news

Recently, Zuckerberg shared his experience in AI, AR, entrepreneurship, and team management in a conversation at the SPC hackathon. He said that the future development of AI will go beyond existing data and enter the field of more complex intelligent behavior, which requires the creation of new data sets.

Zuckerberg believes that AR glasses and VR helmets will become the core of future computing platforms. These devices will become popular in the next 10 to 15 years and will completely change the way people interact digitally.

Regarding entrepreneurship, he shared how he dealt with the emotional ups and downs during the entrepreneurial process, especially as the company continued to expand. Working with trustworthy people is one of the things he values ​​most at this stage.

He also talked about how as the company grew, decision cycles became longer and challenges became more complex, but the support of the team and deep personal relationships helped him stay motivated, and that working with great people not only led to great projects, but also made the work fun and meaningful.

On a personal level, Zuckerberg managed stress and stayed focused during the pandemic by participating in MMA and surfing. He believes that sports can provide an immediate reward mechanism, in contrast to the longer project cycles at work.

The following is the main content of this conversation, enjoy~

host

Mark, your hairstyle is causing a stir. Why did you choose this new look?

Mark Zuckerberg

I hurt my knee, and because I was training and fighting, I cut my hair short, and I'm probably going to keep it that way for nine months.

So by the way, my knee finally healed this week. I'm ready for some violence. But I might keep this haircut. I don't know. It's kind of like when I was younger.

Priscilla or someone told me, hey, you should grow your hair out. And I thought, there's nothing else to do for the next nine months. It looks good.

host

Yes, it does look good. By the way, you have been a major player in the technology space for the past 20 years. How would you categorize the various phases you have gone through? Which phase do you think we are in now?

Mark Zuckerberg

I don't know. I'm trying to not be the protagonist. You know, I think that's worked for me, to be more of a foil. We go through a lot of different phases, right? The initial phase of making Facebook work and getting through the bad times.

With all the politics, volatility, and pressures of accountability over the past decade, this is a very different phase for the company. It’s not over yet, of course.

But I do feel like we now understand these different dimensions and have done a great job of accomplishing what we need to do. I feel more confident about that.

It also allows us to move more towards offense and do more aggressive and exciting things. When I think about the next phase, there are so many ways to do amazing things.

The good thing about social media is that it gives everyone a voice and enables them to take action. A lot of people are using it.

I don't actually think of it as very groundbreaking. Basically, you're giving people the ability to do something very basic and fundamental that just requires really great execution.

When applied at scale, its implications are profound. But scale is what makes it interesting. It's really transformational in a way.

Like you said, in terms of the next phase of my career and what I want to do in the next 10-15 years, I'm more inclined to focus on breakthrough things that can amaze and inspire people at any scale. So obviously, the next generation computing platform projects we're working on, including AR glasses and VR headsets, etc.

I do think we'll eventually get to a point where billions of people will have these devices, just like billions of people wear glasses today.

So I mean, basically everyone who has a phone that's not a smartphone, they upgraded to a smartphone pretty easily, and then everyone else a little bit slower.

Here, everyone who wears glasses will have smart glasses, and maybe other people will choose different form factors over time. But either way, the journey will probably take 10 to 15 years.

This time frame is normal for building something so foundational. But the experience when you try on AR glasses for the first time is very exhilarating. We now have the first working prototype. It's an exciting time.

We had actually been working on it for so long that I had hoped it would be our first consumer product, but actually, I wanted the first consumer version to be even better. So we decided to make the first version a prototype. But when people tried it on, they thought it was amazing.

It's like nothing you've ever experienced before, like being able to interact with holograms in the world, like they're really there. And they appear in the form of regular glasses, which is a very magical experience. For me, a lot of it is just like that. And then there's the AI ​​part, which we'll spend more time on.

But for me, it's more about doing something amazing. I have a saying that I like to tell people, which is there's a difference between doing good things and doing amazing things. It's an inspirational difference. Doing good things is important, too. You can be amazing and not be good.

They're two different directions in a way. But I don't know, now is a phase of trying to do something amazing, trying to build something amazing.

host

It's really interesting. One of the guiding principles we use at SPC is that you are here to throw away the good ideas and pursue the great ideas. I would update that to say you are here to throw away the good ideas and pursue the amazing ideas.

Really, because I think life is too short. Our most scarce resource is time. We can have a lot of good ideas, but what we really pursue should be those amazing ideas. When it comes to the 10-year trajectory, we are LLM Something that is often easily overlooked in the release process, especially in the case of Llama, is that the fair exploration began almost 10 years ago.

