Part IThe Story
In the beginning there were four of them, and they took turns reading what the children wrote.
This was 2006, or near enough: a small office in Menlo Park, California, where a physics simulator had recently been given a ridiculous name and opened to the public. The men in the room were David Baszucki, his longtime engineer Erik Cassel, and the company's first two hires, John Shedletsky and Matt Dusek. They had written the code. They had built the first games themselves, because the creation tools weren't finished and somebody had to show what the blocks could do. When the doors opened, they also became the police. "Erik and I built the first Roblox moderation system right around when we launched Roblox," Baszucki would recall, years later, in a Reddit thread full of people asking why his company charged children money, "and 15+ years ago at the start, Erik, John, Matt and I all were the people working the moderation queues." On another occasion he said it more plainly: "We had just four of us sitting in a room, we were literally the moderators, like we would rotate all the time."
The record doesn't say what the room looked like. It doesn't say what came through the queue on those first afternoons, or who drew the late shift. What it says is that the founders of what would become the largest gathering place for children in the history of the species once knew, personally and by rotation, every flagged thing that happened inside it.
Fast-forward nineteen years. On a November 2025 episode of the New York Times podcast Hard Fork, the journalist Casey Newton asked Baszucki — six-foot-five, sixty-two, a billionaire on most days and a near-billionaire on bad ones — to describe "the scope of the problem" of predators on the platform. Three states had sued. More than twenty federal lawsuits accusing the company of enabling sexual exploitation had been filed that year alone. Baszucki began: "We think of it not necessarily just as a problem, but an opportunity as well." He went on to mention the four men in the room. And then he gave the numbers that the room had become: "At the scale we're at, 150 million daily actives, 11 billion hours a month, like what is the best way to keep pushing this forward?"
The distance between those two sentences — four men rotating through a queue; 150 million daily actives — is the whole of the story. Everything else is an attempt to measure it.
By the Numbers
The Room, Scaled
150MDaily active users, per Baszucki, November 2025
11BHours spent on the platform per month
$20MPrice MSC Software paid for Knowledge Revolution, December 1998
$41.9BEstimated valuation on Roblox's first trades, March 10, 2021
$0Baszucki's base salary after the listing
11.5MShares in the CEO performance award meant to be his sole pay through 2027
$3.60B2024 revenue — against a net loss of $935 million
The Prairie and the Encyclopedia
His parents, Helen and Paul, met at the University of Saskatchewan. Both were descended from Ukrainian immigrants who had settled in Western Canada — settled being the operative verb, the one that sits underneath all the others in this story. David Brent Baszucki was born in Winnipeg on January 20, 1963, and the family kept moving, east across Canada and then south, until it came to rest in Eden Prairie, Minnesota, a suburb whose name sounds invented for a biography and is not.
The boy who grew up there was outdoors a great deal. He rode dirt bikes. He built go-karts. Indoors, by the account his own philanthropy publishes, "he liked to read the encyclopedia," and he liked science fiction for the specific reason that it "transported him to different worlds." These are not exotic details for a future engineer. Rather, they are almost suspiciously on the nose — the kid who builds the vehicle and the kid who wants to leave the planet, housed in one body. At Eden Prairie High School he ran cross country, skied, played varsity soccer, played saxophone in the all-state band, and captained the quiz bowl team, which is to say he was rewarded, publicly and repeatedly, for having already read the encyclopedia. His physics project involved measuring the paint on a local landmark known as the graffiti bridge. The bio printed next to his senior photo in the Class of 1981 yearbook reads: "plans to design space colonies."
He was also, privately, writing games. On a 6502 Apple II. He didn't think this counted. Thirty years later, on a company blog, he would describe that oversight as the central mistake of his young adulthood.
Stanford came next — electrical engineering and computer science, a General Motors Scholarship, and two summers in Flint, Michigan, at Buick product engineering, in a lab focused on controlling car engines with software. It is worth sitting for a second with the image: a lanky Minnesotan in Flint in the early 1980s, writing code that would tell a combustion engine what to do. The go-kart kid, now with a debugger. He graduated in 1985.
And then nothing went the way the yearbook said it would.
Nine Careers in a Spreadsheet
The first business Baszucki ever ran was a window-washing operation. A friend from the dorms suggested it; he and his younger brother, Greg, knocked on doors and washed windows for a summer. "That was the first time I saw we could spin something up from nothing and make money," he has said, and the sentence is notable less for its content than for how small the thing it describes is. Squeegees. Ladders. Suburban glass.
Greg Baszucki — two years behind him at Eden Prairie High, Class of '83 — would become his co-founder at their first real company and, eventually, a member of the Roblox board. Of all the people who appear in his brother's story, Greg is the one with the longest tenure and the least public voice, which is a role that tends to be filled in family enterprises by the person the founder trusts most. The windows came first. Everything else followed from the fact that the two of them could stand on the same ladder.
After the summer came the part Baszucki now tells as a cautionary tale, usually to rooms full of Stanford students. "I can remember in this terrible time right out of college trying to figure out what I was going to do," he told an audience at the Graduate School of Business. "Rather than trusting my intuition, I can remember having a spreadsheet of nine potential careers and then all these metrics — 'it's really good for this, but it's not so good for this.'" He paused, in the retelling, on the absurdity. "It was, like, a really weird way to try to figure out your career."
The spreadsheet produced a salaried job. The salaried job, and the one after it, produced two or three years of what he has called "the absolute worst jobs in the world" and "massive disappointment." At one point he called his father to tell him how bad it was. He does not, in any account I have found, name the employers, and there's something gentlemanly in that — or something strategic, since the omission keeps the story about the spreadsheet rather than about anyone who hired him.
