Scale AI co-founder Alexandr Wang had been at Meta for less than a year when he led a team of fewer than 200 people to build Muse, the core product behind Meta's bet on consumer AI. Behind it lies a wager on direction, seven months of refinement, and a story about how to make a product truly break through.

On October 11, Meta Chief AI Officer Alexandr Wang shared the process of his team building Muse on the well-known American podcast David Senra.

Llama 4 Stumbles, Zuckerberg Steps In Personally

In 2025, Meta's Llama 4 failed to meet the company's expectations.

Wang described the backdrop: "Llama 4 wasn't on the trajectory Meta wanted as a company. It was genuinely a disappointment internally, and it happened at a moment when AI was becoming increasingly important."

It was at this juncture that Zuckerberg proactively reached out to Wang, the Scale AI co-founder. The two had been introduced by a mutual acquaintance earlier but didn't interact frequently, speaking only about once every six months. This time, Zuckerberg sent a message: "Do you have time for a call?"

The initial conversations weren't clear in direction. Wang said several calls consisted of Zuckerberg asking him "what do you think we should do," with no clarity on where things would end up. It wasn't until acquisition terms came up that Wang realized the weight of the situation. Ultimately, the two sides completed the entire deal in about five to six weeks: Meta took a 49% stake in Scale AI, Scale continued to operate independently, and Wang himself along with a few core members joined Meta, with Wang becoming Chief AI Officer.

Why "Personal AI" Rather Than Coding

After joining Meta, Wang and Zuckerberg faced a key strategic choice.

In early 2025, Anthropic's Claude and Opus were far ahead in the coding domain, and the industry consensus was that coding agents were the most important AI application direction.

Wang had a different judgment. He said the first thing he told Zuckerberg was: "Meta is actually a very special company, because Meta's products represent what people want to do—Instagram, WhatsApp, Facebook, these are all places where you choose to follow your interests, hobbies, and friends."

In Wang's view, Meta has the distribution capability to reach 3.5 billion daily active users, which naturally points to consumer AI, not work or coding scenarios.

Consumer AI, personal AI—this has always been the area where we believe Meta can build something truly special. Not work, not writing code, but personal.

This judgment ultimately materialized into Zuckerberg's internal memo "Personal Super Intelligence" published in June 2025, which Wang and Nat Friedman helped draft together.

Nat Friedman's "Fear and Ecstasy" Ignites the Product Direction

What truly convinced the team that the "personal AI agent" direction had disruptive potential was a firsthand experience from Nat Friedman.

Friedman is one of the core figures at MSL (Meta Super Intelligence Labs), an early promoter of GitHub Copilot, and an early Stripe customer. Wang said he was the first person within MSL to deeply use OpenClaw, a personal AI agent product.

"His experience, I can only describe as somewhere between fear and ecstasy," Wang said. Friedman told Muse he wanted to drink more water, and the AI then monitored him through a surveillance camera to see if he was drinking water, and told him "good job."

Friedman then shared a prompt designed specifically for psychological self-analysis. When Wang used it for the first time, he experienced similarly intense feelings:

It was like compressing three years of therapy into a single moment—it could access my emails, my photos, integrate all the information together, and I was extremely vulnerable throughout the process.

In February 2026, Wang and Friedman each wrote memos, separately sent to Meta's board. Wang wrote that he believed super agents were "in some sense the endpoint of consumer products." Friedman's memo contained two lines that later became core product principles: "the clause is the law" and "trust is a must."

7 Months, 200 People, One "Work of Art"

About one to two weeks after the memos were submitted, the team completed the first prototype and demonstrated it to the board. Wang said many core designs in the final version of Muse already existed in that earliest prototype—including the mascot "Jolly" and an early version of the Muse Charm.

But going from prototype to product was a "long and painful" process.

Sometimes the prototype would work, and some moments were magical, but most of the time it didn't work properly and would constantly crash. We had to turn it into a highly reliable, carefully polished consumer product.

This journey lasted 7 months, with fewer than 200 people involved.

Wang said the team built a spreadsheet containing over a hundred specific behaviors and capabilities, setting "release-blocking" thresholds and "passing" standards for each one. After every model iteration, they would check against it. "We had many checkpoints—say 80 items were green, but 20 were still red, and we knew that wasn't good enough," he said. The first checkpoint where everything was green was a specially trained version of Musepark 1.3.

He compared the process to polishing a work of art: "It's a work of art that requires a strong singular vision and intense focus to turn into a complete, beautiful experience."

In Wang's view, this is precisely why Muse succeeded while other early personal agent products failed.

Products from large organizations often become a Frankenstein of different people's opinions and visions—every product manager wants to stuff something in, and what you end up with is a smoothie. Users can feel it.

He particularly emphasized that Muse's product vision primarily came from Nat Friedman: "If there's one person whose singular vision runs through all of Muse, it's him."

Meme Marketing: Making Meta Cool Again

After Muse launched, Alexandr Wang began a密集 round of "unhinged" posting on X—creating various memes using the product mascot Jolly (a little person figure), triggering massive retweets.

He admitted this wasn't entirely intentional.

I'd be in a meeting and a meme would suddenly pop into my head, and I'd feel I had to post it. Sometimes I'd be having dinner with colleagues, and suddenly I'd think of one and just post it.

His logic: the internet rewards risk, rewards the unexpected, rewards things people don't see coming.

Nat Friedman said at internal meetings during that period: "I thought we would learn a lot about Muse users from this, but actually we just learned a lot about what's going on in Alex's head."

Alexandr Wang believes this marketing approach actually drove growth—though he acknowledges it's hard to track precisely.

He also criticized competitors' advertising strategies:

Anthropic did massive outdoor advertising, all about themselves. OpenAI's ad was just two people sitting at a table. I don't understand what that means. You have millions of real use cases, your ad just needs to show what the product can do for users, it's that simple.

Over 200 Direct Reports, Against Bureaucracy

At Meta, Alexandr Wang has over 200 direct reports—almost all researchers who joined the MSL research team (internally called TBD) report directly to him.

He explained this isn't a management structure but an anti-bureaucracy statement.

These are all extremely brilliant people who have already proven they can do exceptional work. I don't need to micromanage them. What they need isn't a manager, but an excellent environment.

He positioned his role at Meta as a "coach" rather than a "founder" as he was at Scale AI:

At Scale, I had strong opinions about every little thing in the company; the company was a product of my effort and perspective. At Meta, I think more about: how do I create an environment where these brilliant people can do their best work?

