Part IThe Story
The menus were the problem, and then they were the solution. In 2015, in a rented two-story house in Kyoto so small that a single shoji screen, paper stretched over a wooden lattice, was all that separated the two bedrooms, a pair of software founders in their twenties had stripped their lives down to two activities. They wrote code, and they ate noodles. They did not bother to dress, clean, or cook. Neither man spoke or read Japanese. That should have made the second activity difficult. But one of them had grown up reading Chinese, and roughly seventy percent of the kanji on Kyoto's menus resembled the hanzi he had learned as a boy. That was enough to tell beef from chicken. It is a small thing, a partial literacy that gets you fed. It may also be the most important fact about him.
"We were just, code, code, code," Ivan Zhao would later say of those months, in a sentence that has since been quoted so often it has become a kind of company scripture. "Then, 'Hey, let's go out for food.' Then, we go eat, go back to work, and do it again." Downstairs were the laptops, eighteen hours a day. Upstairs, behind the paper wall, was sleep. On the other side of that wall was Simon Last, the co-founder who had followed him across the Pacific after the two of them laid off every other person at their company. The company was called Notion. Three years in, it had no product anyone wanted, a codebase they had just thrown away, and runway measured in the cost of rent. They had moved to Kyoto partly because it was less than half as expensive as San Francisco.
The math is funnier than the mythology lets on. They sublet their San Francisco apartment and their office, and so, Zhao has said with evident amusement, they were "actually making money living in Japan." A startup that had burned through most of roughly $2 million in angel money from friends, family, and friends of friends briefly ran at a profit by not being there.
A decade later Notion would pass 100 million users. It would raise $275 million at a $10 billion valuation, be run by roughly 1,100 people alongside more than 700 AI agents, and name its conference rooms after rice cookers. Zhao would be named Entrepreneur magazine's Entrepreneur of 2024 and profiled under the question of whether he is the best-dressed man in tech. None of that was visible from the downstairs room in Kyoto. What was visible was a boy from Xinjiang reading seventy percent of a foreign script, closely enough to order dinner. Not fluency. Not ignorance. Something in between, and workable.
The whole story lives in that gap. A man who belongs partly to several worlds learns to translate each into the others, and then builds a product that does the same: software that looks like a notebook and is secretly a programming environment, an idea about human augmentation dressed as a to-do list. He has a name for it, and the name is about vegetables.
By the Numbers
The Notion Arc
~$2MAngel funding raised in 2013 from friends, family, and friends of friends
4Times Notion was rebuilt during the 'lost years' before traction
$10MAnnual recurring revenue reached before Notion hired its first salesperson
50Employees on staff before Notion hired its first product manager
$10BValuation at the $275M Series C in October 2021
1,000 → 100M+Users, from the first milestone photo to the 2024 announcement
700+AI agents working alongside roughly 1,100 employees by 2026
Seventy Percent Legible
He likes to correct the premise. When Lenny Rachitsky introduced him in March 2025 as someone who grew up "in a small town in China," Zhao answered with the dry precision of a man who has heard the phrase too many times: "I think a small town in China, the definition, it's actually 4 million people." The town was Ürümqi, capital of Xinjiang, in the desert of China's far northwest, closer to Kazakhstan than to Beijing. It sits in what he calls "this Muslim part of China," where homes are full of rugs. Hold onto the rugs.
He was an only child, raised by his mother. When she decided he needed a better education, she took him to the capital. That is where the first act of translation happened, and it was bureaucratic before it was intellectual. A child from elsewhere could not simply enroll in a good Beijing school. "Because I'm from somewhere else," he explains, "in order to go into good school in the capital, you need to win some kind of competition." There were paths in mathematics and paths in programming, the Information Olympiad. He chose programming because he loved computer games and wanted to be at a computer all day. He placed second in Beijing. "Pretty big," he adds. "Beijing is a big city." He had learned to code in elementary school, in QBASIC, using it to solve math problems, and he would later note, with some understatement, that he never expected it to be so useful as an adult.
So the founding gesture of his technical life was a passport. Code was the credential that let an outsider from the periphery into the center. It was not a vocation yet.
At sixteen his mother moved him again, this time to Canada. Classroom English in China had given him grammar and exam technique. It had not given him jokes. "What you're missing is the context, the culture," he says. "So you have to watch SpongeBob or Simpsons to get a sense of humor essentially." He picked up a second language's comic timing from a sponge living in a pineapple under the sea. Every profile of him mentions this, usually as a cute aside, and it is cute. It is also a method. He understood early that knowing the vocabulary of a system is different from knowing what people inside it find funny, what they expect, what they will and will not tolerate. Years later he would make the same mistake with software and correct it the same way.
