information. There’s almost no text left in the interface. Even comments—people respond to comments with videos—so they’ve started to build out this graph of how we communicate with video as the default.

There’s still a long way to go, but if you believe that’s true, then it’s not just about scrolling TikTok at night. If you’re trying to buy a software product, you don’t want to read a whole bunch of material and jump on calls with people. You want to watch a video, and at some point you probably want to watch an interactive video where you can say, “Can you show me how this functionality works?” and the video switches over to that.

When you need customer support, you don’t want to be on the phone or read long articles. You watch videos, and they’ll probably be interactive. Then all the world’s communication is the market.

When we started the company, we always talked about—and it’s still a very important part of our thesis—that our market is not video production. All video production today is an interesting market, but our market is text. Text and slides are the market we’re targeting, and that market is infinitely big.

If you manage to capture just 5% of the world’s text communications and turn that into video, you probably have a company worth more than $100 billion.

Before we go into the future and how we see it, I do want to remain on the funding. Is it bad to raise money to dissuade competitors, win a market, and use capital as a weapon? Is this round a signal to other people saying, “Don’t come here”?

Sure, it’s part of it, but more than anything, this round is about raising the capital to build the best product in the category. Every time you raise money, you send a signal to competitors and VCs, but I have never made a decision at Synthesia based on what our competitors do or don’t do. I think that’s a bad way to run a company.

Do you not pay attention to competitors?

Of course you pay attention to competitors. We have lots of them now, and I think it’s great to have someone to play ball against to some extent.

One of the interesting things about Synthesia is that when we raised our Series A to B to C round, it was very much a secret how fast we were growing and how much people loved the product. From the outside, a lot of people looked at it and thought, “There are a few people in learning and development and training who think this is a cool thing to spend time on, making these AI videos.”

What we saw from the inside was, “Holy shit, this is huge.” Training and learning was the first market we targeted, but this was clearly just the beginning. From the outside, it looked like a cute UK company making avatars. It was fun, but most people probably thought customers were using it to make a funny video for their mom.

For many years, we had to market the category ourselves. Slowly, people started to realize that this maybe wasn’t just a flash-in-the-pan demo, and that there was something real under the hood. Then we started getting a lot of competitors.

We have one particular competitor that literally copied our mission statement verbatim. Everything I say, they copy. The way we talk about the product, the way we talk about the market—everything.

Who is that?

They know who they are. It’s annoying, of course, but that’s capitalism. That’s good. Eventually, you start to learn from your competitors. They do something right, they do something wrong, and it’s actually very helpful.

It’s really difficult to pioneer a space because you have no feedback loop except for what you do and how the market reacts to it. Seeing what competitors are doing, and seeing what works and what doesn’t, is extremely powerful. For us, that’s been a very powerful thing over the last 12 months.

The reason we had a massive reacceleration in the second half of the year is, in large part, because our competitors did a whole bunch of marketing for the category. We clearly had the superior product, so that helped us quite significantly.

You don’t always have to be the first to do something. If you’re good at being a fast follower, that’s also very powerful. It’s a different way of running the company when you have a lot of feedback signals in the market around you, but I think it’s more fun, and it helps you get to the right answer for your customers much faster. Ultimately, that’s the most important thing.

Before we dive into the journey and the elements of it, I do want to discuss the AI landscape today. I speak to so many CEOs, and they say, “Harry, you and your tech bros sell me on the ROI in the enterprise. We’re still waiting.” Where are we in the AI hype cycle today?

I think that’s pretty spot-on. What I see a lot in the enterprise is that buyers don’t really know what they want. A lot of people have been told that they need to have an AI strategy and execute an AI strategy. That means they’re very willing to have conversations and spend their innovation budgets on doing things, but they don’t really know what they need or want.

They don’t understand the technologies well enough to figure out for themselves what they need for their business. That’s both an opportunity and a problem for a lot of AI startups that don’t have real customer obsession.

It’s great because there’s a lot of budget available, and people are extremely willing to do things and sign up for pilots and proofs of concept because they want to deliver an AI strategy to their boss. But when they don’t know what they actually want, it’s very difficult to prove ROI to them.

What’s the problem? Is it implementation?

I think that to be successful today as an AI company, you need to be extremely customer-centric. You need to deeply understand your customers: What are their problems, and how can your product help solve a specific problem?

That sounds obvious. It’s not AI-specific. But today there are a lot of companies doing cool technology, going out, and convincing customers that whatever they’re building is the right thing and will help solve some big problem. Maybe it doesn’t really work, or maybe they haven’t understood the customer problem deeply enough.

Now you have a situation where the AI startup thinks it’s delivering real value, but it’s not a good signal. On the other side, you have a buyer who has just spent a lot of money doing something and says, “It works really well,” because they just spent $100,000 on it.

Eventually, they’ll churn because it doesn’t actually work or doesn’t do the thing they wanted it to do. That’s a problem because the AI startup thinks it’s doing the right thing and delivering value, but at some point it will hit a wall of churn, when all those 12-month contracts begin to phase out because the company isn’t actually delivering value.

Are we seeing that wall-of-churn moment now?

A lot of companies are seeing that, big time. What’s unique about AI over the last couple of years is that people have been extremely willing to part with their money. Consumers don’t mind paying $30 a month to try something that looks cool. At the enterprise level, they’ll sign up for a $50,000 pilot to do something.