AI Can Give Everyone "the Adulthood They Dreamed Of"

Speaking about Muse's long-term significance, Alexandr Wang expressed a broader judgment:

I truly believe in personal AI agents as a long-term form. Humans are born with desires and dreams. AI and personal agents will become the mechanism and bridge to help us realize those dreams.

He believes most people's ambitions gradually get worn down after entering the workforce:

We all become somewhat zombie-like. We lose our agency, we lose our ambition.

He wrote in a post:

The world so far has been shaped by a small number of fanatics who managed to make their will reality. But we've never seen a world where every single human being's agency is fully activated.

He believes the direction Muse represents is a step toward that world.

Full Interview Transcript:

Building Muse, Working with Mark Zuckerberg & The Future of AI | Alexandr Wang, Meta

October 11, 2026 · David Senra

Alexandr Wang is Meta's Chief AI Officer and co-founder of Scale AI. In this episode, we discussed the seven-month process of building Muse, why Meta bet on a personal AI assistant rather than a coding assistant, and how Mark Zuckerberg and Nat Friedman shaped the product. Alexandr also shared how Zuckerberg recruited him after Meta's Llama 4 setback, why he chose to leave Scale after nine years, and how he rebuilt Meta's AI lab around small teams, top talent, and minimal bureaucracy. We also explored how to use memes to make a product go viral, the importance of having a singular vision, and why Alexandr believes AI can empower everyone to pursue grander ambitions.

Chapter 1: The Vision Behind Muse

David: Tell me about the history of Muse.

Alexandr: It starts with the vision when we founded Meta Super Intelligence Labs (MSL). At the lab's inception, Mark published a memo that Nat Friedman and several other colleagues and I were deeply involved in drafting. The memo was called "Personal Super Intelligence." The core idea was: how do we build truly powerful AI that makes everyone's life better? How do we build something that elevates the human experience as a whole?

That memo laid out many ideas and essentially foreshadowed where Muse ultimately went, though at the time it was still a very early concept—it was June 2025, agents hadn't emerged in any meaningful sense yet, and AI that summer was at a completely different stage. But that was the direction we wanted to build, it was MSL's mission, and it was the endpoint we wanted to push everything toward.

Then we started building models, because any frontier model's end-to-end development process takes months: you need to build the entire tech stack, create datasets, pre-train the model, then post-train it, and build the full engineering support system for it—it's a very long end-to-end process.

Chapter 2: Discovering the Potential of Personal AI Agents

Alexandr (cont.): At the start of this year, this entire "production line" started running, and the release of Opus 4.5 showed many of us, including our team, the future of agents, truly bringing the potential of agents to the forefront. That was early this year, and Anthropic was on the rise.

My close collaborator Nat Friedman, I believe, was the earliest person within MSL to use and experiment with Claude. His experience, I would describe as somewhere between "fear" and "ecstasy"—he was truly all in, completely trusting it.

He told some stories, also mentioned in a Stripe Sessions interview: he told his Claude he wanted to drink more water, so Claude monitored him drinking water through a security camera, then praised him for "doing a good job." He has countless such astonishing stories, and they make you realize: there's something quite magical happening here.

I remember something Nat said to me that left the deepest impression: he said, "I think there's really something here." Early adopter Nat Friedman's judgment is a fairly reliable signal—he was one of Stripe's earliest customers, a pioneer in many open-source developer tools, and involved in creating GitHub Copilot and many other projects.

So it was a clear signal: in the personal agent space, there's clearly something quite magical. When you truly trust a personal agent and truly bring it into your life, it can do so much for you.

After hearing all these amazing stories, I started using it deeply. For several weeks, the two of us spent almost every free moment talking to our respective Claudes, immersed in exploring the possibilities of personal agents.

Chapter 3: Experiencing Three Years of Therapy in a Single AI Session

David: Was your initial reaction ecstasy or fear?

Alexandr: At that time Nat had built a prompt for a comprehensive psychological self-analysis, conducting deep psychological probing in multiple waves. He basically wrote a prompt specifically for this, to be used with Claude. That was also one of the first things I did with my own Claude—it was like compressing about three years of therapy into that single moment.

David: Wait, I'm curious—why was that the first thing you did when you got your first chance with Claude?

Alexandr: After I downloaded and installed it, Nat was right there and said, "You should try this." I said okay, did what he suggested—and then that crazy experience happened.

David: Tell me more about that "crazy experience."

Alexandr: It was very, very intense. It really was like three years of therapy, because it was continuous rounds of deep psychological probing—it had other information about me, could access my emails, a large number of my photos, and could look up other information, then integrate all of it together. Throughout the process, I was very vulnerable.

That was February. Both of us reached the same conclusion: there's clearly something very special happening here. We each wrote a memo—both submitted to Meta's board through relevant channels, since it happened to be board season.

In my memo, I wrote that I believed this was, on many levels, the ultimate consumer product form, and that super agents as consumer products represented a clear endpoint in some sense. Nat wrote two, with very memorable titles—"the clause is the law" and "trust is a must"—and these two memos ultimately defined the core philosophy behind how we built these products.

Chapter 4: Seven Months Building Muse

Alexandr (cont.): In February we reached this conclusion: there's something very important here. We then assembled the team and built a prototype within about one to two weeks, demonstrating it to the board—a very rapid iteration cycle. Many design decisions in the final Muse product already existed in the first prototype: Jollybot (that cheerful little mascot) was there, and an early version of Muse Charm also appeared in the first demonstration. Not all ideas and feature tags were there, but many were already in place.

From there—that was February—we launched the product seven months later. Those seven months were truly difficult: going from a prototype that sometimes worked, was full of magic but mostly couldn't function properly and constantly crashed, to a highly reliable, carefully polished consumer-grade product, something we were confident delivering to the consumer world. It was a long and quite painful process.

David: Before we go into that phase, I want to go back to one thing: you built the prototype in one to two weeks and showed it to the board—what role did Mark play throughout this process?

Alexandr: He experienced the same "ecstasy and fear" from this product as all of us. He's mentioned this experience in various interviews: he used it for various household matters, with his kids, and also for his MMA training—having it watch videos of his training and give him improvement suggestions.

During that period, a culture even formed at staff meetings—across a large portion of Meta's teams, people would share their amazing stories of using agents during meetings. That was in February.