In March 2012, by then running a small company with a funny name, he wrote a blog post advising young people on how to get their first job. It is a strange and charming document — he recommends sending a hand-written note to the CEO of every company you admire, then calling the CEO's assistant two weeks later and saying, confidently, "I'm following up on a note I mailed" — but its key passage is a confession:
I remember creating spreadsheets and ranking them by categories, in an effort to figure out whether I should be an inventor, designer, or computer programmer. I should have listened to my intuition. My hobby had been writing games on my 6502 Apple II, which was a good indication of what I was passionate about.
Hold onto the spreadsheet. It will come back, in other forms, at nearly every turning point of this story — as a logic that tempts him, betrays him, and eventually becomes the thing he builds a company to escape. When Lifehacker asked him in 2016 for the one word that best described how he works, he gave them one word. "Intuitively."
The Children Who Wouldn't Do the Experiment
In 1989, four years out of Stanford, the brothers founded Knowledge Revolution. With them, from the beginning or close to it, was Erik Cassel.
Here the reporting voice has to admit something: the public record on Erik Cassel is thin, and almost all of it is refracted through the man who survived him. He was the lead engineer at Knowledge Revolution and its VP of engineering on Interactive Physics. He would follow Baszucki from one company to the next, then into a one-room office, and then into history as half the founding pair of a platform used by tens of millions of children he never met. He died of cancer in 2013. Beyond that, we have mostly Baszucki's pronouns — "Erik and I" — repeated across two decades of interviews like a habit he has chosen not to break.
Knowledge Revolution's first product, Interactive Physics, was an educational simulator for the Macintosh: a 2D world in which students could set up physics experiments — masses, springs, inclines, collisions — and watch the laws play out on screen. It was followed in the early 1990s by Working Model, a more serious mechanical-design tool. The Baszucki Group biography says the company eventually built "a 3D mechanical engineering simulator for companies producing a range of products, from cars to cameras to farm equipment to copy machines." Working Model still exists, now sold by a Michigan company that calls it "the best selling motion simulation product in the world," a claim that may or may not survive scrutiny but has certainly survived thirty years.
The commercial product, though, is not the important part. The important part is what the students did with it. Baszucki told the podcaster David Senra about it in April 2026, and he told it as an origin story that predates the one everybody knows:
— David Baszucki, to David Senra, April 2026Arguably, one could say, I started Roblox when I was two years old, because a lot of this shit really fascinates me.
And then, more specifically: "We had all of these kids, instead of using it to do their physics experiments, they were starting to build stuff. It was 2D at that time. It was just in the pre-Internet time, and so we could see them all trying to build and share their stuff."
A physics-education company noticed that its customers were misbehaving. The kids had been handed a laboratory and had turned it into a sandbox. They were supposed to verify Newton; they were making contraptions. And in a world without the internet, they were trying — on floppy disks, presumably, though the record doesn't say — to share them.
There are two ways a company can respond to customers who misuse its product. It can correct them. Or it can file the observation away for fourteen years and then build an empire on it. Baszucki, writing in the Harvard Business Review in February 2022, described the sentiment that survived the sale of the company with a single unglamorous phrase: "the notion of enabling meaningful co-creation stuck."
Keith Lucas, who worked under Baszucki at Knowledge Revolution and later at Roblox, has written that what he remembers is not the vision so much as the pace. "There is no one better at inspiring team urgency than David Baszucki," Lucas wrote in 2025, describing a boss who was "personally urgent," who "checks in regularly, setting the pace of results by the cadence of his check-ins," and who drove "long-term vision through a series of short-term bursts." Lucas had heard stories about Baszucki's college projects. He had lived the Knowledge Revolution years. What he describes, from the inside, is a man who was already running on the clock that would later define his company. "Working together with urgency and speed," Lucas wrote, "Dave and Erik created Roblox's secret operational sauce — a raw, get-stuff-done power."
In December 1998, MSC Software, a simulation company based in Newport Beach, bought Knowledge Revolution for $20 million.
A Little Astray
Baszucki stayed on at MSC as vice president and general manager from 2000 to 2002. He expected, after the sale, that someone would come and offer him a CEO job somewhere. Nobody did. "Time and time again," he told Fortune in 2025, "you have to participate in making your own reality."
What followed is the most interesting stretch of his biography, precisely because it has nothing to do with Roblox and everything to do with it. It was, by his own count, about two years. In conversation with Senra he described it as a sabbatical and then corrected himself: "In a way, I went a little astray. I started going almost more logical."
More logical. The spreadsheet, returning.
The things he did in those years read like the itinerary of a man trying out several possible selves. He formed Baszucki & Associates, an angel investment firm, which he ran from 2003 to 2004, and through which he put seed money into Friendster — the first great American social network, and the first great American social network to collapse, which gave him a front-row seat to both the dream and the failure mode of online community. From February to July 2003, he hosted a libertarian talk radio show on KSCO in Santa Cruz. And with his wife, Jan, and their children, he crossed the Western United States and Canada in a motorhome.
The radio show is the detail that people tend to skip, because it seems not to fit. It fits fine. A libertarian talk-radio host is a person who has thought hard, in public, about what happens when you let people govern themselves and what the minimum is that an authority must do. Six months on air in Santa Cruz is not a résumé line; it's a philosophical rehearsal. Roblox would become, among other things, an experiment in letting the people build almost everything — the games, the clothes, the economy — while the company retained a narrow and, it would turn out, extraordinarily consequential set of powers. That the founder once spent half a year arguing about the proper scope of government on AM radio is not irrelevant to how the company would later argue about the proper scope of its own.
Jan Ellison Baszucki deserves her own sentence, or several. A Stanford graduate who went on to work as a Silicon Valley fintech marketing executive, she later took an MFA at San Francisco State and won an O. Henry Prize for her first published short story — the first one, which is the kind of debut that ruins writers for the rest of their careers or makes them. Her novel, A Small Indiscretion, became a national bestseller in 2015 and was a San Francisco Chronicle recommended book that year. She now runs the family's philanthropy day to day and describes herself as "a writer, parent, mental health advocate and citizen scientist." The pairing of a novelist with a simulation engineer is the kind of thing that sounds like a contrivance until you consider that both are in the business of building worlds and then seeing whether people want to live in them.