But the real signal is not that you sign a contract. The real signal is renewal. There are too many AI startups that optimized too much for closing new contracts, rather than for renewal. If you optimize for new contracts and not renewals, unless you’ve hit the right thing—which some people do—you’re in for a whole lot of trouble.

What do you think is the biggest misconception people have about where we are today versus where we really are?

I think we’re definitely in a bubble, but I don’t think that’s necessarily a bad thing. That’s how capitalism works. You throw a lot of money and a lot of products at the world, try a million different things at once, and see what sticks. That’s the right way to innovate. That’s Darwinian by nature.

But there’s a lot of money that’s going to go up in flames in AI products that either aren’t that valuable or eventually become features in some of the big cloud providers.

What is getting money today that people think will be very valuable, but you don’t think will be?

Buzzwords always take me out when people say they’re building AI agents to do all sorts of different things. That always lights up my bullshit detector a little bit. When you’re overly obsessed with the technology and the latest buzzword, that’s usually a yellow flag for me.

I’ll give you one concrete example: I really hate it when people say “AI employees.” I think it’s so dumb. It’s not helpful for building useful technologies that people want to adopt, and it’s the wrong way of thinking about it.

You’re going to have AI employees doing all sorts of different things for you? These are algorithms. They’re pieces of software. You wouldn’t say that Miro or Figma is an AI employee that sits and takes people’s designs and puts them into something.

I understand that people think these things will make decisions autonomously, but I don’t think it’s that different from software that we already know.

Did you agree with Mark Zuckerberg on his Joe Rogan episode when he said that by the middle of this year we would have AIs as capable as mid-level software engineers?

In general, the tech industry is very good at setting really high expectations and then not always meeting them. With these hype cycles, I always try to stay positive but rational.

That means saying, “Sure, probably at some point that will happen.” I care most about what I know to be true today and in the next couple of quarters, and I’ll adjust our strategy based on that.

It’s clear that we’re going to have very capable AI that can produce software. But when I speak to a lot of developers, we’re not at the point where you can sit down and say, “Build me the next social network with these functionalities,” and it just does it.

Building software is a lot harder than that. Over the last 2 years, I’ve seen a lot of cool demos, and that’s a good start, but what we really want to see is this applied at scale in production.

You can prompt a Tetris game in a browser, and that’s great. But most software today is very complex. It’s not just technology; it’s a lot of humans, customers, and feedback loops. I don’t think we’re anywhere near being able to automate those things.

I definitely think a great software engineer will be much more productive over the next 6, 12, and 18 months. But there were podcasts, especially 12 months ago, saying, “All software is dead. In the future, you’ll just go to whatever LLM you use and say, ‘Make me a copy of Monday.com,’ deploy it, sell it at one-tenth of the price, and all the existing companies will be destroyed.”

That’s a really naive view of what it’s like to build a business. Technology and lines of code are super important, but as you know, having built your own company, it goes much deeper than that. That view also assumes you’re going to continuously upgrade and improve all of those tools yourself.

Can you imagine having to continuously upgrade and improve 150 different internal tools?

Totally. Unbelievable.

Speaking of a rational and clear mindset around where we are today, everyone acknowledges that we’re seeing the complete commoditization of foundation models. Do you agree with that, and how do you see the foundation-model landscape evolving?

For sure. I think what it always comes back to is that distribution is king, and great products are king. Of course, there will be new LLMs that are better and more powerful, and I’m excited to see what GPT-5 has in store.

But I do think we’re seeing the commoditization of the text-generation layer. For most of the use cases where LLMs will transform the world, the current-generation technology is good enough. It’s about improving the base models, of course, but it’s also about building the product and the scaffolding around them.

We’re seeing xAI and Elon Musk catch up pretty quickly, and Mistral has a really great product. A lot of people prefer those models over OpenAI. OpenAI has a huge distribution moat, and they’ve really managed to capture the consumer version of the world. That will be very valuable, but I think it’s about distribution and product from here on.

You have $10 million. You can put it into OpenAI at a $160 billion valuation, Anthropic at $60 billion, or xAI at $50 billion. Which one would you choose?

I think I’d choose xAI. They’re all great companies, but I’d choose xAI because there’s the most asymmetric upside. If Elon delivers what he usually does—and we’ve all heard about the data center he built in 10 days or something—I would never bet against Elon.

The upside potential is huge, and I think it’s underappreciated that he owns X. OpenAI has clearly managed to capture the consumer as the destination you go to use an LLM. We’ll see LLMs become part of many different apps, and owning X while building the LLMs is really powerful.

X has real-time information that can feed directly into the models, and it already has hundreds of millions of users who, in theory, could start using its LLMs instead of going elsewhere.

What is the differentiator that will create the winner? Is it data? A lot of people have said xAI because of Elon’s access to compute. Is it distribution, where OpenAI would win? Is it data, where OpenAI or Anthropic would win?

There are clearly still gains to be made from scaling, especially outside the text domain—in video, audio, and 3D.

So you agree that scaling laws will continue?

I think they will continue, but I don’t think it’s going to be linear, where whoever has the most compute wins the world. The world rarely works that way. Someone will come up with an algorithm that’s 10 or 100 times as efficient as what we have today.

Compute is important, but I also think data is important. I’d say it’s compute, algorithms, and data.

What everyone is trying to build into AI systems today is control. We’ve proven that these systems are extremely capable at replicating the real world—producing video that looks real, audio that sounds real, and text that sounds real. What we all really want is a deeper level of control over these things.