In fact, I remember one of the last projects I worked on with you at Facebook in 2010 was to use machine learning to weight each node in our graph. I think that project turned out pretty well.

We released the Segment Anything model, which is different from Llama, but even in the first week or two of the release, you saw people running it through videos and the model got better. So it was good for everyone.

I think there is a big bias about the future of computing platforms, that because the iPhone won, the closed source model will continue to dominate. I'm not sure, the future is not set.

If you look back to the PC era, Windows was the leading platform, and there have always been open source and closed source models, each with its own advantages and disadvantages. But a lot of things depend on who is doing it. This applies to startups as much as to lazy big companies.

host

Whether it is an open source platform or a closed source platform, it is a question worth paying attention to which one will eventually become the dominant one in the future field.

Mark Zuckerberg

Whether it is AI orMetaverse, it's all like this.

host

I'm going to skip ahead to share a story that's very important to me. This story is probably a defining moment in my life. It was in 2005, almost when I joined Facebook, maybe week two or three. I was still trying to figure out what to do, and no one was really giving me direction. And I remember Mark walked up to me.

Mark Zuckerberg

There was no clear management structure at the time.

host

Yes, the management structure is not clear and it is total chaos and control.

Mark Zuckerberg

I remember walking up to you and saying, hey, man.

host

You should write a search engine for Facebook.

Mark Zuckerberg

Yes, I told you.

host

Right, and then I said, what team should we work with?

Mark Zuckerberg

yes.

host

You said, this is really important because the first thing people do when they come to Facebook is search for people, so don’t screw this up. My first reaction was, Mark, I don’t know how to write a search engine. I’ve never done this.

Why don't we just go get someone from Google or Yahoo to do it? And he looked at me and said, man, if I can build Facebook, you can write a damn search engine. And you did.

Mark Zuckerberg

Yes, that's true of many things.

host

That culture still defines Facebook today, and to me, it's still that hacker mentality and total accountability culture. If I could give one piece of advice to every startup in the world, it would be this: You can do anything you want to do if you focus on it. It's actually a lot easier than you think. That's the problem.

Mark Zuckerberg

Sometimes it may take longer than you think, but you can do it. A lot of things that seem obvious are not actually done by other people.

It's one of the strangest things, but it's also comforting because a lot of times we think other people have these things figured out, but they don't.

host

We're adults now. Back to Llama and open sourcing Llama. I'm curious, what practical applications do you hope to see with Llama? Do you have a specific vision?

Mark Zuckerberg

There are actually a lot of cool things that are going to be developed, but I don't have a specific vision. From my perspective, I just want everyone to use it.

Because, the more people use it, the better Llama becomes. Maybe this is a very selfish answer, but it makes me better at building what I want to build.

Frankly, people should be relieved by this answer because a big question about open source strategy is: Why are you doing this? You trained this model and then gave it away for free, is that sustainable?

For example, when training Llama4 or Llama5 costs billions of dollars, will you still give it away for free? My answer is: yes. I don't think of it as "giving it away for free", you are all helping me improve it.

So I don't really care what you do, as long as it's responsible. Of course, safety is important. The discussion about AI ethics, safety, and security is very important. We invest a lot of time in the training process, which is one of the main reasons why open source is difficult.

If we were just developing the model for ourselves, we could make certain assumptions, like we would have multiple layers of stacking, in addition to the model itself, to filter out bad stuff, like bad queries or bad outputs being generated.

We tried to replicate those capabilities in the open source model and built the whole system, which is deployed as a whole system, but fundamentally we have to put more effort into training the base model because it might be just one model without all those extra layers.

Maybe this is a negative answer to your question, but actually one of the benefits of open source platforms is that people will build things that you can't even imagine. So I don't know what you should let people build, you figure it out. But I am a little worried that the opportunities and challenges around security will be greater.

The discussion about open source security often revolves around whether open source is fundamentally less secure than closed source. If you do it well, open source software can be more secure, just as open source software has always gotten more secure over time.

Because it's open source, people can review it more. Although it seems counterintuitive, people actually fix bugs faster. Llama will be the same, people will point out problems, and we will continue to release small updates and also large updates.

Developers will continue to want to push out the latest model and apply their tweaks and fixes on top of it, so the entire stack, from silicon to data center, will get better.