Looking back at this whole cycle now, it's quite interesting, because the Claude phenomenon clearly had a peak and then faded. It didn't persist, and I think many people experienced this feeling: there were those magical moments, but most of the time it wasn't usable, problems kept emerging, and then it gradually faded.

So the core challenge in building this product was largely: how do you turn it into a truly reliable consumer experience that you're confident putting in front of ultimately billions of people, with the assurance that most of them will have an excellent experience?

David: Why were you able to do it when Claude's developers couldn't?

Alexandr: First, we have tremendous respect for Peter Sainberger and everyone who contributed to Claude. I think Peter is absolutely a visionary, and having something like Claude emerge from that beautiful mind of his is a kind of magic.

What makes Muse so good, perhaps unglamorously, is the sustained polishing and attention to detail required to make the combined model-plus-product experience truly excellent. We iterated very, very heavily on the model itself. We aligned our overall model roadmap toward the broader direction of building personal super intelligence, specifically the personal agent direction.

David: Tell me more about that, and how it differentiates you from other competitors.

Chapter 5: Why Meta Bet on Personal AI

Alexandr: It's actually interesting to go back to that moment. If you go back to early this year, February, Opus and Claude were far ahead of everyone in coding and growing explosively. That was the beginning of Anthropic's full rise, sweeping across the industry—a very alarming moment: oh god, their lead in coding is too big—plus they had Mythos, and the lead was very clear.

The mainstream view in the industry at the time was: coding agents are the only important product form, or the only important technical capability. There was enormous inertia across the industry driving everyone to pursue this use case and compete with Anthropic on coding agents.

But precisely because of our firsthand experience and the conviction we formed around the idea of personal agents, we chose to step back and think. I think the reason we started with personal super intelligence from the beginning is also because we believed Meta had a unique advantage in this direction.

David: Let's talk about that now.

Alexandr: The first thing I told Mark—even before the Scale and Meta partnership deal was finalized—was: if we have AGI or ASI, Meta is actually a very special company.

One of AI's promises is this: in each of our lives, there are things we "want to do" and things we "have to do." AGI's promise is that we'll barely need to spend time on "have to do" things anymore, because AI will handle a large portion of them for us, and we can spend all our time on "want to do" things. In my view, Meta's products represent exactly what we all "want to do": I use Instagram, WhatsApp, or Facebook, and these products let me follow my interests, hobbies, and the people I like and care about. Meta's product portfolio happens to cover the domain of "things we want to spend time on."

In the long run, Meta is an incredible winner in all of this. This was one of my earliest convictions about Meta and super intelligence, and these ideas ultimately fed into the concept of "personal super intelligence."

You can look at it from several angles: one is the company's DNA—Meta's heritage is connecting people through friendship and family, helping people discover their interests, all very human things. Another is the business perspective—Meta has 3.5 billion users using its apps daily, and the natural domain where Meta is destined to succeed is consumer AI, not work, not coding, but personal.

David: Right, exactly.

Alexandr: So, personal agents, personal AI, and consumer AI have always been the areas where we believe Meta can make extraordinary contributions.

So we arrive at early 2026: you have the prototype, internal support, everyone knows Mark has endorsed this, and then you begin seven months of difficult product refinement.

Yes. There was a shadow hanging over us: personal agents had "come and gone"—we always believed that was because the experience wasn't good enough, wasn't polished enough. But we knew this was something we had to take seriously in developing the product.

We focused one of the top priorities of our model development roadmap on building personal agent capabilities. We had a long spreadsheet listing over a hundred specific behaviors and capabilities the model had to have to support the ideal personal agent. The process was extremely meticulous. Over time, we established evaluation criteria for all these items, figuring out how to train the model to be very, very good at each one. We would also regularly review this spreadsheet, identify where we were falling short, and ensure continuous improvement. It was an extremely detailed process of constantly polishing every corner of the model to support the entire product.

David: Is this also part of Meta's culture? This culture of continuous product polishing?

Alexandr: Yes. A large part of Meta's culture is: once you know what to measure, there are very talented people to optimize those metrics, polish all the details, and push these things to the extreme. In the pre-AI consumer product era, this manifested in how to optimize conversion rates, recommendation rates, every step of the new user onboarding flow; in the AI era, at least starting with Muse, it largely manifests in polishing every detail of the model to perfectly fit product needs.

Chapter 6: Mark Zuckerberg's Obsession with Improving Products

David: Have you read Jeremy Stern's in-depth piece about Zuck on Colossus?

Alexandr: I saw it. Honestly, it was so long I didn't finish it.

David: It was 15,000 words. I just had Jeremy on the show—he's my favorite author.

Alexandr: I saw that episode. I think he's now the best writer in the world on entrepreneurs and investors, truly number one globally.

David: In that piece, Zuck talked about his advantages over Elon and Sam Altman. He said: my skill set is very boring and low-key, but I'm very good at building teams and then continuously, slowly improving products over a long period. Compared to Elon and Sam Altman, I won't lose control of the company, and I don't need to raise any more money, so I can just keep doing this for a very, very long time. When you were talking about polishing Muse for seven months, that piece kept surfacing in my mind.

Alexandr: I think that's completely right. It requires tremendous patience—seven months, maybe it doesn't sound that long, but in the AI era, seven months is equivalent to a decade in traditional industries. That's true for all of us. But I really have to give Mark a lot of credit—it requires great patience, restraint, and trust in the process.

You can go back and look at what Mark said on every earnings call throughout this period—from February all the way to September—he would mention that we were building personal agents. We weren't really hiding it; we were building it and talking about it, but internally the company was full of doubts throughout this period. The mainstream Wall Street narrative was: oh god, Meta is burning so much money, can they actually get anything done in AI?

David: More than that, it was: Zuck can't win AI. That phrase was repeated over and over for a long time. Going back to Jeremy Stern's piece—he started talking privately with Zuck's competitors, including a former researcher who had worked under Zuck and now went to a competitor's lab. That person said: "I don't want to compete with them, because Mark is like..." I think the line was "Mark is that damn terminator"—he just won't stop. This researcher also talked about one of Zuck's superpowers: he's never satisfied with the product's current state, always thinking about continuous improvement; but he's not the kind of jerk who destroys team chemistry—he is indeed a dictator, but not rude, doesn't hurt the team. The piece specifically mentioned that team cohesion is extremely important to him.

Alexandr: I think all of that is true. One of the most magical things about Muse is that we spent a considerable amount of time polishing all the details before finally launching the product. I think the personal agent product category is one balanced on a knife's edge: if it can reliably create these magical experiences, it has the potential to become one of the greatest consumer products ever. But you're dealing with models, with non-deterministic things, with agents that can fail and be unreliable—once users have an unreliable experience, they'll think it's garbage.