The motorhome is the detail I find myself returning to. Wikipedia has nothing to say about it. The family foundation mentions it once, in passing, in a list. A man who had sold his company, failed to receive the job he thought he'd earned, and was trying on the costumes of investor and broadcaster, loaded his family into a vehicle and drove across two countries. There is no record of what he thought about on the road. But when he got home, the foundation says, "David was joined by Erik Cassel... in a one-room office in Menlo Park. They began writing the simulation code that would ultimately become the foundation of Roblox."
There's a reading list attached to his 2025 appearance on Tim Ferriss's podcast that includes, among the expected titles, a remarkable concentration of navigators — Roald Amundsen, Captain James Cook, Ferdinand Magellan, Joshua Slocum, Robert Falcon Scott — and a copy of Mutiny on the Bounty. Make of it what you like. The boy who planned to design space colonies grew up to admire men who left the map, and at least one story about the crew deciding the captain had it wrong.
A Room in Menlo Park
The prototype went through names like a band in a garage. In December 2003 it was eBlocks. Then GoBlocks. Then DynaBlocks. At some point it became Roblox — "robots" plus "blocks" — though the sources disagree about when: one account says January 2004, another says 2005, and the company's own founding mythology seems not to care. The paperwork is firmer. Roblox Corporation's original certificate of incorporation was filed with the Delaware Secretary of State on March 23, 2004.
The idea, as Wikipedia's careful summary puts it, was "a physics sandbox with creation tools and a social networking aspect" — which is to say, Interactive Physics plus Friendster, the misbehaving children plus the network they never had. And the decisive choice was made early: the platform would rely entirely on user-generated content. Roblox would provide the tools, and later the servers. The users would provide everything else.
This is the kind of decision that reads as obvious in retrospect and was terrifying at the time, because it means that on launch day you have nothing to offer but potential. Baszucki and Cassel worked alone on the early versions. They made their own games on the platform before the creation tools were finished. The avatars weren't animated. Features existed only in their most simplified form. They advertised on a few websites, and a few dozen playtesters showed up.
When he told the story to Guy Raz on How I Built This, Baszucki didn't dress it up:
We had some people play it, 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 where we're going to have to buckle down.
Another nine months. The phrase has the shape of a pregnancy and the mood of a prison sentence. In mid-2006 they hired Matt Dusek, who worked on the platform's communication features, and John Shedletsky. Roblox was formally released on September 1, 2006, with Roblox Studio — the creation tool — arriving that same year. The early community made paintball games, haunted houses, model trains you could ride.
And then something changed. Baszucki has described it to the Stanford GSB podcast with an almost religious plainness: "Something really magical happened and that is the day we launched the ability for people to create stuff. We just knew we had hit virality."
It's worth being precise about what this means. The thing did not go viral when it was a game. It went viral when it became a tool. The children who refused to do the physics experiment in 1990 had finally been given the right instrument.
There's a smaller decision from these years that Baszucki returns to in interviews, and it tells you more about his temperament than any mission statement. The original Roblox avatars were crude — blocky, programmer-drawn, faintly ridiculous. People advised him to make them more sophisticated. He shipped them anyway. He has cited this, on Shane Parrish's Knowledge Project, as an example of pairing a long-term vision with small, fast steps, and of trusting his instinct when it conflicted with feedback. Those blocky figures became the most recognizable silhouettes in children's media. He was right, in the way founders are sometimes right: not because he knew, but because he refused to wait.
In 2008, the company stopped making its own games entirely. From then on, every experience on Roblox would be made by someone who did not work there. The corporation would build the world's floor and walls, and its weather. It would not build the houses.
It would, however, run the bank.
The Price of a Penny
On August 16, 2007, John Shedletsky posted an announcement on the Roblox blog under the headline "Builders Club is Here!" The subscription let members create more games, sell virtual clothing, strip outside ads from the site, and receive a daily allowance of Robux. Turbo and Outrageous tiers followed, because of course they did. In those years the platform ran two currencies — Robux and Tickets, which everybody called "Tix" — a dual system that lasted until April 2016, when the company killed Tickets on the grounds that new players found two currencies confusing. Baszucki wrote the farewell himself.
The economy was the second great bet, after user-generated content, and it was a bet about trust. Baszucki has said — in the Knowledge Project conversation — that the company has held the value of a Robux near one U.S. cent for twenty years. Twenty years of monetary stability is longer than many national currencies have managed. A central banker would understand the logic instantly: an economy where children trade hats and swords and pets can only function if the unit of account doesn't move under their feet.
By August 2011 the platform had about 6.8 million monthly active users and was the second-most popular kids' entertainment site. Profits that year were up 75 percent over 2010. In June 2011 the company raised just over $4 million. In August, the first in-person Roblox Rally in San Francisco drew more than 1,000 people. That same month, VentureBeat profiled an eighteen-year-old named Luke Weber — "Stickmasterluke" on the platform — whose games had accumulated more than ten million plays. A teenager with ten million plays, in 2011, building for free inside someone else's machine. He was the proof of concept and the open question at once: if the kids were the creators, who would pay them?
The answer arrived around 2013 as the Developer Exchange, or DevEx, which let creators convert earned Robux into real money. At first, payouts were capped at $500. The cap disappeared; the threshold settled at 30,000 Robux; the exchange rate settled at $0.0035 per Robux. In 2017 the company said creators had collectively earned at least $30 million. In 2020 roughly 345,000 developers earned money through the program, and the community was projected to take home around $250 million. Several individual games crossed $10 million in revenue. The platform said more than two million developers were using Roblox Studio to make more than 20 million games a year — and that most of them were minors.
Notice the two numbers sitting next to each other. A Robux costs a player roughly a penny. It pays a creator roughly a third of a penny. The difference — the platform fees, the app-store cuts, the hosting, the moderation, the margin — is the entire business, and it's also the entire controversy.
