The Exchanges

Q I, I, I guess, I read at a certain point, you sort of realized that, that if you wanted to grow that business, you would have to Pivot, right? And then, and I guess that pivot was you, you sort of built a three D computer assisted, like design product for, uh, for like Engineers, mechanical engineers, people who wanted to see how something's going to work before it's built.

A That's right. And we entered a whole new market, a whole new type of customer, um, Kodak, Xerox, Ford, people making mechanical things where our product, which was called working model, which I love that name. It really described what we were trying to do. Um, way back then when this wasn't widely available, you could simulate how that elliptical trainer would work. You could simulate the car suspension. You could simulate the camera shutter mechanism. And so we built a pretty substantial business around that. Uh, we were partners with all of the three D design software at the time. And that's how we, that's how we ultimately built knowledge revolution to the point where it was acquired by another company.

AI assessment note: “That's right. And we entered a whole new market, a whole new type of customer”

Q How did you and Eric work together? Was it, I mean, you know, your partner's in this, but, but, I mean, I'm assuming he had a, probably strong views, and you had strong views, and there were things that maybe you want to try, or things that he, would you, did you argue, or was it, was it just easy?

A Some areas super easy, and some of the, um, core areas of even inside Roblox today, the way we run the in-memory database, the way we do networking, these are, these are still foundational things that Eric put together. Other areas we did disagree on, and one was that very first Roblox avatar, which became the Roblox blocky classic Avatar. We had a lot of debate. Eric thought, oh my gosh, this is still too rough. We need better looking avatars. But we came together with the notion it's better to, you know, get this live and get it in users hands as long as it's safe and robust and see what the feedback is, rather than spend, you know, another five or six months getting better avatars. That has, that's something we always came together, and I think In general, we were very aligned on this initial architecture. We would, we were probably whiteboarding an hour a day as we would go, so we were able to build things very quickly. You know, generally, the alignment was really high.

AI assessment note: “Some areas super easy... Other areas we did disagree on”

Q were probably 26, 27. You decide to start your own business, um, a software company, which we'll get into, with your brother, Greg. Um, you called it Knowledge Revolution, which is, I'm not, not judging you at all, but it's kind of a boring name. Um, a little bit boring. I'm not, no judgment, but it is, but what, what was it? What, what was the idea that you had?

A I was going to take two or three months off and decompress after my two difficult jobs. And we were right in the hardcore educational computing transition from Apple II to Macintosh. A lot of schools still had Apple IIs. A lot of schools were starting to get Macs. And I, I immersed myself in the educational software market, physics, math, history, played with everything. And it's interesting when you immerse in something from the outside, it always looks shiny and mature, but as I got into it, I started looking at the physics software market, and at that time, the physics software market, even very awesome, a lot of stuff going on, but typically, 10, 50 pre-made simulations to help animate and visualize classic physics things. I started going, oh my gosh, wouldn't this be cool if instead of these things being pre-canned, this is more like a physics construction set. And wouldn't it be cool if you could build all of these things from one simple system? Um, students could do it. You could model textbook experiments. That was the genesis of the idea for this product called interactive physics and the company knowledge revolution.

AI assessment note: “instead of these things being pre-canned, this is more like a physics construction set.”

Q of landed on. And, and when you, when you kind of came to that conclusion, had you already been talking to, to Eric Castle and, and your brother Greg or others who had worked with you at Knowledge Revolution about what you could do? Like, did you start to like call people up and say, hey, let's get the band back together? Do you remember what, what you were doing?

A Yeah, there was an early experimentation phase where I had a rough idea of what I was interested in. I was interested in the idea of a simulation on top of which people could build puzzles or scenarios and the idea of unlike a knowledge revolution where we had to be really accurate in what we built because people were designing stuff. I was very interested in the notion, could we make something that's very scalable so it can do a lot of things, but not be perfectly accurate? And I started down the route of making this simulator more as a puzzle experience rather than an immersive avatar type experience, and that's where Eric, um, joined me, and we started working on this together.