This is something that needs product quality and value to reach a certain threshold before it can truly resonate broadly.

Chapter 7: How Meta Determined Muse Was Ready to Launch

David: I want to talk more about this process. I'm curious how many resources were invested, how big the team was, how many people were involved. Do you have data, like: I use Muse, nine times it's a great experience, two times it's inconsistent or terrible, and then I churn?

Alexandr: You could see this happen with almost every other agent product earlier this year—almost everyone churned, and the reason was unreliability.

We did extensive user testing on Muse, organizing many rounds of testing where brand-new users downloaded and used it from scratch, observing what that process felt like. One of Muse's core premises—which I think is where it truly delivers—is that it's fundamentally different from most consumer AI products people have experienced: it's not a chatbot, it's an agent that can do what only agents can do. For developers, this might be old news—they've been using agents for a while; but for most people, just this leap requires explanation and very thoughtful onboarding.

We also did testing across different model versions and continuous A/B testing. The product experience is extremely sensitive to model quality—to the model's reliability in completing tasks, in communicating with users, in proactively confirming when guidance is needed.

David: But how did you know it was ready to launch?

Alexandr: Part of it was very quantitative: going back to that spreadsheet with over a hundred rows, we set "release thresholds" and "good standards" for each row. We had many checkpoints—80 rows green, but 20 still red—and we knew that wasn't good enough. It wasn't until we had the first model checkpoint where all rows were green—that was a specially trained version of Musepark 1.3.

That's the very quantitative part. There's also the experiential part: we tried the model and could feel it was dramatically better than any version we'd iterated on before.

David: Going back to the seven months you mentioned, how many people were working on this, model work and product work combined?

Alexandr: Fewer than 200 total. By big company standards, that's actually very few.

Looking back, I think that's quite remarkable: the entire executive team, everyone knew this could be one of Meta's most important products this year, but we had that organizational restraint—keeping a truly small and focused team, not letting it balloon into a project involving half the company, because we knew this was a singular work of art. It needed strong conviction and focus to coalesce into a complete and beautiful experience.

David: Tell me more about that. Why do you call it a work of art and say it needs a singular vision or conviction?

Chapter 8: Why Great Products Need a Singular Vision

Alexandr: I think there's a common criticism of products from large organizations, and this problem is now starting to spread in many AI labs: the product becomes a Frankenstein monster—almost like a Cronenberg-esque mishmash—pieced together from a bunch of different people's perspectives, visions, and beliefs. It's a bit like PM hell, where every product manager has one thing they want to stuff into the product, and ultimately it all gets mixed together into a smoothie of various things.

As a user and consumer, you can feel the difference—you can feel that a product was made by many different people, each with their own goals, haphazardly stitched together. Muse had to be the exact opposite: one clear conviction, one complete and unified experience that we present to the world.

David: I love that point. This is completely coincidental—I don't know when this episode will come out, but we're recording this on the 15th anniversary of Steve Jobs' passing. When you talk about consumer products with a singular vision, you naturally think of Steve first. I've read all the books about Steve and Apple's history, and I've done about 15 to 20 episodes about them on my other podcast Founders. My favorite line to describe Steve's influence is: Apple is Steve Jobs having ten thousand lives. The products embody his taste, his perspective, his convictions—what he wanted to see in the world is what end users got. In Muse's case, Nat Friedman deserves enormous credit—his taste and aesthetic sensibility ultimately determined and shaped the product's appearance.

David: Have you publicly discussed Muse's user data?

Alexandr: We've said "millions," and I've retweeted charts that others posted, which suggested Muse is the fastest-growing consumer AI app ever.

David: But you won't say specific numbers.

Alexandr: We said "millions."

David: Right, millions could be hundreds of millions, that's also millions. So you just didn't say the modifier in front.

Chapter 9: Using Memes to Make Meta Cool Again

David: I want to talk about your presence on X—and I say this with affection—you've been a bit "unhinged" recently. Since Muse launched, your public posting style has changed noticeably. Is this a marketing strategy? What's going on here?

Alexandr: Several things are happening at once.

David: It's certainly more fun than what you did at Scale.

Alexandr: Definitely more fun. The past few weeks have been the happiest period of my career, bar none. Several things are happening:

First, we felt it was important to find ways to make Muse break through. This is an era of constant bombardment with AI products, startups, new versions, new features—for most people it's just a flood of information. We held large team meetings specifically to discuss this: we need to break through—Meta has strong distribution capabilities and can ensure everyone knows about this, but that's not enough to truly get the world's attention in a fun, culturally infectious way.

As it happens, I know some friends in my life who are obsessed with internet memes and internet culture—I affectionately call them "my neurotic friends"—who think about memes and the internet all day long.

David: I know some people who are obsessed with Muse, and they're also obsessed with your X timeline. They're convinced you have a hundred Muses helping you think of memes.

Alexandr: Actually, every single tweet on my X timeline is written by me. For a while, I really struggled to concentrate in meetings because meme ideas kept popping into my head and I couldn't resist posting them. Once I was having dinner with colleagues, and halfway through I suddenly said, "Wait, I've got it!" and posted another meme. It felt a bit like being an engineer debugging back in the day—you can feel that intuition of "this one's going to blow up."

Second, I think it's important: how do we make Meta cool again? For some reason, despite having unparalleled cultural software like Instagram and Facebook, Meta hasn't been cool for quite a long time. So I wanted to make Meta cool, fun, a bit playful.

As for that "letting loose" feeling, it happened naturally over time. When Muse first launched, I was mainly sharing various interesting use cases, retweeting dozens of people's Muse scenarios. Then for a period, I posted a few tweets without much thought, and those turned out to perform really well.

David: Have you seen that pattern of "the longer you think, the worse it performs"?

Alexandr: Exactly. Some of my tweets I really didn't think much about—I posted and didn't even look at my phone, then hours later I'd check and think, "Wow, this one took off." Then I remembered something about the internet—not deliberately, but I recalled: the internet rewards risk, rewards the unexpected, rewards things people don't see coming. So I just let go.

Many of my colleagues were a bit surprised by this. Nat had a line that he said at several meetings throughout that whole meme-posting week: "I thought we would learn a lot about Muse users from this, learn how people use Muse, but actually we just learned a lot about Alex's brain."