On August 19, 2021, the investigative YouTube channel People Make Games released a video alleging that Roblox took an outsized share — 75.5 percent — of the revenue generated by games on its platform, and likened the arrangement to company scrip, the private currency that coal towns once paid their miners in. Roblox's chief product officer, Manuel Bronstein, told Axios the company intended to give more money to developers. The company also asked People Make Games to take the video down. The channel responded by making another one.
Baszucki's own defense of the model, offered in an October 2021 Reddit AMA to a user who asked whether a game aimed at children should contain any purchases at all, was that the company had always tried to do several things simultaneously: prioritize safety and civility, "balance openness and freemium," and support the growth of a developer base "as more and more of them wanted to make a living and build studios on Roblox." He ended on the line the company returns to whenever the question comes up: "The vast majority of players can enjoy Roblox without spending any money on the platform."
Both things are true, which is the problem with them. The company built one of the few genuinely functioning creator economies on the internet, one that has turned teenagers into studio owners. It also built a slot through which children's allowance money passes into a vault and comes out, at a rate of a third, on the far side. The Robux holds its value. The question of who it's valuable to depends on which side of the exchange you're standing on.
What Erik Didn't See
The middle years were long and mostly unglamorous. In May 2012, a stripped-down iOS app arrived that let you socialize but not play. On December 11, 2012, the full iOS version shipped. Android followed on July 16, 2014. Xbox One followed on November 20, 2015 — and for the console launch, out of concern about Entertainment Software Rating Board standards, the company did something it had sworn off in 2008: it chose the games. Fifteen of them, hand-selected by Roblox staff. It was the first time in years that the corporation had stepped back into the role of curator, and it did so not because it wanted to but because a gatekeeper larger than itself required it. The pattern would repeat.
Erik Cassel died of cancer on February 11, 2013. By December of that year the company had 68 employees.
That is almost all the record says. The Baszucki Group biography adds a single sentence that does more work than any obituary: "David continues the work of Roblox in Erik's memory." It's the kind of line one writes for a foundation website, and it's also the kind of line a person could build a life around.
What Cassel didn't see is easy to enumerate. He didn't see the guests disappear — in January 2014, most of the platform's 10 to 12 million monthly players didn't even have accounts, and in October 2017 the company ended guest play entirely, forcing everyone to become someone. He didn't see December 2016, when monthly users reached about 30 million, or April 2019, when they reached 90 million. He didn't see a version of the company with 163 employees, or 830, or 2,474. He didn't see the pandemic.
In the spring of 2020, as schools closed, Roblox became the place children went instead of the playground. Between March and April it reported more than 120 million monthly players — a population that, the company said, included more than half of all American kids under sixteen. The role-playing game Adopt Me! passed 10 billion visits by July and set a platform record of more than 1.6 million users at once. Roblox grossed $2.29 billion in 2020, the third-highest of any game in the world, behind two Tencent titles. In February of that year, a Series G round led by Andreessen Horowitz had valued the company at $4 billion. In January 2021, a Series H led by Altimeter and Dragoneer valued it at $29.5 billion.
In July 2020, Baszucki formally announced the plan that Roblox's developers would build a "metaverse," borrowing the word that would, within eighteen months, be seized by Mark Zuckerberg and very nearly ruined. He cited Ready Player One and its sequel as inspiration and described a thirty-year vision in which users would have "a fully-fledged digital identity." The boy who wanted to design space colonies had found his colony; it just happened to be made of blocks.
The company considered a traditional IPO, filed for one, and then switched to a direct listing — no new shares sold, no underwriters' pop, just the doors opened and the existing owners allowed to trade. On March 10, 2021, Roblox's Class A shares began trading on the New York Stock Exchange under the ticker RBLX, at an estimated $41.9 billion valuation on the first buys. By the close, the figure being reported was $45 billion. Bloomberg calculated that day that Baszucki's wealth had soared to $4.6 billion. Cassel had been dead for eight years and one month.
The compensation documents from that winter are, in their way, a portrait. In a November 18, 2020, letter, Baszucki's base salary was listed as $500,000. A revised letter, dated March 3, 2021 and effective February 21, raised it to $800,000 — and then specified that on the date the registration statement became effective, "your annual base salary will be reduced to $0." In its place was a performance award of up to 11,500,000 restricted stock units, which the company said it intended to be "the exclusive cash and equity compensation that you will receive from the Company through 2027." He announced he would give away any future compensation from the listing to philanthropy. The corporate charter, meanwhile, had authorized 65 million shares of Class B common stock — the high-vote class that keeps founders in control.
So: zero salary, giant performance grant, a pledge to give it away, and a dual-class structure that guaranteed he'd be around to earn it. Business Insider ranked him the seventh-highest-paid CEO of 2021, at $232.8 million, as the grant was valued. That December, a New York Times investigation reported that Baszucki and his relatives had used a tax break intended for small business investors to legally avoid tens of millions of dollars in capital gains taxes. Nothing illegal was alleged. Something was, nonetheless, being alleged — that the man who'd pledged the money away had first made sure the government got less of it.
In May 2021, in the course of the Epic Games v. Apple litigation, Apple argued that Roblox developers don't really make games. Roblox, which needed to stay on Apple's good side, agreed. From then on, the things on Roblox were "experiences." The company that had been founded on the conviction that children could make real games had, for reasons of app-store diplomacy, officially stopped calling them that.
The Arm That Was Broken
The interview with Shane Parrish opens in a place the host himself calls unexpected. "I want to talk about how keto changed your family's life," Parrish says, and Baszucki answers at length.
"Then this thing happened with my son where he got super bad bipolar, probably 8 to 10 years ago," he said. "We tried every medication. We tried every solution to get him off medication. And here we were four or five or six years later, doctors had told him, 'You're going to be on five medications for the rest of your life,' he wasn't quite himself, lots of hospitalizations, all kinds of stuff. And then out of the blue, someone I knew in the metabolic health community just said, 'Hey, by the way, keto solved my bipolar.'"