AI assessment note: “and that's where Eric, um, joined me, and we started working on this together.”

Q Collapse if you pulled out a brick or you could build things. I mean, did you see it as As an educational tool? Did you see it as something that, I mean, did you see it as something that could be, you know, turned into a tool like that? How were you thinking? Did you see it as a game, a video game, a massive online, you know, multiplayer game?

A Even, even early on, there was a bit of a thesis that if we could mirror and really couple raw things Three D cloud base. We, we used to call it digital stuff, simulated stuff in the cloud, that this is such a fundamental technology that on top of it, people would play. Someday they'd go to school or do simulations. They might work together. It was the notion that this was going to be a universal type technology that happened really early on. And then we released Roblox studio. Um, and the second we released Roblox Studio, it was one of those days where you go, oh my gosh, this is going to be absolutely amazing. We cannot believe in a couple days this early community at that time of 500 or a thousand people have all downloaded Roblox Studio, and we're seeing creations better than anything we can do. It's just such a fun experience when, when you think Think that that's going to happen, and then you see it actually happen.

AI assessment note: “on top of it, people would play. Someday they'd go to school”

Q first started to, like, do this kind of monitoring, right, early on, it, it must have been more like a cat and mouse game, right? Like, people doing things that were not what you intended them to do. I mean, you put a bunch of kids in, in a room together, and, and eventually some of those kids are gonna start beating the heck out of each other, you know?

A That, that's exactly right. There was, I remember when it was the four of us We all ran the moderation system. So we were on the front lines acting as moderators. And as we went, we were constantly building in, you know, any image that's ever uploaded, any sound that's ever uploaded needs to go by human moderators. Uh, the tension that for many creative developers, uh, who are very brilliant, very smart and very creative, it's actually fun to Push on Roblox and see, oh my gosh, what can I get away with? And it's, it's an area where we continue to, you know, go even further and get even better to this day, but it really started in that first month.

AI assessment note: “That, that's exactly right. There was, I remember when it was the four of us”

Q All right, so let's go back to Outlines of what would become Roblox, but, you know, at the time, I'm thinking there were some, I mean, this is still, you know, early internet, um, and most people had dial-up, and there were things like, I think, like Second Life and SimCity, right? Those were already out there at the time.

A That's right. Um, they had all, to some extent, built these systems that would You know, bring people together in immersive three D environments. Uh, well, we were in the midst of the very early phases of what we were doing, which was more of this super scalable physics puzzle system. We had some people play it, uh, but we knew within a week, oh my gosh, this is not going to go viral. We have a lot more work to do. And Eric and I had to say, okay, we've got another nine months or 10 months Where we're gonna have to buckle down, really make this multiplayer, avatar-based, you know, co-reality, if we want to get the, that virality, which was really our vision all along. It was just almost a realization there were no shortcuts to doing that. So, we shut that puzzle down and went back to work again for another nine months.

AI assessment note: “That's right. Um, they had all, to some extent, built these systems”

Q I read some of that stuff, and I'm not, not picking on you, but I read it, and I think, really? Is, is that really where we're headed, and is that healthy for us? It doesn't seem like we were, we were Emerged from the savannas of Africa to, to live in a virtual world. I just, I just don't think it's, it's where we were supposed to be headed.

A Yeah. I share your sentiments. And at the same time, I tend to be pretty optimistic. I'm optimistic that people over time will balance texting and phone and video and three D as ways to connect. And I'm, I'm, I'm optimistic that You know, in schools, um, there'll be a balanced amount of reading about history or using immersive three D technology to go to ancient history. And I'm, I'm, I'm kind of optimistic that the dystopian matrix in the VAT, um, type that's your world is hopefully a long, long way off. And we'll see a really lovely, more balanced use of this technology that We would use as a utility, and society will come up with a good balance on.

AI assessment note: “I'm optimistic that people over time will balance texting and phone and video and three D”

Q have that, like, a company that has user-generated content can really control it. Like, you can just build in algorithms or systems that can automatically, um, eliminate it, uh, because, you know, you're smart people, and you develop the system, and, and so you should be able to Automatically eliminate content that's objectionable or offensive or scary or dangerous for kids. Um, first of all, is that, is that possible?