That period was really fun, and I felt like I was drawing on years of accumulated internet knowledge. I think it actually had a real impact on growth—which is a bit funny, though we don't have precise tracking data, so it's hard to determine.

David: I definitely think it had an impact. I'm hearing about it everywhere, and the people I hear about it from have all downloaded the app and are now using it.

Continuing what you were saying—you have distribution capability, but that doesn't mean you can make people actually come back and make the product truly sticky. So how did you do that?

Alexandr: I think the key is: how do you get users to reach the "wow, you can do this" moment as early as possible? This product is naturally suited to that, because if someone has never used an AI agent before, many of the things it can do will blow their mind.

Another key is: how do you capture those stories of people having "wow" moments and use them to spread the product more widely.

David: But how did you use those stories? You weren't running ads either, right?

Alexandr: We ran some ads. But honestly, at first it was just me retweeting interesting cases—whenever I saw someone doing something interesting with Muse, I'd quote-tweet it.

I've never understood why Anthropic and OpenAI's ads are so bad. I remember Anthropic did massive outdoor advertising, all about themselves; OpenAI's was the same, with just an icon in the ad (I think most people wouldn't recognize it as their logo), then "ChatGPT" written on it, with two people sitting at a table. I thought, what does this even mean? You have so many use cases, the ad just needs to show the product's benefits, it's that simple.

David: You have millions of ready-made materials—just use them to run ads. It's really interesting how such basic mistakes get repeated over and over.

Alexandr: One question we kept coming back to when building Muse and doing other things at Meta is: why would anyone care? They don't care about you, they don't care about your company. Everyone acts out of self-interest.

There's a very old book called Scientific Advertising, written in the early twentieth century by Hopkins, a top copywriter and one of the founders of the most profitable advertising agencies in the twenties and thirties. He said in the book: this isn't just about advertising, it's about human nature that doesn't change. His core point was: your ad talks about yourself, about your product, about "I have this thing, come buy it"—but people aren't interested in that, they're all self-interested. You just need to do one thing: show what this thing can do for them, and start with that. Those who absorbed these ideas written in the early twentieth century built massive companies and advertising empires in the fifties, sixties, seventies, and even today.

Alexandr: Right, exactly. A few key insights: first, most people in the world don't care about OpenAI or some research lab, don't care about these companies or our research—they care about what this can do for them. We also had a gift, which was that mascot—that little joyful character.

I remember the tweet that made me realize "this has potential": I was driving to work, used Muse to generate an image of the little guy carrying a briefcase, and tweeted: "MFW (My Face When) going to work on a Monday." I immediately felt it was one of my best tweets recently, and then I realized I could put him in all kinds of situations.

I also realized how to create surprise moments completely different from any other product marketing you've seen. I don't remember the last time someone used bold memes to market a product; I'm not even sure anyone has done it. But I felt: wow, this is something we can do with this little guy, this is fun, this will spark curiosity in ordinary people.

Chapter 10: Building Trust with AI Agents

David: Okay, but that might get them to learn about the product, maybe even download the app. But once the app is on their phone, how do you keep them using it?

Alexandr: Users who truly fall in love with the app go through a cycle—we observed this pattern as early as February: you give it a small task, it does it, and you think, "Wow, it did it." Then you give it a slightly bigger task, it does that too, and you think, "Wow, it can do that, and this too." Then you give it a slightly more complex problem. It's a bit like a trust fall game with an AI agent—at first you don't quite trust it, but you're curious, so you give it a small try, then slightly bigger, then bigger.

When I talk to college students, some already use AI in a way that's a complete trust fall: they just record voice messages, pour all their thoughts into the AI, and trust it to organize everything, give them a clear plan and action items. I think the key for successful Muse users is getting them into this trust fall process, using it for one thing, then a bigger thing, then bigger, and succeeding each time.

David: I'm curious how much of this is word-of-mouth driven. Obviously, after talking to you I'm going to download this app no matter what. But what really got me interested was a screenshot a friend sent me showing what he was doing with Muse. He left his ID on a FlexJet plane, and he told Muse: I don't want to be involved in this process, you figure out how to get this ID back, my ID is there, I'm here, figure out all the steps in between. Then Muse arranged a courier, got the ID, got through his security system, and delivered it to his office. That's the best marketing: just three or four screenshots, back and forth, him saying "I don't want to be involved, you handle everything," and Muse getting it done step by step.

Alexandr: Yes. Modern product marketing is largely screenshot-driven. A modern consumer product must be able to speak through screenshots—screenshots must be able to spread, whether through group chats, word of mouth, or social media. I think this has always been an important driver. This wasn't deliberately designed, but I think the little guy's presence in screenshots is a big deal—having Jolly or your Muse avatar appear in a screenshot immediately conveys "this is a Muse screenshot," and that's the best marketing you can have.

Chapter 11: The Future of Human Ambition

David: So do you use Muse all day yourself? What are your personal usage habits?

Alexandr: I mainly set up a lot of workflows and use it between meetings. At work I'm supposed to listen carefully and engage with others, so in practice I set up a bunch of workflows and use it as a "second brain."

David: You talked about this in the article you wrote. It's very short, and the last time I checked it had 5.5 million views. You described Muse very well—"if everyone had a second mind that transcends themselves." I think this is philosophically one of the most interesting questions about "what it means to be human when you have powerful AI"—what that relationship actually looks like, different people imagine it differently. The most worrying scenario is of course AI dominating us while we passively follow. But before you go on, what's your own view?

Alexandr: I truly believe in Muse, in personal agents as a long-term product form. Humans are born with desires, wishes, and dreams—that's human nature. I think AI and personal agents like Muse will become the mechanism and bridge that lets us realize these dreams, and keep realizing them. Then humans will dream bigger, bigger, bigger. Children will dream of having their own galaxy, and then Muse helps them realize those dreams.

Bezos wrote in a shareholder letter something like: the best quality of consumers is that they are never satisfied with their current choices—"divine discontent." I think this is a model for our relationship with AI. There are other models—I think AI will increasingly participate in the process of scientific discovery, becoming more deeply embedded in every part of the economy. But I do believe this form will persist: everyone having an AI that supports them in pursuing what they want.

David: So you see AI as something that removes "the burdens of life," right? The things we don't want to do. You said in that article that the world is full of gatekeepers and obstacles, and Muse can help you bypass these, no longer consuming your time, letting you truly focus on what you want to do.