We have done similar things before, such as the open source computing project, designing our own data centers and servers, and then open sourcing the designs, and eventually the entire industry standardized around our designs, the supply chain developed, and the price came down. This is good for us. We are now in a similar stage in AI and related fields.

host

I'm curious, given your different roles in these areas, what do you think about the scaling prospects of these models? I mean, they still follow the scaling law, how much more do you think we have to go? For example, Llama4, or GPT5, etc. are very interesting, but what are your current views on the future development?

host

What are the limiting factors?

Mark Zuckerberg

Yeah, it's hard to say. There's a question about computing power. I don't know the exact limit of scaling, but obviously I'm betting that it will scale, right? Because we're doing all this infrastructure.

It's hard to guess the exact timeline, these are things that take 5-7 years to build to get to the massive scale that you want Llama 10 to be. But you have to prepare in advance because you can train the model in 6-9 months, but getting the required computing power is a long process.

So it will work, but a lot of it depends on the data. A lot of it will be done in different areas, right? Areas that didn't have data on the web before.

What happens now is that companies go out and search all the content on the web and try to find patterns in it, and that's your pre-trained model. And then you fine-tune it to build different applications or features that you want. When you start getting closer to models that behave more intelligently, there are no datasets like that yet.

You now need to create this dataset yourself. Many new jobs in the future will not just take existing data and then build the computing power to process and train it.

I actually think that training in the future will be more like reasoning today, either trying to experiment in some area, like how do you create an intelligent dataset? There may be a certain amount of manual work involved, or it may be letting the system play by itself and generate data through games.

That's just one example, but it's something that doesn't exist today. So there's the question of how far this can go? My guess is very far. Yes, I'm an optimist, but it's hard to say for sure.

host

Do you want to talk about some of the retrospectives on Facebook in the media? By the way, speaking of the early days of Facebook, I remember when we were working on News Feed, I was part of multiple News Feed projects.

Mark Zuckerberg

You were one of six people who helped build the first News Feed.

host

I was sitting next to you at the time.

Mark Zuckerberg

That's a good thing, but you know,

host

There were a lot of war rooms and closed-source meetings in the early days of Facebook, especially after we launched News Feed, where I think I didn't leave the office for a whole week.

Mark Zuckerberg

Yes, there were protesters in the streets. We were already under a lot of pressure. You could walk out the front door and people were really angry about this.

host

Right. I'm curious, can you share a recent war room experience you had? For example, when the weights of the Llama model were leaked, what does the Facebook war room look like today?

Mark Zuckerberg

I think there are two different parts to what you're talking about. First of all, I'm not as involved in the day-to-day operations of the situation room as I was in the early days of handling the news feed.

The way we approach these large-scale strategic issues now is different. I've been trying to learn how to run the company more smoothly, and obviously I haven't been completely successful in that regard.

If you look at the volatility of the company, you're like, "Ah, that's hard to do." But I'm going to keep working on it and hopefully find a way to make it smoother without compromising on the really cool things that are out there.

The obvious way to make things more stable is to push less, but I can't do that. However, I do think we used to be too impatient. And I think sometimes we made up for our lack of experience with more effort.

But over time, you try to get things done in a more stable way. When the first version of Llama was released, we were going to do an academic release, but it leaked.

That's kind of cool, we're trying to figure out what to do with it. It's not a fire-and-burn situation. It's actually great, people really want it, we just have to figure out the right way to handle it.

The many different crises we face right now, I think it's an unfortunate phenomenon, probably partly because we've become a larger company, and partly because the world has changed, and the challenges we face now are more social and political than they are technical.

You see, the challenge with News Feed at the time was that it was a big change, but at the same time, you launched News Feed and traffic to the site went up 50% overnight because people could now find all this interesting content.

Then we had to deal with that and give people the control they wanted, but now a lot of things have changed. For example, people wanted this open source Llama, but we originally intended to release it as an academic open source, and then it immediately became widely used.

So we had two points of debate internally. One was whether this was a good thing and whether we should continue to do it, which was resolved very quickly. We thought it was a good thing and we should continue to do it.

The second problem is that political and social debates sometimes change very quickly. The open source debate is going to be one of the most important technology policy debates of the next five to ten years.

We've made a lot of progress, and part of the reason we've made progress is because a lot of people are using Llama. Quite frankly, entrepreneurship and startups are more popular than big tech companies.

So I think what we're doing, although whether it's popular in itself remains to be seen, but when there are a lot of people using it, then it's OK. People will be more willing to listen and think that this is a good thing.