A I think that we're always balancing, you know, when a new creator comes on the platform and they're at a computer tech camp, can they share that creation immediately with their friend? We're always balancing, does the average person on Roblox ever run on across any bad content? And so we're always balancing the kind of the control we put on that user generated thing. With yes, our ability to make sure there's virtually no bad content on the platform. So it is a balance. We, we have made the decision not to filter every single, um, Roblox creation through a day of moderation. At the same time, we constantly get better, um, through algorithmic social human review. And, and right now where we're at is it's very, very, very infrequent that anyone on our platform runs into intentionally bad content.

AI assessment note: “So it is a balance. We, we have made the decision not to filter”

Q Eric and Had been battling, um, cancer, and quite an aggressive cancer for a few years, and, and, um, and unfortunately succumbed to it in, in early 2013. Um, when that happened, I mean, he was the main, right, I mean, he had, he had sort of built the back end of, of this thing, um, Aside from, I'm sure, everybody being devastated, what did it mean for, for Roblox?

A You know, he had a cancer that was basically not curable. He knew he had, you know, between one and three years, I think, at the time, I forget exactly. It's just like, oh my goodness. So, what was interesting with Eric was, um, in the midst of that, he took a few weeks off, and with his family, um, But then didn't change anything. It was almost a testament to what we were doing together was his optimal lifestyle, which made me really open my eyes a bit. So, Eric kept working for quite a while. We all knew he was battling cancer. I tried to make it very easy, you know, hey Eric, if you want to chill out, we got this.

AI assessment note: “So, Eric kept working for quite a while. We all knew he was battling cancer.”

Q Legends. And at that time, in 2007, eight, I mean, was it hard, was it hard, like, was it hard to kind of make that argument to people, like, hey, you can buy things in the game? I mean, were people at that time saying, but it's not real, it's not real, stop, what am I buying? I mean, was that, I would think that would have been the pushback.

A We, there was a time in Roblox It's history where we were growing very rapidly on users and hours and all of that was good. But we had a really, it was kind of a difficult time where users were growing, but monetization was going the other direction. And we, we did maybe the first wave of our response is, did we break something? Did we, do we need to tune something? What's a spreadsheet of the 50 little tune-ups? Um, that we can do, and those were not successful, and it was, is a little bit scary. We knew in the back of our minds that, well, there's this really big system thing, and what that would mean is we have, we take our virtual currency, the developers themselves can bring it into their experience in any way they want. We will then cash out those developers, And so we said, okay, all hands on deck, we're going to build this system, not knowing whether it would work or not, but having a lot of intuition that it felt right, and that we, we sure think it's going to work, and everyone we talk to think it's really cool. And then we had a day similar to when we launched Roblox Studio, where we turned this on, And within two or three days, we had that same feeling, just, oh my gosh, problem solved. So those are the big things that are hard to build that, that power these types of systems.

AI assessment note: “users were growing, but monetization was going the other direction”

Q know it's just a stock price, but, uh, you know, does that stress you out? Do you get, I mean, you know, you are running a public company Do you, and everybody says, no, I don't worry about the stock price, of course, but I mean, how could you not worry about it when there are millions of shareholders out there? Does it, doesn't it stress you out at all?

A It's this interesting mix where we, we have these shareholders. We used to have our VCs, then our employees and our angels, and now we have people all around the world. So we have this amazing responsibility for them. At the same time, we, I mean, it's pretty fun to run this company. We have three billion of cash in the bank. We generate cash. We have strong values and amazing people. And so we, we keep coming back to, for our shareholders, we have to keep having this vision of the future. And just as we did five years and 10 years ago, have those five or six or seven really hard platforms, strategic things that we've got that, this, this funny feeling that if we do this, they're going to amplify and multiply in multiple ways. And that's just really fun.

AI assessment note: “we have this amazing responsibility for them. At the same time, we, I mean, it's pretty fun”