They tried it, under medical supervision. Within three weeks, still on heavy medication, his son said something like I feel a little different. The real person started to show through. Jan wrote about it in the San Francisco Chronicle under a headline that does not hedge: "Doctors Said My Son's Bipolar Disorder Couldn't Be Healed by Diet. They Were Wrong." She hopes to write a memoir about it.
The family's philanthropy is downstream of this. In March 2021, the month of the listing, David and Jan launched the Baszucki Group and the Baszucki Brain Research Fund to make grants for bipolar research. In December 2021, the University of California, San Francisco launched the Baszucki Lymphoma Therapeutics Initiative with $6 million from the family over five years. In September 2022, Baszucki, Sergey Brin, and Keystone Capital chairman Kent Dauten committed a combined $150 million toward bipolar disorder research and treatment. Jan founded Metabolic Mind, a nonprofit studying the link between metabolism and psychiatric illness.
What's striking is how the experience rearranged his categories. "I never thought of it as a super physical thing," he told Parrish, "'cause now I think of it just the same as a broken arm." A disorder of the mind, reframed as a disorder of the body's energy supply. An engineer's revelation: the software was running on bad hardware.
And then — this is the part that makes the story his, rather than simply a family's — he turned the instrument on himself. Baszucki wears a continuous glucose monitor. He has said he tries to make big decisions when his metabolic state is optimistic and low-carb, not in the sluggish aftermath of a heavy meal. He tells managers never to send an email when they're angry; wait until morning. Check your state before trusting your judgment, as the episode notes summarize it. The man who once ranked nine careers in a spreadsheet has spent his later life trying to measure the instrument doing the ranking.
Read one way, this is the logical endpoint of the spreadsheet — the quantified self, applied to intuition. Read another, it's the opposite: an admission that the gut is a physical organ, and that trusting it requires keeping it in good repair.
The family foundation lists David's philanthropic interests as "metabolic health, democracy and astronomy." Space colonies, still. In December 2025, the New York Times reported that he had given at least $2.5 million toward the construction of the White House State Ballroom, which the paper noted in the context of post-election donors who had benefited from the new administration. The foundation's bio does not list ballrooms among its interests. People contain multitudes; so do their checkbooks.
Not Just a Problem
By November 2020, when Roblox filed its registration statement, the company had more than 830 full-time employees and 1,700 "trust and safety agents" — roughly two moderators for every employee. The four men rotating through a queue had become a contracted army larger than the company itself. The filing listed child sexual exploitation among the risks to the business.
The army was not enough. A 2020 Fast Company investigation described the company's efforts against sexual content as whack-a-mole. Wired in 2021 documented games in which players role-played as Nazis and recreated real massacres. In July 2024, Bloomberg Businessweek ran a cover story headlined "Roblox's Pedophile Problem," attributing predation on the platform to moderation that leaned too heavily on artificial intelligence and to user anonymity. That October, the short-seller Hindenburg Research accused Roblox of inflating its user metrics and called the platform, in a phrase that would follow the company for years, "a pedophile hellscape for kids." Roblox refuted the allegations. Turkey banned the platform in August 2024. Qatar, Iraq, Algeria, Palestine, and Egypt followed over the next eighteen months; Kuwait banned it and then un-banned it after the company agreed to new rules.
Against this backdrop, Baszucki gave a series of interviews that did not go well.
On March 13, 2025, the BBC asked him what he would say to parents who didn't want their children on the platform. "My first message would be, if you're not comfortable, don't let your kids be on Roblox," he said. "That sounds a little counter-intuitive, but I would always trust parents to make their own decisions." It was, unmistakably, the libertarian talk-radio host speaking — parental sovereignty, consumer choice, exit as the ultimate form of voice. Ellie Gibson of the Scummy Mummies podcast said it risked sounding "a bit of a get out." Mumsnet's Justine Roberts pointed out that "if you've got multiple children you're looking after... you probably can't 24/7 watch everything they're doing." The BBC, for its part, created a fifteen-year-old account and a twenty-seven-year-old account on unlinked devices and got them messaging each other. In the same interview Baszucki said, "We do in the company take the attitude that any bad, even one bad incident, is one too many."
In July 2025 he suggested Roblox might be a good platform for online dating, a place where lonely adults could meet — "a virtual date to start, and then if they connect, move to the physical world." PC Gamer called the remarks "tone-deaf" under a headline about horrifying every parent alive.
Then came Schlep. Schlep was a YouTuber who ran sting operations on Roblox, posing as a child to catch adults who tried to groom minors, and publishing the results; The Times of Israel described him as a "self-appointed watchdog with Yiddish name." His career is a short, sharp biography of the age: a young man who decided the platform's police were insufficient and deputized himself. On August 9, 2025, Roblox banned him, terminated the accounts associated with his group, IP-banned him, and sent a cease-and-desist letter saying that vigilante activity violated its terms and created an unsafe environment. The company's position — that amateur stings could lure predators by impersonating children, and that freelance policing does more harm than good — was not crazy. It was also delivered to an audience that heard only that Roblox had punished the man catching predators.
A Change.org petition demanding Baszucki's resignation passed 100,000 signatures. Representative Ro Khanna launched his own petition. Members of the Roblox Video Stars program quit in protest. The stock fell more than 11 percent. On August 14, Louisiana Attorney General Liz Murrill sued, alleging that by shutting down independent efforts to expose dangers, Roblox was failing in its duty to protect children; her office called the platform the "perfect place for pedophiles." Qatar blocked it. Indonesia threatened to. Kentucky and Texas sued.
There was a precedent, if anyone was looking. In November 2021, Roblox had sued a banned YouTuber named Benjamin Robert Simon — Ruben Sim — for $1.65 million, alleging harassment of employees and fake terrorist threats that shut down part of that year's developer conference. He settled for $150,000 in January 2022 and was permanently barred. The company had shown, once before, that it would use the courts to enforce its monopoly on governing its world. The difference was that Ruben Sim was a villain in nearly everyone's telling. Schlep was a hero in many.