Before we actually released Llama2, which was the first commercial open source model, there was a huge debate internally, mainly about the risks and concerns that dominated public opinion until many startups started using this technology. The question we needed to solve internally was how to execute this plan responsibly and securely in the context of open source.

My view is that if we do it well and we work hard on security, over time you build trust and you build a community. They're not going to want to take open source software away from us.

You see, today's big companies are built on open source software. If this debate goes well, the next generation of big companies will be built on open source AI. I think there will be some challenges along the way, but this will be one of the most important debates and we need to be careful about it.

That's why sometimes we have to be more cautious and sometimes we have to move slower in what we do, but that doesn't mean we have to spend all night in the Situation Room solving problems.

So from the leak of Llama 1 to the release of Llama 2, it was probably a few months, right? Six months, eight months or so, and while we were working through all these issues, the team was working on improving the model, but we were dealing with liability in the background, making sure we were doing a good job on that front and laying the foundation for the future.

host

That's great. All right, let's switch gears and answer some audience questions. What advice would you give to technologists who want to become entrepreneurs? When I work with a lot of entrepreneurs, I find that they often run into local optimal solutions. What advice would you give them?

Mark Zuckerberg

The idea of ​​"from -1 to 0" that you mentioned fits my philosophy very well, which is the way to explore the unknown. Looking back at when I first started my business, many people decided to start a business, and then they became very focused and started doing something. But the problem is that once you have a group of people doing something, it is difficult to change direction.

Of course, you can persevere through willpower, but it takes time, and in the early days you are actually exploring a very dynamic and broad space.

You need to be able to not only have meetings and make decisions in your head, but you also need to be able to pivot and change direction on the fly. For example, you might suddenly decide that I’m going to do something completely different now than I did a few hours ago.

I remember in college, before I started Facebook, I didn't think it was going to be a big company. Dustin and I started Facebook when I started it and he joined me because I was still in class and locked in a computer science lab writing code in PHP.

Dustin joined me at that time to help with some operational matters. He is really an amazing person and the company's operations are running smoothly thanks to him.

After that first spring, the spring when we launched Facebook, we went to Silicon Valley and thought it would be a great place to spend the summer, that's where all the companies started, and we thought, surely what we were doing now is not going to be the ultimate direction of the company.

Even though we were close to a million users, we still felt like this wasn't going to be a big company. What's the lesson here? I actually did a lot of different projects during college.

I like to build things, like I didn't know what courses to take, so I built a service that would scrape the course catalog, let people enter the courses they've taken and the courses they plan to take, and then show the relevance and reviews of all these courses.

Part of it is interesting because I started this project to answer my own questions about taking classes, but people actually like clicking through to see what classes other people took because people really want to get to know other people.

So all of these projects led to the original Facebook. But I probably did about 10 of these little projects. I was still in high school, doing angel projects.

I would say that when I’m working on a project, if a final is coming up, I might quickly make a tool to help with studying, like I did for a history course I took at Harvard on “Rome in Augusta.”

I spent the last few hours writing a tool that would grab all the artwork and let people input their historical significance, and then share it with the class so everyone could use it to learn.

Frankly, I really believe in this way of constantly building different things and not getting stuck in a fixed direction too early, especially in the startup stage, because when you're still exploring, it's hard to predict what the world will be like in the next five to ten years, but it's even harder to figure out your wedge into the future.

My advice is not to lock in one direction too early, but to build a lot of interesting things, learn different things from them, and try not to fall into a fixed pattern too early, because it is actually very difficult to fully learn and change direction when needed. From this perspective, my understanding is very consistent with your philosophy, which I think is very good.

host

In that sense, when people ask me about different projects I worked on at Facebook, I can tell them about some that were successful, but I was also involved in many that didn't work out.

We have an attitude internally that you should pursue something you think could become a great product very hard, but not so hard that your ego becomes tied to the project.

Because the whole point of building and trying things is that a lot of things aren't going to work, and that's OK. I still admire that about Facebook to this day, that you guys keep pushing hard but also accept and understand that not everything is going to work, and that's part of the hallmark of the company.

Mark Zuckerberg

Yeah, I've probably made more public mistakes than anybody in this business. There's a Michael Jordan commercial that I find very inspiring.

host

Yes, just like the ad says.