In my world, take some of the big video-generation models. Compare what we do with what Sora or Runway does. Runway’s models are extremely capable and powerful. You type something in, and it produces basically anything you ask for. That’s a really powerful demo.

But to really use this, you have to be able to say, “I want this same character in a different scene,” or, “I want the character to say this particular thing.” There are so many layers of control, and they’re very hard to build in. Every time we add them, it decreases the overall fidelity, because you’re trying to make the model do what you want rather than simply replicate what the real world looks like.

Getting those layers of control in is going to be difficult. I think it will require a lot of algorithms.

To what extent is Grok’s lack of layers of control a problem or a benefit?

There’s control over content moderation, and there’s control over the output. I’m talking about controlling the output. Today it’s a slot machine: You type something in, get something out, and then try to change the prompt and pull the slot machine again.

That’s frustrating when you’re trying to get to something specific, because the model doesn’t produce exactly what you want. You want some degree of control in there.

On moderation, it will be interesting to see how it pans out. We’re definitely in the middle of a vibe shift, with Zuckerberg also pulling back on moderation. In general, I’m in favor of that. I don’t think humans moderating content is the right way forward.

I think Mark made the right decision to pull back on moderation. I don’t have the actual details of the statement, but directionally, it’s correct that having humans sit down and evaluate content is not the right way to do it.

Products like Wikipedia show that the collective power of people working together to arrive at some sort of truth is really powerful. Community Notes is taking that Wikipedia way of thinking about the world and trying to implement it into every piece of content.

It’s not easy, and it isn’t solved yet, but I do think that’s the right way for us to have some degree of control over what people do and say.

Where it gets messy is with gray content. We have that problem at Synthesia as well. We have an internal product for content moderation, and it’s been a journey for us.

The hard thing is that you have green content—content everyone agrees is great. That’s 99.9% of content. You have red content—hate speech and violence—which most people agree is bad. Then you have the gray middle, and that’s where it gets difficult.

I think that’s the content Zuckerberg is talking about when he says he’s going to loosen controls. Take cryptocurrency content, for example. When is it an enthusiastic entrepreneur who started a new coin and genuinely thinks it will change the world? When is it an outright fraud, trying to get well-meaning people to put their money into something they think will go up 10 times? Those lines are very difficult to draw.

Is Synthesia liable? Are you the arbiter of justice about what is good and what isn’t? If someone puts out an opinionated rant that’s fair but upsets a lot of people, are you the arbiter of right and wrong?

Today, we are, and that’s a decision we’ve made. We’ve had a lot of discussion about our approach to this problem. For me, it comes back to the customers.

We’re an enterprise product. It’s important that the avatars we have aren’t seen being used in all sorts of wacky content online. As a company, we don’t see ourselves as having to uphold any kind of right to free speech.

Frankly, from a business perspective, having someone pay me $30 a month to create questionable conspiracy content is not good business for me. All those things lined up. Our enterprise clients don’t want avatars affiliated with content that doesn’t match their brand, and economically it just doesn’t make sense.

We’ve taken a very strict approach, which means we are acting as the arbiters of truth. We definitely have people who are unhappy with that.

Staying on models before we go down this rabbit hole: I speak to many investors, and when Synthesia comes up, they say, “Amazing company, amazing what they’ve done, but OpenAI is going to move into this.” It’s the most obvious play for OpenAI to move into. Customer service is always the other one.

How far do you think model providers go into the application layer, and how do you assess that?

I think that’s classical VC brain, with too much focus on the technology. There’s a big misconception that Synthesia is an avatar company. It’s a fair assumption because you go to our website, and avatars are a headline feature. They’re one of the things that made us what we are today.

But if you talk to our customers, they don’t buy us for the avatar model. They buy us for the workflow. We’ve taken the entire video value chain: When you have an idea and make a draft for a video, you don’t need to use a camera or a voice-over artist because we have AI models for that. We give you a great editor, like using PowerPoint or Canva, and we give you a distribution mechanism with an AI video player designed to serve multilingual content.

The list goes on. When you talk to our customers today, the reason we have million-dollar-plus contracts isn’t that people say, “I want to make an avatar video.” It’s because they want to convey a message to someone with the highest efficiency and engagement, using the best workflow.

That’s so much more than just the models. Of course, we want to win on the models, and we are going to win on the models. Our specific niche is humans presenting content.

Sora, Runway, and all those companies are doing amazing things with AI video, but they have a much wider remit. They’re trying to build true foundation video models. Maybe we’ll be customers of those companies one day, or maybe we’ll be partners. The only thing I care about right now is voiceovers and humans presenting to the camera.

In that niche, and when you take the entire platform as a whole, that’s how we win. The winning companies over the next couple of years will be the ones that aren’t overly obsessed with the AI component of their platforms, but that build for workflow.

As we move into 2025, the models are very important for us, but we’re expanding across the value chain. We increasingly want to be the publisher of videos for customers. Instead of using YouTube or another enterprise video-hosting platform, customers can use us directly. That’s where you’re going to see a lot of value being created.

There will be a few foundation-model companies that become very big. They already power us—we use OpenAI and Anthropic.

What do you use most, and has that changed over time?

As a company, we use OpenAI the most.

Wow.

But that is shifting more toward Anthropic. We’ve shifted some workloads to Anthropic, and we’re also using Gemini.

In general, it’s about pricing. Some models are better at certain things than others, but we have a lot of workloads. Moderation, for example, is a big one. Pricing is important, as is the scaffolding and infrastructure around the models.