Which brings us back to November 2025, and Hard Fork, and "not necessarily just as a problem, but an opportunity as well." Kotaku called it "the most astonishingly tone-deaf response." The conversation, by the host's own subsequent framing, "got tense." What Baszucki was trying to say — that the safety work was also the work of building the future of communication — is a defensible idea, the sort of thing a product person says sincerely. It is also the sort of thing no parent wants to hear about predators.
Six months later, at Stanford, he gave a different answer to the same question — what happens when engagement and safety pull in opposite directions — and it was the clearest thing he has said on the subject in public:
— David Baszucki, Stanford GSB View From The Top, 2026We always go safety. I think we have to design our product assuming an 11-year-old is just handed a phone.
The policies that followed were the most restrictive in the platform's history. That November, the company announced facial age estimation as a condition of chatting. In December 2025, age checks became mandatory for communication in Australia, New Zealand, and the Netherlands; in January 2026, everywhere. Users were sorted into age bands that could only talk to nearby bands — a twelve-year-old, for example, only with users nine to fifteen. In April 2026, the company reached a $12 million settlement with Nevada, with funds directed to youth programs and law enforcement and a commitment to keep messaging unencrypted so it could be monitored. In June 2026, Roblox divided itself into tiers: Roblox Kids for users under nine, Roblox Select for nine to fifteen, and standard Roblox for sixteen and up. Unverified users would be confined to games rated 9+ or lower. Publishing a game now required an age check. In August 2026, the European Union designated Roblox a Very Large Online Platform under the Digital Services Act. Russia, which had banned it, lifted the ban that June.
Nineteen years after four men took turns reading what the children wrote, the company had rebuilt the room — with walls this time, and doors keyed to the age of your face.
None of this resolves anything. More than 25 million users played a single game, Steal a Brainrot, at the same time that summer. Grow a Garden, a game made by an anonymous teenager, pulled roughly 22 million concurrent players and briefly broke Fortnite's record. The misbehaving children of 1990 had become the industry. The question of who watches them remained, as it had always been, the question.
Builderman
Baszucki's username on the platform is builderman. It has been since the beginning. Somewhere along the way, his users turned it into a character — a mascot, a meme, a minor deity of the Roblox cosmology, the man who appears when the game welcomes you.
On Christmas Day 2024, a horror game on Roblox called Forsaken — made, like every game on Roblox, by people who do not work for the company — added a new playable survivor. The game's wiki describes him as "the big boss of Roblox HQ himself." He costs 500 of the game's currency. He has 90 maximum health, slightly below average. His playstyle is Support. He carries a hammer, and with it he builds two things: sentries that slow the killer's path, and dispensers that heal the people around him. His lore says that he "tries his best to show care for the ROBLOX interns and other Survivors, doing what he can to protect them due to his mistakes." His description ends with a line that the fans who wrote it could not have known would land the way it does: "As a promise with Shedletsky, he'll do all he can to help everyone."
When you select him, he says, "Welcome to ROBLOX!"
And then the killer comes, and he kneels in the dark, and he starts to build.
Part IIThe Playbook
What follows is an attempt to extract operating principles from a career that has, at different moments, looked like a masterclass and a cautionary tale — often in the same quarter. Some of these principles are things Baszucki says he does. Others are things the record shows he did, whether or not he'd describe them that way. A few are lessons he paid for in public. They are offered with that ambivalence intact, because the most useful playbooks are the ones that include the plays that didn't work.
The principles
- 1.Watch what users do with your tool, not what you sold it for.
- 2.Get out of the content business you are best at.
- 3.Ship the blocky version.
- 4.Let the second idea marinate for a decade.
- 5.Run your currency like a central bank.
- 6.Pay creators in real money — and be honest about the spread.
- 7.Treat bureaucracy as entropy, not equilibrium.
- 8.Audit the instrument before trusting the reading.
- 9.Hold a thirty-year vision through two-week sprints.
- 10.Make your pay a public statement of time horizon.
- 11.Design for the most vulnerable user from day one — and speak like it.
Principle 1
Watch what users do with your tool, not what you sold it for
Interactive Physics was sold as a laboratory. Students turned it into a sandbox. A less attentive company would have treated this as noise — or worse, as misuse to be corrected in the next version's tutorial. Baszucki treated it as the most important market signal Knowledge Revolution ever received, even though he couldn't act on it for more than a decade. "Instead of using it to do their physics experiments, they were starting to build stuff," he told David Senra. The insight wasn't about physics. It was about a latent demand — to make and share — that the product had accidentally revealed.
The deeper discipline here is to treat deviance as data. When users route around your intended use case, they're telling you what they actually value, often more honestly than any survey. TechCrunch later reported that successful Roblox games defied free-to-play orthodoxy, and that adding tutorials actually decreased engagement. The users, once again, knew something the experts didn't.
Tactic: Once a quarter, pull the ten most "off-label" uses of your product and ask whether any of them is a bigger business than the one you're running.
Principle 2
Get out of the content business you are best at
In 2008, Roblox stopped making games. This is a remarkable decision when you consider that Baszucki and Cassel were, at that point, among the most experienced builders on their own platform — they'd made the first games themselves, because there was no one else to make them. Exiting first-party content meant surrendering the thing they were best at in order to become the thing nobody else could be: the substrate.
Today, as Baszucki told TechStuff, "Roblox doesn't make any of the games on our platform." The payoff is the shape of the hit list: Adopt Me!, Grow a Garden, Steal a Brainrot — games no Roblox employee would likely have greenlit, made by people the company never hired. A platform that competes with its own creators eventually loses them. A platform that refuses to compete gets Grow a Garden for free.
Tactic: Identify the content or service your team produces that a third party could produce instead, and set a date to stop.