Alexandr: I think Muse can give everyone the adulthood they dreamed of as children. Every child has grand ideas and dreams. When they imagine their lives, they imagine becoming an astronaut, saving the planet, changing the world—they think big.

But for various reasons, after people finish school, enter the workforce, and work in a position for a while, those hopes, ambitions, and the ability to dream get slowly drained away.

David: Not drained—shattered. Life beats them out of people. As adults, I think we all underestimate how much we're like the walking dead. We lose our agency, we lose—I think "ambition" is a great word. AI's promise broadly, and the promise through products like Muse, is to let this state continue from childhood: dream big as a kid, use AI to help realize it, then dream bigger, realize again, dream bigger again, experiencing a rising sense of agency throughout life—rather than agency flourishing and then being crushed.

Alexandr: What you just said reminds me—you mentioned earlier that unsettling AI psychological review you and Nat went through, with waves of deep psychological probing. What if it were the opposite? Not that, but actually giving you more confidence—making you understand that the world is malleable, and that with enough energy and drive, you really can change the world around you.

I think this is the trait the greatest entrepreneurs display: they dream big, achieve it, dream bigger, maybe work in one direction for ten years, achieve it, dream bigger. Elon is obviously a prime example, and Mark too. Many iconic entrepreneurs have had this firsthand experience. I think this should become the norm for everyone.

The tragic part is that for most people: you have bigger dreams and ideas, you enter the corporate workforce, slowly become the walking dead, and at some point you want to pursue your dreams again, but it's really too hard, because you have so many commitments—maybe a family, maybe various ties—and you're almost stuck in that "walking dead" state, unable to pull yourself out.

There's one more thing about Muse, before we get to Scale AI and your decision to partner with Meta. You wrote in that article: "The world so far has been shaped by a small number of fanatics who somehow made their will reality, but we've never seen what humanity looks like when every individual's agency is fully switched on."

I think the world I'm describing—where every individual's agency strengthens throughout their life, where everyone has the ability to realize their wishes and dreams—will go crazy in a very beautiful way. It will be interesting, artistic, and cool, reaching into different corners of the world, incredibly diverse, almost unimaginable: billions of people suddenly having the ability to create extraordinary things because of abundant intelligence resources. I think that's part of the promise. In my mind, it's a bit like the interdimensional cable TV in Rick and Morty, manifesting in the human world in some way—that would be truly remarkable.

David: I keep hearing a line from that show echoing in my head: the universe eats smart people.

Chapter 12: The Mark Zuckerberg Call That Changed Everything

David: Okay, I want to trace back. What I'm personally curious about is: you founded Scale AI, you were running Scale AI, and then one day Mark Zuckerberg contacted you and said—I assume he said—"Hey, can you hop on a call?" Can you tell me how that happened?

Alexandr: I think the message was something like: "Hey, do you have time for a call?"

David: Did you have a prior relationship with Mark? Had you spent time together?

Alexandr: I have a friend named Alex Schultz, who I've actually known since Scale was about one year old, from San Francisco. He's now Meta's Chief Data Officer, previously a longtime Meta executive. He introduced me to Mark around 2021 or 2022, when Mark was becoming increasingly interested in AI and Scale was doing a lot of work in AI.

I remember it took about a full year from the introduction to the first meeting being scheduled. Why so long? Mark's calendar is obviously extremely busy, he has countless things to handle, and he's very good at prioritizing—he'll spend a lot of time on things that truly matter, sometimes even too much time. The only way he can do this while also being a good father and partner is by ruthlessly prioritizing.

We met a few more times after that, and I would seek his advice—getting advice from Mark Zuckerberg is of course an amazing thing. I remember at that time, I would ask him questions on WhatsApp and he would reply with very long messages. I was confused: how does he have time to write such long replies? Later I understood—he just writes extremely fast, and can quickly write rigorous and detailed replies to a large volume of WhatsApp messages. That's one of his impressive abilities.

Meta is an important Scale customer. Before this event, we talked about once every six months or so—that was our frequency of interaction. Then came that call. At first it wasn't clear where it would end up, because the first conversation was just—it was after Llama 4—essentially asking: what do you think we should do?

David: Why is it important to emphasize that this conversation happened after Llama 4?

Alexandr: Llama 4 wasn't on the trajectory Meta wanted. That was genuinely a setback internally, and it happened at a time when AI was clearly becoming increasingly important—AI would obviously be crucial to Meta's future.

The first few calls were phone conversations, and he was somewhat asking for opinions: what do you think we should do? Where do you think we should focus?

David: This matches what I've heard privately about him—he has this amazing network of world-class entrepreneur advisors, not exactly Meta's board, though some board members are involved. I've heard he constantly seeks advice from these people, sending messages like: "Here's the situation I'm facing, what would you do?" Over and over. I think this is a remarkable quality, because it requires humility—continuing to do this after all his incredible achievements.

Alexandr: From his first contact to our announcement of the partnership, it was probably only five or six weeks—quite fast.

David: This is also something world-class entrepreneurs have in common. We had Jonathan Ross, the founder of Grok, who first publicly talked about the $20 billion Nvidia partnership on my show, and he said: "From Jensen's first call to the money hitting the account, it took three weeks."

Alexandr: That's impressive. I had a lot of thoughts about Meta's AI strategy at the time. The earliest one was what I mentioned before: Meta is genetically a company focused on what people "want to do," and in this great AI transformation, it's actually one of the most uniquely positioned companies. I conveyed a lot of ideas to Mark, and at the same time, there was another parallel conversation: "Maybe we can explore whether there's a possibility of partnership."

David: So when he said that, was it completely unexpected for you?

Alexandr: Completely unexpected. The whole thing was like a series of crazy events. At the time I thought: "Oh, that's nice, but impossible..."

David: When they say "we need to find a way to work together," you know what that means, right? Acquisition. They always say that—"we need to find a way to work together."

Alexandr: I probably didn't have that awareness at the time. I thought: "Maybe he's just testing the waters, just saying it, but does this really make logical sense?"

David: Why did you think it didn't make sense at the time?

Alexandr: Meta's signature acquisitions are Instagram and WhatsApp, both very clearly product-logic-driven decisions. Scale is a company with deep enterprise and government sales—a completely different business model from Meta. So the logic I thought of was: if Meta wanted to enter those areas, maybe it would make sense, but this wasn't that situation—Mark was clearly mainly interested in getting the Llama project and the overall AI project on track. He wanted talent.