At the University of British Columbia he made a choice that looks idiosyncratic and turns out to be structural: he skipped computer science. "Did not study computer science since I already knew how to code," he says. He studied cognitive science, the discipline of how minds work, along with art. He took up photography to train his eye; in a 2016 interview with a Chinese tools site he said that taking photographs in college was how he developed his sense of aesthetics, that design was self-taught, and that "learning by making" was the best way to learn anything. Other accounts add that he had studied traditional Chinese watercolor as a boy. By graduation, he noticed something odd about his social life. "Most of my friends are artists," he says. "I'm the only nerd in my art friend circle."
The artists needed websites, as artists do, for portfolios. He built three or four of them and noticed something that the people inside either world might have missed. His friends had taste, often better taste than the engineers he knew, and almost no ability to work in the medium that was eating everything else. "People don't know how to create with the software media, computing media," he remembers thinking. Paint and film were open to them. Software was not. It came pre-assembled by strangers, sealed.
The kid who had used code as an entry ticket in Beijing was now watching people be denied entry. He could read both scripts. Most of his friends could read one.
The Paper That Made Him Swear
The second translation went backward through time.
Trying to build a creative tool for his artist friends, he did what an obsessive does and read the history of creative tools. That history kept dead-ending in the same place: Northern California, the 1960s and 1970s, in the labs around Stanford and Xerox's Palo Alto Research Center. "You realize the first generation of computing pioneers," he says, "they actually had the best ideas." He names them in a litany he has repeated across a decade of interviews: Douglas Engelbart, Alan Kay, Ted Nelson. For these men, he says, "there shouldn't be a separation between builders and users. It's the same medium."
Then he read Engelbart's 1962 paper, Augmenting Human Intellect. His reaction is recorded with a candor that no press office would have written: "When I read that paper, it's like holy shit. If you are making software, if you know how to code or design, this is the highest leverage thing you can do for other people."
Engelbart, a Navy radar technician turned Stanford Research Institute engineer, spent his career arguing that computers should amplify human problem-solving rather than merely automate clerical work. He is remembered mostly for the mouse, which is roughly like remembering Gutenberg for the font. Kay, a jazz guitarist and biologist by temperament, helped build the Xerox Alto and its environment, Smalltalk, in which nothing was a sealed application and everything was an object a user could open and change. Nelson coined "hypertext" and spent decades as the prophet of a web more deeply linked than the one we got. All three have in common an inheritance that the industry took the wrong half of. Zhao fell in love with the half that got left behind.
He tells one story about that loss with a heat he rarely shows elsewhere. When Steve Jobs visited Xerox PARC and saw the Alto, Zhao says, he saw the graphical interface: windows, icons, a mouse. He did not see the deeper idea underneath, the malleable object environment. The PARC engineers apparently tried to show him. "Steve Jobs say, 'I don't like this direction of scroll bar direction,'" Zhao recounts. "Then people just instantly change the scroll bar direction for him. That's the power of the original Smalltalk Alto system. He only saw the graphic user interface." The personal computing generation of Jobs and Bill Gates built its empire on applications, sealed boxes made by specialists for everyone else. "Then that has all the apps we have today," Zhao says, "and has the SaaS route we have today."
The man who would quote Jobs approvingly on Twitter ("there's just a tremendous amount of craftsmanship between a great idea and a great product") also holds him partly responsible for the shape of the industry Notion would have to fight. He is not alone among technologists in this complaint. What is unusual is that he turned it into a business plan.
After UBC he moved to San Francisco and took a job as a product designer at Inkling, an education publishing startup. The work taught him how software teams actually ship. In his spare time he was building the first versions of what would become Notion. In 2011, before Inkling, a young entrepreneur named Akshay Kothari had tried to hire him. Kothari was then running Pulse, a news-reading app that LinkedIn would acquire. Zhao turned him down. Remember the rejection. It will come back as an investment, then as a partnership, then as a man with a paint roller.
The company was founded in 2013 with Simon Last, a programmer whom Zhao found, by one account, through a small Twitter community of people doing strange and interesting coding projects. Last was a student at the University of Maryland. Zhao offered him an internship; Last dropped out to go full time. Last's biography, as much as it is public, is almost entirely about the work: an engineer gifted at avoiding premature optimization, the right instinct for a two-person company and nearly the wrong one for a hundred-million-user one. He shares Zhao's single childhood obsession. "Lego is the only thing he wanted for every Christmas," Zhao says. Zhao says the same of himself: "It's the only toy I ever wanted."
Two only-Lego boys, one from Ürümqi via Beijing and Vancouver, one a Maryland dropout, set out to build what Zhao calls "Lego for software." The original 2013 seed deck, which Zhao has since published as a public Notion page, described the company they would build. It estimated fifteen months. "It took us 10 years 😅," he wrote on LinkedIn in 2024, emoji included.
Sugar-Coated Broccoli
The first products were what you would expect from a person who had just had a religious experience reading Engelbart. They were developer tools, made easy enough, the founders hoped, that anyone could build software. Nobody wanted them.