Mark Zuckerberg

I failed again and again, but in the end he concluded, "That's why I succeed." That's a good point. Walter Isaacson's biography of Albert Einstein makes a similar point: Einstein had many theories that were wrong, but he also had some very good theories, and that's what's important.

host

Just to ask a related question, I've worked for and helped many startups, but when I think back to the early days of Facebook, I feel very inspired.

That was a very unique experience, and I'm wondering what you did right in terms of recruiting? What were the key hires you made and why were those people the best choices?

Mark Zuckerberg

There were two factors. On the one hand, there were a lot of very excited and inspired people fresh out of college who had used our product and thought it was great and wanted to work at the company. On the other hand, I had failed for a long time at building an executive team.

We went through a lot of executive turnover because I had this Platonic model in my head of “what a head of engineering should look like.” They were probably all pretty reasonable people, but it was just a huge culture mismatch.

Until Peter Thiel took me out to lunch one day and said, "Clearly, this isn't going to work." He said, "You don't mesh with these executives, and you should work with people you want to spend time with." Because you were in the office almost all day, and your whole life was devoted to this job.

It's a bandwidth issue in early teams, you want everyone to speak the same language and have similar basic assumptions. Of course, you also want to challenge each other, and people with different opinions are important, but we also have enough common values ​​that things start to work smoothly.

When you mention this chaos, it's really part of who we are. Yes, these guys may be less experienced, but we are more in the same boat.

The great thing about the company now is that these people are now grown up, they are all very experienced, excellent executives, very experienced in the industry. We have been working together for 15 years. People ask me how to build this kind of team, and my answer is that this kind of team cannot be built in a vacuum.

What I do is I have a relatively large management team because we have so many different products. I want to be very hands-on, and to do that I don't want to just work through six or seven people.

I have a team of about 25 people who are responsible for about 15 product groups, ranging from the main application to advertising, AI research, smart glasses,Mixed RealityHelmets, all those different things, all the business functions. We spend a lot of time together every week. One of the ways I run the company is that I want it to be a tight-knit team.

One of the things I'm really proud of is that these product people are really good. Of course, there may be differences in the business side, like you want to hire a very experienced general counsel, which is very important. But Chris Cox, as the chief product officer, he is really outstanding.

host

Hey, everyone is busy.

Mark Zuckerberg

Yeah. I remember, Dustin and I had a huge argument about this because Chris was clearly a very promising young guy, and Dustin felt like we really needed him to be an engineering manager. But Dustin, you already have five engineering managers. I really need a head of people. Hiring was really hard back then.

host

He's a good personnel manager.

Mark Zuckerberg

Yeah, he was brilliant. He was very generous enough to do HR for a few years and then come back to product. But today, one of the things I'm most proud of is that all of our top product managers, none of them started out as product managers.

They all started at a lower level and worked their way up. We hired a lot of people as directors, and some started as individual contributors.

We even have people who started in executive roles who are now part of the management team and do a lot of different things. Some people started in non-technical roles and moved into technical roles. And vice versa. It's really cool. We have a really great team.

You can get through a lot of it on your own, but I really think that as you get further along, it really becomes a team sport. How much fun you can have and how hard the lows are really depends on the people around you. It's really hard.

Things always take longer than you expect, and you run into problems. That's probably one of the reasons why I've always been passionate and happy, is that I really like the people I work with, and they are my closest friends.

host

I'll follow up on that by asking, how do you manage the emotional ups and downs of starting and running a company? How has that evolved over the past twenty years?

Mark Zuckerberg

Yes, it's complicated. As the company grows, the cycles get longer. In the past, like you said, when we launched the news feed, we built the first version of the news feed in a relatively short period of time, and then people were immediately dissatisfied, but it was resolved a week later.

Now as we've gotten bigger and bigger with what we do, while it's fun, it also brings a fair amount of startup intensity. You're constantly going back and forth between feeling energized and feeling like you're about to collapse. I don't miss that feeling, it's exhausting. I don't think I could do it again.

Now the cycles are longer, and it takes a certain amount of willpower and the support of the people around you. It's also possible that you don't fully realize how bad it is when you're in it, until you get out of it and look back and realize, wow, that was a thrilling experience.

host

Hearing you say that, I feel like even for Mark, the energy of starting over is very strong. During the pandemic, I personally started to learn some new sports, such as learning to snowboard, becoming an intermediate surfer, learning to surf, and learning to seesaw, which is all great.

These sports give me the opportunity to connect with nature and learn new things. You’ve obviously started some new activities like MMA, seesaw, etc. What do you like about these sports? Is there any transfer learning between these sports and work?