David: Yes, he wanted talent. The whole thing was shrouded in a certain ambiguity and uncertainty.

Alexandr: When did it become clear? When we started talking numbers, I thought: "Oh, okay, so it's real." That's when things became clear.

Chapter 13: Leaving Scale AI, Joining Meta

David: At first you were still thinking "does this make sense," then you kept talking—how quickly did he convince you this was a good path?

Alexandr: Part of it was that the deal structure itself was fascinating. What we ultimately landed on was: Meta invests and holds a 49% stake in Scale, Scale continues to operate independently, and I and a few people join Meta.

I was a bit incredulous about this plan at the time, because it was a very unusual deal, but I think it was truly a "three-win" design: Scale's shareholders got a good outcome, Scale continues to exist—and I truly believe Scale's best days are still ahead.

David: I'm curious—could he change your mind in a single call? Or was it more like "Mark, I need a few days to think about this, it's too crazy"?

Alexandr: It was several weeks of thinking: "Is this real? Does this make sense? And if it makes sense, how do I feel about it?" The dominant emotion was: this is crazy. Then as I started seriously considering whether to make this decision, that itself was no small matter. There were many important conversations to have with colleagues at Scale and our investors.

Giving up or handing over your "child" is of course extremely difficult. How many years did you work at Scale?

From founding to... nine years. And you're still very young, not even 30—that's about a third of your life.

Paul Graham has long said: if you treat your company as your life's work, you'll behave completely differently. I genuinely treated Scale as my life's work, throughout my entire time there. It was a very difficult emotional process.

But what ultimately made me see it clearly was a combination of factors: this really was a good outcome for all parties, and I also saw the potential of what could happen at Meta. At that time, Meta did look a bit like a "mess" in AI, maybe not the most attractive place to work in AI—but that actually triggered my entrepreneurial instinct: you could essentially rebuild it, like founding the lab all over again.

David: Yes. Precisely because there was so much work to be done, it cleared space for you to build many things from scratch, establish the right principles, build the right culture, and create the conditions for building something extraordinary.

Chapter 14: Rebuilding Meta's AI Lab from Scratch

David: Did you discuss it with Mark like that? Like "we need to hit the reset button here"?

Alexandr: He didn't seem interested in minor fixes. His attitude was: "This path doesn't work, tear it down and start over." That's exactly how it was. The whole conversation evolved continuously. I very firmly believe one thing—Elon has also proven this at different stages of his career—which is: you only need a very small, extremely strong, highly technical, very flat team. Do that, and you can achieve a great deal very, very quickly.

This is something I'm very confident about, especially in AI—this is the right approach, particularly when we're facing the need to move fast. This was one of the things I discussed extensively with Mark early on, and I think he was excited about it too. It ultimately became one of the core pillars of our overall strategy: small, flat, highly technical teams, extremely high talent density, and how fast you can move under these conditions.

David: From the outside, Mark's approach looks like "going all in"—that's how I understand him. Have you read the book The Mind of Napoleon?

Alexandr: No.

David: It was published in 1957. I found it through an interview—I think it was Tyler Cowen interviewing Sam Altman, asking what the most important book he read that year was, and he said The Mind of Napoleon. The book is hard to find, 300 pages of Napoleon's own words, organized by theme. In it, he repeatedly emphasizes: hesitation is fatal. But he also says that before taking action, you do a great deal of deliberation, examining from every angle to make sure you're making the right decision. I feel there's an echo of this in how Mark operates: do a lot of deliberation, make sure you're on the right path, but once you've made the decision, go all in without hesitation. I feel that echo, right at this moment—he came to you, and then you started rebuilding the entire organization.

Alexandr: I think he's really internalized the belief that "failure doesn't matter; what matters is how big you win when you win." This underlying logic allows him to truly act fearlessly in many situations, because he's very comfortable taking massive risks.

David: He has a great line: in a constantly changing world, the biggest risk is doing nothing.

Alexandr: Yes, he's deeply internalized that. Of course there's a parallel universe where Meta looks more like Google—slower, more bureaucratic, a very different-looking company. But precisely because he's internalized these lessons so deeply, the entire organism stays dynamic, stays vital, stays alive.

Chapter 15: From Founder to Coach

David: You were the founder of Scale, running the company yourself; now you're rebuilding this lab and working with Mark—how does your approach differ in these two environments?

Alexandr: At Scale, I deeply internalized an idea—though I can't remember where this advice came from—which is: if you want to be a good founder, you must be able to do every job well. I had very strong opinions about how every little thing in the company should work. This created a company that was largely a product of my personal effort and perspective, with this style permeating every aspect of the company—but it also made me a bottleneck in many important areas. This has pros and cons for a company.

I think if you're a first-time founder, you should absolutely adopt this approach, because it has a much higher probability of success than not doing it. Scale was my first company.

David: Right.

Alexandr: But at Meta, the situation is completely different, because we have to accomplish an enormous amount in an extremely short time. I don't have the luxury of "being good at every job." I'm not an AI researcher, not a product designer, and don't have the various specialized capabilities that are crucial to organizational success.

So at Meta, I adopted a completely different philosophy—the core being: how do we create an environment that attracts top talent in every field? Recruiting the best and most exceptional people is the first priority.

David: Why do you feel you can do this at Meta? Is it just because there are more resources?

Alexandr: I think what made MSL so attractive early on was that it offered the opportunity to build a lab from scratch with a very small team, and we had a clear direction—personal super intelligence. This direction resonated with many people. I think everyone working in AI wants it to genuinely help ordinary people, wants it to mean something to their mom, grandmother, grandfather, wants it to be practically meaningful to everyone in the world.

At the same time, because we're doing this with Meta's resources, we're not limited by compute, not limited by infrastructure, not limited by distribution channels—these are precisely the most common bottlenecks for startups. But this is also a genuine opportunity to leave your mark, exert deep influence, and shape things with your own hands. I think that's quite exciting for many people.

David: So this is why Mark is willing to spend what seem like astronomical sums to recruit a few top people—because he knows these people will in turn attract more top talent. But you need that starting point.

Alexandr: Exactly right. Another point that may be underestimated: OpenAI and Anthropic's stock prices have risen enormously, so if you consider how much these people could earn by staying there, it's actually quite a substantial number. For many talents, compensation might ultimately be roughly comparable. Of course, we often overlook how much the early employees and researchers who joined these labs have seen their stock appreciate, and what their compensation levels are now.