How did you think about whether to build your own models? How do you think about that?

We only want to build models if we think we can be the best in the world at them. In general, that means they’re in a narrow domain and directly tied to a bigger workflow.

We always work backward from what our customers are trying to do. If we think we can be the best in the world at something, and there are compounding advantages to building that model, then it’s great for us.

Specifically, for humans presenting to camera and dialogue-driven content, that’s what we want to be the best in the world at.

Do you think we’ll have a world of many smaller, specialized models, or will we have 3 massive general models?

There will be some concentration among the big companies, but I also think we’ll see many specialized models. It’s interesting, though, because “specialized models” is still very technology-centric.

There will be many different products that people use in their daily lives to run their businesses, and a lot of those will probably use open-source models tuned specifically for something and integrated into an overall platform.

If you think of that as a specialized model, then yes, we’ll see many specialized models. But it’s very tech-industry to talk about models all the time. Most people will interact with models in things like ChatGPT, but a lot of the things you don’t notice as a consumer—when you’re using someone else’s platform or product—will be workflows in the background driven by LLMs.

You’ll definitely see companies building their own specialized LLMs, but I think we should think more about products than models.

Final one on models. We talk way too much about models, but GPT-5 has taken so long. Everyone’s asking when it’s coming, and it’s been delayed repeatedly. How do you think about the delay, and do you think it will surpass expectations or fall beneath them when it’s released?

It will be incredibly difficult to recreate the ChatGPT moment when it first came out. As humans, we saturate incredibly easily. We often forget how crazy GPT-3 actually is when you use it.

My sense is that when GPT-5 is released, it will probably be good. A whole bunch of tech people will be very excited about everything it can now do. Maybe I’m wrong and it will be a superintelligent thing that self-improves and saves the world, but I think it will come out and be good—definitely better than the previous models.

Most people probably won’t care that much, because the bar for someone to really care is incredibly high. There’s all this talk about whether something is AI or not AI, and those definitions are real, but what’s the definition of AGI?

If you took ChatGPT 200 years back in time to England and showed it to someone, you’d be burned at the stake. If you did it 50 years ago, people would think, “This is AI. I’m talking to a computer that knows everything about the world. This is absolutely batshit crazy.”

Today, people say, “It’s just a stochastic parrot,” and I don’t know where the truth is in that. I think it will be very powerful, but I don’t think it will have the cultural moment ChatGPT initially created.

There will be a small group of people who care deeply about which benchmark it beats and how it performs in different areas, and that’s awesome. But for most use cases and most businesses around the world, the current models are already pretty good. They can solve a lot of problems.

One of my favorite quotes is from Eleanor Roosevelt: “Great minds discuss ideas, average minds discuss events, and small minds discuss people.” That’s what I aspire to—not necessarily what I am, clearly.

I want to discuss the future across a few different segments, starting with content creation. This is my life. What does the future of content creation look like in 5 years?

In very broad strokes, what happened is that we democratized distribution with the internet. There are no gatekeepers; anyone can make a website and share content online. That’s been very powerful.

Then we saw, to some extent, the democratization of content creation because all our smartphones have cameras, and the price of camera equipment dropped. The setup we’re sitting with today would probably have cost 100 times more 30 years ago.

What’s about to happen is that we’re truly going to democratize content creation. We’re going to change the world of making content from something that requires capturing things with sensors in the physical world to something where you can generate everything digitally.

That’s going to have huge implications. Think about text, which we don’t really think of as a technology anymore—it’s a fabric of society. Text went through this journey: The printing press was one of the first major inventions around text, then we invented keyboards and computers, and now everyone can create content. It’s entirely digital.

With music, we’ve also seen it go from something we generally do in the real world and capture with sensors to something we can do digitally. Most music today isn’t made with real instruments; it’s made with software instruments.

When I open my MacBook at home, I can recreate almost any song I want without real instruments. I can synthesize it on my computer, and that changed the music industry in many ways.

We’re going to see the same thing with the video and audio industries. You’ll be able to bring your ideas to life without needing much more than your imagination.

The most obvious effect is that the cost of creating content will go to zero. It costs basically nothing to write a book today. In the not-too-distant future, the purely technical cost of creating a Hollywood film will also be close to zero.

That’s going to mean a deluge of content—an absolute deluge. It means that people like you, YouTube creators, and everyone else can compete with Hollywood in terms of what they can bring to life.

Today there’s still a separation between Hollywood content, like the Netflix series and high-production-value films we watch, and YouTubers, who make prank videos, podcasts, and other fun things. Even for you, it would be very difficult to make a Hollywood-quality production because it would be extremely expensive.

That separation is going to flatten, which means the best ideas and the best content will win. I think that’s going to be pretty exciting.

I think we’re so far away from this. I create content for a living, and when I look at the clipping tools, they’re so far off what my team can do upstairs. It’s incomparable. I don’t see that being 2 years away.

It’s also highly ambiguous. There’s a lot of knowledge involved in deciding which parts are important. Depending on what we wanted to achieve with this episode, we could cut 6 different sections as the best highlight reel.

I find these clipping tools are still far off. I think content is less democratized than ever.

I totally agree with you. The important distinction is that when I say you’ll only need your imagination, I don’t mean it won’t require skill.

It means the technical barriers will be lower. You won’t need a camera, and you won’t necessarily need to be amazing at visual effects. A lot of that work will be done automatically, but it will put more emphasis than ever on being a great storyteller, understanding what’s important, and asking the right questions.