Mark Zuckerberg

For me, the reason I like MMA is that, as I mentioned earlier, when you're running a larger company, especially when you're doing hardware, the cycles are very long. For example, you're training some basic model, and it might take nine months to work and then see how it works.

In contrast, in MMA, you need to focus immediately or you can actually get punched in the face. And when you can fight back, it feels great. I can no longer do that in the office, and now I have to release stress in the ring.

host

What is that?

Mark Zuckerberg

Alex, great post. But similar to your experience, I was also very competitive in sports when I was younger, then started a company and stopped most sports.

Basically, I stayed in shape, like lifting weights, but I didn't have specific fitness goals. Then in the early days of the pandemic, when everything went remote, I started running. Our family spent some time in Hawaii, and I started surfing, which I thought was really great.

You're out in nature, enjoying the beautiful scenery, you're getting a physical workout, and there's a goal to not get swallowed by the waves. It's fun and strategic.

Then, I no longer lived in Hawaii, so I started looking for alternative land-based sports. I decided to learn a martial art, and things just escalated from there.

We started training with a few friends, and then one of them decided to go compete. I thought that was pretty cool, and I was really curious to see how he would perform.

I myself would not have gone to the fight because I thought it would be absolutely humiliating. I would have been choked or something. But he went to the fight and did well, so I said, I've made an agreement with Priscilla that I can have an MMA fight.

I felt that in this kind of work environment, you can't take too many hits to the head, so I decided to just play one match. Then I injured my knee and now I'm back.

host

Has a date been set for that game? When is it?

Mark Zuckerberg

I haven't decided yet, I need to be confident with my knee. I guess in the spring, give myself some time to recover.

host

By the way, let me tell you another story about Mark's competitiveness.

Mark Zuckerberg

Yeah, it's a funny story. It was around 2009 when we decided we were going to have a push-up competition to see who in the office could be the first to do 10,000 push-ups.

According to the results, Alex Himalworn won, and he is now in charge of our AR glasses project. Mark was right behind. I still remember that everyone in the office was trying their best for this competition.

It was not because of the competition, but because he had been doing so well. He was an average engineer then, but now he is one of the top people in the company.

host

I have one last question for you. The world often asks me what kind of person I am, and I usually try to avoid answering that question.

Mark Zuckerberg

Sometimes I’ll glance and walk away.

host

Sometimes I tell them that you are one of my closest friends that I can lean on in my darkest moments. It always amazes me that the more famous you become, the more time you seem to make for us. I want to ask you, what are some things that are important to you that most people don't know how to appreciate?

Mark Zuckerberg

We talked about the physical activity part, and we talked a little bit about relationships and trust, but back to our original discussion, which is the difference between building something good and building something amazing.

There’s another dimension to this phase of my life, beyond just doing amazing things, and that’s doing them with people I really enjoy.

Sometimes I want to be involved in a project simply because I like that person, and I’ll go out of my way to find opportunities to work with them because they’re good people and I want to help them or learn from them and they’re interesting.

There are a lot of different things going on. Obviously the majority of my energy is still focused on Meta. One thing that's been really fun for me is that my relationship with Priscilla has expanded. We have a family now, but we also run CCI together, which has been really cool, seeing her growth, and this huge organization that we run.

Some things can be a little random, like I want to be involved in a project that aims to raise the highest quality beef in the world. Why? I don't know, but I want to do it with some great people.

Those are the things that are important to me. Part of it is that the company is big enough and successful enough that we need to continue to build momentum, I guess, and keep moving forward. But it's not because I want to hit a certain valuation that drives me to do these things, it's because I care about what we're building.

Being able to work on great things and being able to build really deep relationships is the most important part for me at this stage. Part of that is, when you're young, you don't have 20-year friendships because you're young.

Now we're at a stage in life where I have a lot of those friendships, which is really great. We're still young enough to continue to build those relationships, and it takes 20 years, but you can do it. That's what's important to me.

You've said so many nice things to me, and you're some of my closest friends. You're doing amazing things, and I'm forever grateful for the work you did in the early days of Facebook, because without it, we wouldn't be where we are today.

I joke that you guys are just really good people. Those who get to work with you and be mentored by you are really lucky. That's part of why I wanted to come here and do this, because I believe in what you're doing.

Philosophically, it's the right force to be in the world. If you compare it to some other incubators of different ideas, it may be a different philosophy, but at this stage in my life, that's what's important to me.