Principle 3
Ship the blocky version
The original Roblox avatars were programmer art, and people told him to make them more sophisticated. He launched them anyway. On the Knowledge Project, the example is offered as a case of pairing an ambitious long-term goal with rapid iteration. The first release in 2006 was, by his own account, a dud: "We knew within a week, oh my gosh, this is not going to go viral." The response wasn't to retreat into a longer build cycle; it was nine more months of buckling down, followed by the moment they released creation tools and "just knew we had hit virality."
The principle is not that polish is bad. It is that polish applied before you know what people value is wasted, and often counterproductive — the crude avatars became iconic precisely because they were simple enough for children to understand, customize, and draw. Baszucki has said he favors "a few excellent tools over complicated products." Simplicity, here, is not a shortcut. It's a design strategy for a user base that includes ten-year-olds.
Tactic: Before your next launch, list every feature that exists to look impressive rather than to be used, and cut half of them.
Principle 4
Let the second idea marinate for a decade
The observation that kids wanted to build and share was made at Knowledge Revolution around 1990. Roblox launched in 2006. In between came the sale, two years as a corporate VP, an angel portfolio that included Friendster, six months of libertarian talk radio, and a family motorhome trip across the West. Baszucki describes part of this period as going "a little astray." The record suggests something more useful: an interval in which he absorbed the social-network wave firsthand (and saw its failure mode up close), rehearsed in public his ideas about self-governance, and let the original insight combine with new infrastructure — broadband, cheap servers — that hadn't existed when he first saw it.
Founders are trained to believe that speed of execution is everything. Baszucki, a famously urgent operator, nonetheless sat on his best idea for fourteen years. The timing wasn't procrastination; it was the internet catching up to the idea.
Tactic: Keep a written file of observations that seemed important but premature, and reread it every time a major enabling technology shifts.
⏱
The Long Gestation
From the misbehaving students to the misbehaving platform.
1985
Graduates from Stanford in electrical engineering as a General Motors Scholar.
1989
Co-founds Knowledge Revolution with brother Greg; Erik Cassel leads engineering on Interactive Physics.
1998
MSC Software acquires Knowledge Revolution for $20 million in December.
2003
Hosts libertarian radio on KSCO Santa Cruz, February–July; angel-invests via Baszucki & Associates.
2004
Roblox Corporation incorporated in Delaware on March 23.
2006
Roblox released September 1; Roblox Studio follows the same year.
2008
Company stops making its own games entirely.
2013
Erik Cassel dies of cancer on February 11; Developer Exchange arrives around this time.
2021
Direct listing on the NYSE on March 10; Baszucki Group founded.
2025
Schlep ban in August; facial age estimation announced in November.
2026
Global age checks for chat in January; age-tiered accounts in June.
Principle 5
Run your currency like a central bank
By Baszucki's account, a Robux has held near one U.S. cent for twenty years. In 2016 the company eliminated its second currency, Tickets, because two units of account confused new players. These are the decisions of someone who understands that a virtual economy is a trust machine. If the currency inflates, every item a child has saved for loses value; if it's confusing, newcomers stall at the door.
The rarely appreciated point is that stability is itself a product feature. Creators set prices, players save, limited items trade on secondary markets — all of it depends on the unit staying put. Most companies with an in-app currency treat it as a pricing lever. Baszucki treats it as monetary policy, with the conservatism that implies.
Tactic: If your product has any internal unit of value — credits, points, tokens — write down the conditions under which you would never change its price, and publish them.
Principle 6
Pay creators in real money — and be honest about the spread
The Developer Exchange turned a hobby platform into an economy. By 2020, some 345,000 developers were earning money and the community was projected to take home around $250 million. Teenagers like Luke Weber became, effectively, studio founders. No other children's platform of comparable scale had offered that.
The critique is also part of the playbook. Players buy Robux at roughly a penny; creators cash out at $0.0035. People Make Games put the platform's share at 75.5 percent and compared it to company scrip. Roblox's initial response — asking the channel to take down its video — produced a sequel. The lesson isn't that the take rate was indefensible; app-store fees, hosting, and safety operations are real costs. It's that a spread you don't explain will be explained for you, by someone less sympathetic.
Tactic: Publish a plain-language breakdown of where every dollar a creator generates goes, before a critic does it for you.
⚖
Bets and Their Bills
The major structural decisions — and what each one eventually cost.
| Decision | Phase | Payoff | Bill that came due |
|---|---|---|---|
| Rely entirely on user-generated content | Scan | Infinite catalog at near-zero content cost | Moderation burden scales with every creator |
| Ship crude avatars despite advice | Try | Iconic, customizable silhouette | Years of "kids' game" stigma |
| Stop making first-party games (2008) | Try | Hits from anonymous teenagers | Less control over what the platform contains |
| Developer Exchange for real money | Scale | Hundreds of thousands of earning creators | "Company scrip" critique over the spread |
| Direct listing, $0 salary, giant performance grant | Scale | Signal of alignment with long-term holders | $232.8M pay headline; tax-break scrutiny |
| Ban vigilante predator-hunters (2025) | Risk | Preserve the company's monopoly on enforcement | Petitions, lawsuits, an 11% stock drop |
Principle 7
Treat bureaucracy as entropy, not equilibrium
One of Baszucki's more pointed claims, as summarized from his Knowledge Project conversation, is that "leaving an organization alone doesn't keep it stable." Bureaucracy, in his model, compounds unless it is actively destroyed — the way disorder accumulates in any closed system. This is a physicist's view of organizations, and it suits a man whose first company sold simulations of forces acting on bodies.
The operational corollary shows up in his insistence on naming a single decision-maker and giving that person room to think, rather than pursuing group consensus. It shows up in Keith Lucas's description of a leader who "focused on what mattered over the merely important, keeping the team small through repeated growth surges." The company went from 68 employees at the end of 2013 to 163 by the end of 2016, a period in which monthly users nearly tripled. The headcount stayed lean because someone was always pushing back against the natural tendency of every process to grow.