If you’re creating fictional content, you’ll need to combine things that work in the cultural zeitgeist while still standing out enough to be interesting for people to watch. All those things will be more important than ever.

What will be less important is knowing the right people to distribute your content to or spending 8 years becoming the world’s best visual-effects artist, because you can get a lot of help with that from AI systems.

The storytelling piece will remain exactly the same. I also think the value of real content will be higher than ever, because we’ll have a lot of AI-generated content, and that will be a category of its own.

Just as we still like to watch people play the piano, or like knowing that the artist we’re watching is a real person, I think that will carry over into this world as well.

Will we have more human-made or AI-made content in 5 years?

Definitely AI-made content, but I think what’s more important is what content surfaces. If you go to YouTube today, I don’t know the exact statistic, but probably 95% of all videos have fewer than 10 views. The content is already there; the question is what gets surfaced.

That was my point. When you reduce the cost as much as AI does and increase supply so much, what’s the future of discovery?

I think TikTok is a really great product. This may be an unpopular opinion, but when I look at my TikTok feed today, I’d say 90% of it is highly informative and educational. It’s about music production, the music I like, politics, and technology. I see your face once in a while. It’s an amazing feed of interesting people based on my interests.

What TikTok has done really well is build the graph based on your interests, rather than your social connections. That model is extremely powerful.

I would like to see a bit more control over your algorithm. Today, there’s an intuitive thing everyone does: If you see content you don’t want to see, you swipe quickly because the algorithm notices that you swiped past it. You’re almost training the algorithm yourself.

I think that’s an excellent way of doing it, but I’d like more control. You should be able to say, “Don’t show me these particular topics. I’m not interested in them.”

TikTok has also done a great job with its interest graph. When you post a piece of content, it shows it to a couple hundred people and quickly senses whether those people like it. If they don’t, it goes in the trash. If they like it a little, it shows it to more people and tries to figure out whether it’s great content.

If you look at TikTok’s usage, that’s clearly working very well.

Should TikTok be banned?

TikTok has built an amazing product. I don’t think it’s a brainwashing machine or a propaganda machine.

The denigration of social media is unfair. To your point, you said your feed is educational. Mine is motivational speeches, air-fryer recipes, and workout routines. It’s highly additive to my life in positive ways.

Just as a diet can consist of chocolate and sweets or vegetables and fruit, you choose your own content diet. That’s a human responsibility, and I think it’s unfair to demonize social media.

I totally agree. That’s the paradox. You and I may have built our algorithms to be productive for our lives, but it’s also true that a lot of hours are wasted watching dumb content.

I don’t think that’s necessarily bad. We all need to watch dumb content once in a while. I do, too.

In general, though, I’m optimistic about the social-media and content landscape. We have the internet, so information is free. We’ve increased the democratization of content creation, both through cheap cameras and through AI.

We have Community Notes emerging as a system for figuring out what’s real and what isn’t. There are a lot of problems in the ecosystem, but I think we’re finding our way toward building an awesome media ecosystem driven by citizens and experts rather than journalists and politicians.

We still have some of the control that was good about the previous generation of media, where a few corporations owned the newspapers and content was reviewed by 6 different people before it went out the door.

That’s what always happens in the world. We invent new things as humans, and we learn as we go. There are definitely negative consequences, and no one would disagree with that, but I think we’re finding our way.

I’m actually very optimistic about the future.

I also think the quality of the content you consume is inherently better. I mean this in the nicest way, but I think the content we produce in venture capital is inherently 10 times better than what a journalist at a large publication produces, because it’s our job every day. We live and breathe this.

Exactly. I gave a TED Talk a few months ago about this. My provocative opening statement was that we may be the last generation to read and write.

There’s a lot of moral panic about TikTok—for example, that people’s attention spans are decreasing massively. Maybe those are real problems, but what if we’re just tired of overly dense, long, boring content?

A 2-minute TikTok video may be a better way of understanding a topic than a 20-minute YouTube video, which may be a better way of understanding it than a 300-page book.

What do you say to the idea that we’re just dopamine junkies, getting jacked on 2-minute clips because we can’t be bothered to read a 5-minute piece?

The dopamine-junkie idea applies to everything. That’s how we govern our lives most of the time. I understand the analogy, but is it all dopamine addiction if you do sports every day, or whatever else makes you feel good?

Most things we do are attempts to make ourselves feel good in some way.

One of the problems is identity misuse or fraud. How do you think about the future of identity verification, both for Synthesia and across platforms?

I do think we need some degree of identity verification for the big spaces we use. We need to know where content comes from, who created it, when it was created, and how it was created. We need identity verification to some extent, at least in certain spaces.

But I also think we need to take that all the way down to the content level. Imagine you had Shazam for every piece of content on the internet. It could tell you, “This picture was originally uploaded 5 years ago. It was first uploaded by Harry.”

In the future, it could also say, “This was partly made with Synthesia,” or, “You’re watching an adapted version,” or, “This was made with AI for translation.” Those are the kinds of signals we can give around content.

If you had that, plus identity verification and Community Notes, every time you scrolled past a piece of content you’d see whether it was verified. Today, only celebrities used to have verified accounts because people tried to impersonate them. When you saw the trusted verification sign, you knew it was actually the celebrity posting.

I think we should flip that around so that everyone and all content is verified. If we have the provenance trail, we know where content came from, maybe how it was created, and other information around it. That gives you a green check mark.