Our conviction in founding MSL was just two things: first, recruit the smartest, most exceptional people; second, personal super intelligence—how to create an environment at Meta and MSL where everyone can do their best work.

At Meta, I do something more like traditional management. The core question I'm always thinking about is: how do I create an environment where these brilliant, truly special people can fully realize their talents? I play more of a coaching role, whereas at Scale it was completely different—I was the "author," the one who needed to implement my perspective into everything we did.

David: Tell me more about that "coaching" feeling.

Alexandr: I think my job is, first, to help establish the North Star and the direction—personal super intelligence, winning the consumer AI race, building AI that's meaningful to everyone; and second, to discover special talent within the team, then give them the environment and resources to fully realize their abilities.

A mistake big companies often make is: great talent is right there, but they don't get the opportunity to stretch their wings, so they often choose to leave for smaller companies or startups.

David: You mentioned Jeff Bezos's shareholder letter earlier. This is also a point he repeatedly emphasized, in both the books I've read about him and his own shareholder letters—if great people can't build things, they leave.

Alexandr: Exactly. AI is a very special thing. The word "research" seems to have lost its original meaning now, but by definition, humanity doesn't yet know where these models' capabilities end, or what can be built with them. They are truly unknown entities we don't fully understand. So this really is science, really is research—we're exploring the limits of this technology, exploring what can be built with it. This process requires a lot of trial and error, and requires smart people to generate extraordinary insights.

This isn't a process that can be pushed forward with extremely high pressure and extremely intensive work methods. People need space, time, and resources to do their best work. I think about this in many dimensions. Take Nat Friedman—he's absolutely brilliant. As I said before, if there's one person whose core vision runs through Muse, it's him. What matters now is putting him in a position where he can fully express all of that. We also have some excellent researchers with very exciting ideas—how do we create an environment where they can explore freely? Most attempts won't succeed, but some will achieve amazing results, and those successes are enough to justify all the investment.

There are two kinds of problems, or two modes: one is called diamond mining, the other is called skyscraper building, depending on the return distribution of what you're doing. In some domains the distribution is a super power law—most of what you do is nearly worthless, but a few things can be worth a million times more than everything else combined. Venture capital is a typical diamond mining industry—it's all about identifying those special ideas and seeing them through.

That's easier said than done—just look at your 19th birthday party.

Chapter 16: Future Billionaires at Alexandr's 19th Birthday Party

David: Ha, that's really incredible. Some listeners don't know what I was just talking about—can you briefly say who was at your 19th birthday party, and why investing in that table would have been a good trade?

Alexandr: It really was a crazy thing. At my 19th birthday party, I was a freshman at MIT. It was January, and I was doing a winter internship at a trading firm called HRT. There were about ten or twelve of us, almost all Harvard or MIT students, most of whom already knew each other from various math, science, or computer science competitions.

Among them was Jeffrey Yan, who later founded Hyperliquid—very successful; Scott Woo later founded Cognition—also very successful; another person was Jesse Zang, who founded Decagon—also quite successful; another was Vicky Yei, who I think is an excellent researcher now working at Anthropic. Everyone was extremely outstanding.

I recently talked with HRT's founder—HRT itself has now grown incredibly impressive.

David: What does HRT do?

Alexandr: HRT is a proprietary trading firm, trading with its own money. In Q2 2020, they set a record of $11 billion in revenue and about $8 billion in profit. The company has fewer than a thousand employees—truly astonishing. I recently met with HRT's founder, and they said they were considering making some private investments. I said: "Just invest in the interns, write seed checks for all the interns."

I grew up participating in math, science, and computer science competitions and Olympiads, and met all kinds of people from across the country in that circle. Seeing the people I competed with back then become so successful now is really a surreal feeling.

Chapter 17: Diamond Mining vs. Skyscraper Building

David: It'll be fascinating to see what happens with your group over the coming decades. I interrupted you earlier—anything more you want to add about diamond mining and skyscraper building?

Alexandr: The core is: is the return power-law distributed—a few things worth a million times the investment—or is it more like building a skyscraper? I would classify data labeling or data businesses as the latter. Amazon Prime delivery is a similar example: returns are relatively linear, and you have to build a process and optimize it to the extreme.

If you're building a skyscraper, it's a highly operational business, and it's all about squeezing every bit of efficiency out of the process—finding ways to improve it by another 2%, another 1%, another 0.5%, continuously optimizing.

Diamond mining is completely different—it's like saying: "Alright, everyone go explore, try all the crazy ideas," and then you just nurture the ones that look promising.

David: Is that what it feels like at Meta now?

Alexandr: Yes, that's what it feels like at Meta. And skyscraper building, overall, is what it felt like at Scale. This is a very fundamental paradigm difference.

Chapter 18: Why Alexandr Has Over 200 Direct Reports

David: Back to the management differences—your management approach at Meta and Scale is fundamentally different. Why do you have so many direct reports? It seems like over 100? What's the exact number?

Alexandr: Over 200.

David: How? Why?

Alexandr: Going back to MSL's founding philosophy—what kind of lab did we want to build? The core idea was: how do we maintain extremely high talent density? How do we let these excellent people do their best work here? How do we build a technology-centric organization where technical people can truly realize their potential, even accomplish their life's work?

One of the design principles around this is: extreme flatness. In practice, all researchers who join the TBD team report directly to me. This is more of a statement—TBD is a symbol of anti-bureaucracy. We recruit brilliant people who have clearly proven over their careers that they can do exceptional work. I don't need to manage them in detail; they're fully capable of doing excellent work autonomously.

But we do need technical leadership. Some people are responsible for setting clear technical direction for their respective groups. We have a "Pod" structure—each Pod has a technical lead with final say on technical decisions. So technical leadership exists, but from a "people management" or bureaucratic perspective, we are anti-bureaucracy.

On the other hand, we do make a lot of collective decisions. This is an extremely talent-dense team, everyone has brilliant insights, and we want to find the path forward that everyone collectively agrees is most correct through debate and discussion. Everyone in the lab has done exceptional things—we want to hear everyone's opinion.

This is of course an unconventional management approach. If I claimed to be the best manager for over 200 people, that would obviously be a lie. But the point is—they don't need a manager at all; what they need is a good environment.

They're all extremely smart, all have super abilities, and just need the space to apply them.

David: So maybe not "coach." You're more like a guardian of a good environment, letting them do great work.

Alexandr: Yes, exactly.

David: Alex, this has been really wonderful today. Thank you for taking the time.

Alexandr: Thank you.