Tactic: Assign one person, every quarter, the explicit job of killing a process, meeting, or approval step — and report what was killed.
Principle 8
Audit the instrument before trusting the reading
The young Baszucki trusted a spreadsheet over his gut and lost years to it. The older Baszucki trusts his gut — but wears a continuous glucose monitor to check whether the gut is in good working order. He describes making big decisions when he's in an optimistic, low-carb state rather than a sluggish one, and he counsels managers never to send an email in anger. His son's recovery from bipolar disorder, which he credits to ketogenic therapy, convinced him that the mind runs on physical hardware, "just the same as a broken arm."
The non-obvious lesson isn't about diet. It's that intuition and measurement aren't opposites. "Trust your gut" is useful advice only if you know when your gut is compromised. The executive who never checks their own state is trusting an uncalibrated instrument.
Tactic: Before any irreversible decision, write one sentence about your physical and emotional state, and postpone if the sentence makes you wince.
Principle 9
Hold a thirty-year vision through two-week sprints
Baszucki has described a thirty-year plan for Roblox in which users would have "a fully-fledged digital identity." He calls the current moment the "prehistoric era" of gaming and compares today's games to "the horse and carriage." This is a man who plans in decades. And yet the people who have worked for him describe him almost entirely in terms of speed — urgency, check-ins, short bursts. The company values include both "Take the Long View" and "Get Stuff Done," with no apparent sense of contradiction.
The trick is that the long view sets direction while the sprint sets pace. Without the vision, iteration becomes thrash. Without the sprints, vision becomes a slide deck. Lucas's summary is the cleanest version: Baszucki drives "long-term vision through a series of short-term bursts," and is "responsively dynamic, extending goalposts if ideas get traction and stopping projects that don't yield results."
Tactic: Write your company's ten-year destination on one line, then require every two-week plan to name the specific step it takes toward that line.
Principle 10
Make your pay a public statement of time horizon
On the day Roblox's registration statement became effective, Baszucki's base salary dropped from $800,000 to zero. In its place: a performance award of up to 11.5 million RSUs, intended to be his only compensation through 2027, and a public pledge to give away any additional compensation. The structure tells shareholders, employees, and creators exactly what he's optimizing for and over what period.
The caveat matters as much as the principle. The award produced a $232.8 million headline in 2021, and the Times reported on the family's use of a small-business tax break to avoid tens of millions in capital gains. Signals of alignment are powerful precisely because they invite scrutiny of everything adjacent to them. A founder who says "judge me by the long term" will be judged by every short-term choice that seems to contradict it.
Tactic: If you restructure your compensation to signal long-term alignment, assume every related financial decision will be read against that signal — and make them accordingly.
Principle 11
Design for the most vulnerable user from day one — and speak like it
Roblox began with its founders personally working the moderation queue. By 2020 it had roughly two trust-and-safety agents for every full-time employee. It still ended up the subject of a Businessweek cover, a Hindenburg report, bans in a half-dozen countries, and lawsuits from multiple states. The structural lesson is that safety cannot be a department sized to the problem as it existed last year; it has to be a design premise sized to the problem as it will exist at a hundred times the scale.
Baszucki eventually said it about as well as it can be said: "We always go safety. I think we have to design our product assuming an 11-year-old is just handed a phone." The age-checked chat, the age-banded communication, the Kids and Select tiers — these are what that sentence looks like in product. The companion lesson is about language. "If you're not comfortable, don't let your kids be on Roblox." "Not necessarily just as a problem, but an opportunity." Each was defensible in context and devastating out of it. When the stakes involve children, the founder's vocabulary is part of the safety system.
Tactic: Have someone who is not in the industry — ideally a parent — read your planned public answer to the hardest safety question aloud, and rewrite anything that sounds like a product roadmap.
Free playbook
Get The Business Model Playbook
58 business models, one visual page each: how the money flows, the metrics that matter, and who runs it. Free when you join the Faster Than Normal email.
Free. No spam. Unsubscribe anytime.
Part IIIQuotes and Maxims
In their words
— David Baszucki, to Stanford students, as reported by Fortune, December 2025A lot of my development has been trying to, over time, ignore advice I've been given.
— David Baszucki, Fortune, 2025Time and time again, you have to participate in making your own reality.
— David Baszucki, on his son's recovery, The Knowledge ProjectI never thought of it as a super physical thing, 'cause now I think of it just the same as a broken arm.
— David Baszucki, How I Built This with Guy RazWe had some people play it, but we knew within a week, oh my gosh, this is not going to go viral. We have a lot more work to do.
— David Baszucki, Reddit AMA, October 28, 2021Ha! I happen to play Roblox in both my free time and my work time.
For readers who want the conventional biography laid flat, a small shelf of short books now exists, including David Baszucki: Co-founder and CEO of Roblox - Building a Digital Playground for a New Generation and David Baszucki biography: The Man Behind Roblox.
Maxims
- Misuse is a message. When customers bend your product toward something you didn't intend, they are showing you the bigger market.
- Build the floor, not the furniture. A platform that stops making content stops competing with the people who make it better.
- Crude ships; perfect waits. The blocky avatar became iconic because it existed, not because it was sophisticated.
- Some ideas need the world to catch up. A decade between insight and launch is not failure if the infrastructure wasn't there.
- Currency is a promise. Hold the value of your unit steady and an economy can grow on top of it.
- Explain the spread before someone else does. Every take rate is defensible until it's described by a critic.
- Disorder compounds. Bureaucracy grows on its own; simplicity requires someone whose job is to destroy it.
- Calibrate the gut. Intuition is an organ, and organs can be in bad repair.
- Say it like a parent is listening. When the users are children, how you describe the problem is part of how you solve it.
- The room never really gets bigger. Whatever the scale, someone is still responsible for reading what comes through the queue.