Something unverified should stick out like a sore thumb. If there isn’t a fingerprint for the content, or we don’t know where it came from, that should be what sticks out—not the verified sign on a celebrity’s page.

Does that not inherently lower the value of your content, though?

Why would it lower the value of content?

Because your content is AI-generated, so it wouldn’t have the green tick.

No, I don’t think it lowers the value of the content. We’re already in a world where most content we consume is, to some extent, fake. Most pictures on Instagram have been touched up in some way. When an image leaves the sensor, it already goes through color correction and a whole bunch of other things.

The line will continue to blur. I think it will be less about how you made something and more about the content itself. You could make a video of yourself talking about something, and in a couple of years no one will be able to tell whether it’s real or not.

What matters is that you made it and that what you say in the video is something you actually wanted to say. It’s similar to how we don’t care whether a Hollywood film used a green screen. You care whether it’s a good film or a bad film.

We should evaluate content based on the content itself. How it was produced is certainly a signal, but so is when it was produced, who produced it, whether you’re watching the original version, and whether you’re watching an edited version.

It’s difficult, but you can imagine a world where you build a chain showing where content was originally created and what happened to it afterward. Maybe you’re watching a 25-second outtake from Harry’s podcast, and you can click to watch the full segment. You can trace it back to the full episode.

The reason I mentioned Shazam is that it exists because this was a commercial problem for YouTube. Record companies told YouTube that people were uploading videos using copyrighted music, and they would sue YouTube out of existence if it didn’t fix the problem and pay them.

YouTube’s solution was to detect whether something was copyrighted. When you upload a video, it scans it against a database of billions of songs. From a 4-second clip, it can identify the song within milliseconds. If it’s copyrighted music, YouTube either takes the video down or, most of the time, puts an ad in it and pays something back to the rights holders.

You could imagine the same thing happening with every kind of content. You upload a video, and it says, “This is a clip from Harry’s original podcast, released 2 weeks before Victor put it on his YouTube channel. Click here to watch the full thing.”

You begin to build this chain of content and identify where things came from. That requires a centralized database—not ideally a decentralized one—where, every time you make content, you register the first time it was created.

I hate to say it, but it could actually be a blockchain. You register when a piece of content was first created, then upload it to different social platforms and distribution systems, which check it against that record.

One of the big problems today is that people take a picture from a war 5 years ago and say it happened yesterday in some country. With a system like this, it would say, “This picture was uploaded 5 years ago.” You’d see an automated Community Note saying, “This picture was taken by a BBC journalist 7 years ago.”

A lot of this is hard, but we already have many pieces of the technology. When we think about identity verification, moderation, and content provenance, we’re beginning to see the birth of an entirely new content ecosystem that could look very different from what we have today.

That leads to my next question. How will that new content-creation ecosystem change the labor market for content? Which roles are legitimately threatened, and which will be created?

Very technical roles, like camera operation, will probably fade away to some extent. More emphasis will be placed on creativity, storytelling, and being relatable—all the things that make great content. Those will be more important than ever.

What about a thumbnail designer or a visual-effects artist? Who is threatened, and whose job is going to be created?

I think most of those roles will naturally transition into something different, assuming you’re open-minded enough to do it.

Take visual effects. As we talked about with clips, AI can make a first pass that’s perhaps better than random, but much worse than what you can sit down and do yourself. You plus AI can probably be really, really good.

If you’re a visual-effects artist, it may mean you become like a software engineer. You’ll be 100 times as efficient, and instead of relying on 20 other people to create a digital clone of me by modeling my face, you’ll prompt it or control it directly.

If you’re good at that, you’ll probably be even more powerful in this new world.

I think we’ll have many more content creators than we do today. Before the internet and computers, how many people wrote things? Very few. In the 1930s and 1940s, most people didn’t create text. There were specific people, like secretaries, and later there were typewriters.

Now all of us produce content all the time, and that will be a big trend.

Most of the limit on how quickly AI becomes part of daily life is not actually the technology. As I said earlier, we have very powerful LLMs that can do a lot of great things. They’ll get better and better, but the rate-limiting factor is people.

People have to use these things, trust them, pay for them, buy them, and integrate them. That’s always the case with technology. It takes a long time.

The final topic I want to discuss is location. You decided to do this in London, which I love as a Londoner, but it’s a different choice. Why did you choose to do it in London?

The honest answer is that I couldn’t get to the US. I didn’t have an amazing CV before I started Synthesia, and I couldn’t get a US visa. You need sponsorship if you want to go to the US, and I couldn’t get it. I wanted to start a company, knew a few people here, and moved on a whim.

How would Synthesia have been different if you had been in the US?

I still think the chance of success is higher if you’re in the US, particularly on the West Coast. There are a lot of advantages. Over the last 6 or 7 years, it’s gotten much better to build here, but I still think there are advantages to being in the US.

Europe is catching up pretty quickly, though.

What are the advantages of being here?

There’s one big one, and it goes both ways: talent. If you look at the ecosystem in the Valley, one factor is the price of talent, which is abnormally high.

There’s also loyalty, which I think is less discussed. In the US, you have a lot of people who are very driven and view their career as building a portfolio of stock options in different companies, hoping one of them explodes and makes them very rich.

That’s a rational way of thinking about it, and it makes sense, but it also means the relationship people have with their employer is more transactional. People are more willing to jump ship. You have 1 or 2 bad quarters, and you may lose a lot of good people because there’s another cool company on the block.

In Europe, people think less like that, for better and for worse. Stock options in Europe are still viewed as a random lottery ticket rather than part of compensation, partly because no one in Europe knows anyone who got rich working at a startup. We haven’t had that many successes here.

Are you seeing that attitude change?

A little bit, but it will take a long time. No matter what I tell people at Synthesia, and no matter how much they read the news and see what’s happening, what they care about is whether they know someone—a cousin, for example—who made $2 million by being early at a startup in the US.

Especially in the Valley, everyone knows someone who made $2 million working at a startup, and they know several people. That makes you feel like you can do it as well.

Here it’s abstract. You hear that some companies made money, but before we see that happen more broadly, it requires a bigger ecosystem with more exits, more large companies, and so on.

What you get in Europe is more loyalty. People care about working somewhere they like, working on interesting problems, and having great colleagues. That’s a benefit that’s less discussed.

Do you think people work as hard?

I don’t have a strong sense that people work harder in the Valley. I know a couple of great friends who have moved over there and opened offices, and I haven’t heard them say, “People work insanely hard.”

There’s probably a percentage of people who work extremely hard, but we have those people here as well. That’s more of a personality trait dispersed across the world.

If you take an average software engineer in the Valley, I don’t think they work harder than people in Europe.

Why do you think it’s gotten better to build here? I feel like it’s gotten much worse in the last 6 months. Do you share my opinion?

It’ll be interesting to see what the new government does. They have a lot of great cards in their hand. There are a lot of great companies building here, and what’s important about the London ecosystem is that we’re building global leaders, not just regional winners in areas like expense management.

You have ElevenLabs leading in voice, Synthesia leading in enterprise AI video, and Wayve, which is one of the top 3 companies in the world doing self-driving. We have a lot of great companies becoming global leaders.

It’s important that the incoming government doesn’t screw this up.

How could they screw it up?

Through overtaxation and all the usual things people talk about. What they can do to make it better is fix the UK’s brand problem.

What do you mean?

I think we disagree on this, but I think the UK is a pretty good place to build in general. I’m from Denmark, and that would have been a terrible place to build a company. It would have been almost impossible.

I think there are a lot of good things that have happened in the ecosystem. You have the talent, the venture ecosystem has improved, even though it’s still very private-equity-centric, and the country has gotten better.

But there are still ways to screw it up. I see a lot of people leaving, especially after the recent administration. My phone is literally full of messages asking, “Where are you moving? Where should I move to?”

From billion-dollar founders?

Every single day.

You’re sitting down with Keir Starmer, and he asks what he can do to make the best developers want to build here and the best founders want to stay here. What would you advise him?

The obvious things are corporate tax, entrepreneurs’ relief, and all those measures that make it attractive to build here. We’re in stark contrast to much of Europe, which is a terrible place to build.

Don’t screw up the fundamentals. Don’t raise taxes too much, and make sure we still have great relief schemes for people who create jobs and economic productivity.

On the brand problem, it starts with fundamentals: crime on the rise, the cost of living, and all the things you’ve been vocal about. Those are big problems.

But it’s also the consistent negativity from this Labour government. Stand up and say, “London is open for business.” It’s negative, negative, negative. Sell the story.

I completely agree. That comes back to the brand problem. As much as we have people here who want to leave for the US, one of my learnings has been that a lot of people want to come to London as well, especially from the US.

In our journey, I’ve found it easier to move people from the Valley to London than from the Valley to New York, which feels counterintuitive. A lot of people want to experience the European lifestyle. They like that London is a cultural melting pot, along with all the things I love about London.

If we worked more on the brand, we could attract a lot of great people to come here and build. Of course, it starts with the fundamentals.

If you look at California versus London in terms of taxation, I’m not sure there’s a massive difference. California also has very high taxes and can be pretty rough. I wouldn’t over-index on that.

I think it’s a case of doubling down. There are a lot of good things happening in the UK, and if we don’t screw it up, we just need 1 or 2 companies to become billion-dollar-plus companies headquartered here. Then we’ll really start to see the ecosystem grow.

A couple of $10 billion companies would crystallize those exits.

Wait for it.

And they IPO.

Yeah, that’d be great.

I wouldn’t know where they IPO, though.

That’s the point.

Synthesia on the Nasdaq?

It’s a joke, but the London Stock Exchange needs to fix that. It needs to put some liquidity into the system. I’m not a financier, so I don’t know exactly what it needs to do, but it’s a disaster.

We’ve unfortunately seen great companies suffer because of it.

You had Nick from Revolut on the show, and he said, “I’m highly rational. This is not a rational decision,” but it really isn’t. It was the most elegant way of saying, “No, I’m not doing that.”

It’s not very attractive.

Do you think that’s actually a problem?

The answer is, “Who gives a shit?” Liquidity is liquidity. Whatever market you get it in, it still makes a thousand people in London who hold early stock into millionaires.

I don’t think it’s in my top 3 things that need to be fixed.

What’s the top thing that needs to be fixed?

Keep the reliefs for entrepreneurs and people who create things. Maybe even make them better, because that’s very attractive.

You could also subsidize GPU costs to incentivize UK data centers. That would be an actual competitive advantage.

Obviously, don’t overregulate. We haven’t seen much regulation yet, but there’s been a bit of a vibe shift from the last administration to this one.

Don’t look at Europe as an example. Europe has all the AI regulation and none of the AI companies. Staying the course and not screwing it up is really important.

I want to do a quick-fire round with you. I’ll give you a short statement, and you give me your immediate thoughts. Does that sound okay?

Sure.

What do you believe that most people around you disbelieve?

I think playing computer games is probably one of the best things a kid can do for future success.

Why do computer games make for better entrepreneurs?

Because Toby at Shopify said that if you can run a clan, it’s better than university. That’s exactly what I did: World of Warcraft.

No, actually, what I think it is, is that we’re probably the first generation that had access to complex strategy games. What you learn throughout life is decision-making, and computer games are a microcosm of the universe.

You can run so many simulations. It may look to parents like you’re playing Warcraft 3, Red Alert, or World of Warcraft, but what you’re really doing is training yourself to make a lot of decisions rapidly and understand the implications of each decision in the game.

Life is a game. Careers are a game. Building companies is a game to some extent. We’re the first generation that could sit by ourselves and get so many iterations in, which would otherwise have been impossible.

Before the computer era, you couldn’t simulate lots of small decisions about how to run a business. You had to actually build and run a business. With a computer, you can play RollerCoaster Tycoon.

RollerCoaster Tycoon isn’t the same as running a real theme park, but it may be 20% of what it’s like to be the CEO of a real theme park. That teaches you a lot.

The decision-making and sense of strategy you build from games are highly valuable because they create a 2-way feedback mechanism. You can read books and understand the world through them, which is obviously powerful, but what’s powerful about computer games is the constant feedback loop.

What trait are you slightly ashamed of that has contributed to your success?

I’m a generalist in the purest sense of the word. When I was younger, I wanted to be an artist and a great computer programmer or developer, but I never really excelled at any of them.

I still have a dream of making music someday, but I think what I’m really good at is knowing a lot of things about many random subjects. I’m extremely curious. I’m not an expert in anything specific, but I have a very wide range of interests.

My neural network has been trained on random things, from how elevators work to black metal in Norway to how people make technology. I used to get annoyed with myself and wonder why I couldn’t just become the world’s best music producer, programmer, or AI researcher.

I’ve realized that my power is being decent at many things. In my role, what you do is make decisions and get ideas. You can’t force those things.

I truly believe that companies are subconscious, and the more varied things you put into your brain, the better you get at making analogies and pattern-matching.

Vu told me you were the resident DJ, which I thought was a brilliant description. “The heaviest things in life are not iron or gold, but unmade decisions.” What unmade decision weighs on you most?

I studied a combination of computer science and business at university, and I really regret not taking a pure STEM degree—a pure computer science degree—because I think studying business is mostly useless.

You have to learn business by doing it. There are probably 10 basic ideas in business, and if you learn those, you’ll be pretty well covered.

That goes back to the point about being a generalist. I should have done a pure computer science, math, or engineering degree.

What have you changed your mind on in the last 12 months?

One thing I’ve changed my mind on during this AI hype cycle is the importance of ignoring the noise. When you’re in the middle of these cycles and see everything getting funded and all these competitors appearing, you get a lot of noise.

You start asking whether you should do certain things, listen to what the market says, or copy what a company is doing. I’ve relearned the lesson that you should never do things out of guilt because other people are doing them.

Don’t look too much at competitors or at what the broader AI landscape is doing. Listen to the customer. They’re the ones who pay your bills, and ultimately they’re the ones you need to make happy.

I had a period where I may have over-indexed on what other companies were doing and on the AI space generally as the most valuable signal, instead of the customer.

It sounds more like relearning a fundamental lesson.

You can only choose one investor from your cap table for your next company. Who do you choose?

I can’t answer that.

You can. I won’t let you out of this room. Choose 1 investor.

That’s a hard one. I’d take Mark Cuban again because he truly believed when absolutely no one else did. I’ll be forever grateful to him for doing that.

That’s a clip. You’re thinking about clips when you’re me these days.

What about the way your parents brought you up would you deliberately do differently with your future kids?

My parents’ generation saw computers as inherently socially isolating—you were sitting alone and dopamine-addicted to clicking buttons. That’s exactly what people say about TikTok today.

The reality is more nuanced. When I played computer games as a kid, it was a very social thing, both because I played with other people online and because our parents had a hard time believing you could be friends with people you hadn’t met in real life.

They also thought it was a waste of time. If you’re playing Angry Birds, maybe it is a bit of a waste of time, but if you’re playing strategy games, I’d put that alongside solving math problems or reading STEM books.

Everything should be done in moderation, but I’ll definitely encourage my kids to play great computer games.

Every parent listening is questioning their parenting decisions.

Final one: What question are you never asked that you wish people would ask you?

The one I didn’t prepare for.

There’s one I find interesting. I presume you’ve had some secondaries now. Has having some secondaries had the mentally freeing effect that people often say it does?

I definitely think so. Not having to worry about financial stress, and not having everything you own in 1 stock, frees you up to make better decisions.

I’d encourage everyone to take secondaries off the table when they can. I’m 100% certain that if we hadn’t taken any secondary at all, we would have thought about selling much earlier.

You might make more rational decisions when you’re not carrying all that financial pressure.

Dude, this has been one of my favorite shows. Honestly, it’s painful. It sucks when someone you met years ago is now a guest on the show.

It sounds dickish, but we have a high bar for what you have to be to get on the show, and that just shows what a fuck-up I was. Thank you so much, dude. You’ve been amazing.

Thanks, dude. This was great.