Zhao describes the failure without self-pity and also without the usual founder spin. "We tried that a couple of years and learned that actually most people just don't care," he says. "The majority of people, they wake up, they have report due, they need to get their job done, they don't care creating software to optimize whatever they're doing. They don't care." They showed the product to friends and investors. "It did not resonate with people."
He has called the first three to four years of the company "the lost years," and the phrase has some truth beyond rhetoric. In his own count, they rebuilt Notion four times. One version, he wrote years later, had a feature where "you'd press a key to x-ray all blocks." He follows it with a parenthetical that every founder should tape above their desk: "(what were we thinking?!)"
What they were thinking was what Zhao wanted. That is his diagnosis. "The first version of Notion was more about what I wanted than what people wanted," he has said, and elsewhere, in an earlier interview: "We focused too much on what we wanted to bring to the world. We needed to pay attention to what the world wanted from us." The diagnosis is about maturity before it is about strategy. "Sense of maturation is you don't see the world just from your perspective but from outside your perspective," he says. "We were young. Took us multiple years. It hit your head straight into the wall to realize that."
The insight that came out of the wall was a sort of benevolent deception. If people would not use a tool for building software, perhaps they would use a tool they already used every day, one that happened to contain the power to build software if they went looking. Productivity tools. Notes, documents, task lists. The things a billion people already did on computers. He put it to Sequoia this way: "We realized, in order to be a big software tool, we have to start from a product that is already ubiquitous, that people recognize, and hide the tool-making, hide the programming ability, like how you sugarcoat broccoli. Nobody wakes up wanting to make software. Well, very few people."
— Ivan Zhao, on Lenny's Podcast, March 2025We call it sugar-coated broccoli. People don't want to eat the broccoli but people like sugar, so give them the sugar then hide the broccoli inside of it.
The metaphor is good, and slightly sly, because it describes the grammar of his whole life. A desert kid wins entry to Beijing by being good at the thing the gatekeepers measure. A Chinese teenager learns Canadian humor from cartoons. A programmer studies cognitive science and makes portfolio sites for painters. Each time, something that would be rejected in raw form gets wrapped in something the receiving culture already recognizes. It is not deception, exactly. Zhao is insistent on this point when he talks about marketing years later: "You can fool other people once or twice, then there's no future. It has to be actually tied back to something genuinely the value creating." The broccoli has to be real. You are only choosing the form it arrives in.
Knowing what to build did not mean they could build it. In the middle of the productivity pivot, the founders realized they had chosen the wrong technical foundation. At the time, before React won, Google was promoting Web Components, a standard that, Zhao says, "feels like a Lego-like, the building block-like." Of course two Lego obsessives bet on it. It was too new, and it was unstable. "It don't know where the bug come from," he says. "It's from your source code or from the underlying libraries?" The app crashed. Data was lost. One later account quotes Zhao describing the feeling at the bottom as "kind of like despair... I almost never had that feeling before."
The sentence is worth setting beside something else he says about himself, more often and with more conviction: "I'm lucky and Simon lucky that high is never too high, low is never too low for us." Both are probably true. Equanimity is not the absence of despair. It is what you build after you have felt it once.
A Wall Made of Paper
So: the layoffs. Notion was then five people. "The layout I brought back to me and Simon, two people," Zhao says, transcription errors and all. "And morale obviously there was really low. You have to say goodbye to your teammates." In a short Chinese-language interview in 2016, a year after it happened, he listed founding his own company as the turning point of his career and then described, in two flat sentences, the decision that nearly ended it: the previous version of Notion wasn't stable enough, so he laid off everyone at the company and moved with his co-founder to Kyoto to rewrite it from the beginning. There is no adjective in it. That is how he talks about the worst year of his working life.
Then there was the money. "Chinese mom always can help," he says, "and I'm a single child." His mother had helped from the start. As a Canadian he needed to register a company to move to the United States, and she helped him raise the first money. He paid her back. When the money ran out again, he asked to borrow it a second time. "Hey mom, can I borrow that just to bridge us?" Several accounts put the loan at $150,000. Nearly a decade later, announcing 100 million users on LinkedIn, he gave her a parenthetical: "Notion almost died (thanks for the bridge, mom!)."
The woman who moved her son from Ürümqi to Beijing so he could get into a better school, then from Beijing to Vancouver so he could get a better education, then wired money into a dying San Francisco startup so he could keep chasing a 1962 research paper, appears in his public accounts only in fragments. She is the least documented and possibly most consequential investor in a $10 billion company. Her investment thesis, as far as anyone can tell, was the same one three times: move the boy to where the door is.
Japan had always been at the top of the founders' list of places to go. "Let's just go somewhere that we've never been to change the scenery a little bit," Zhao remembers. Kyoto was a pragmatic choice that looks poetic in retrospect, a city defined by craft traditions, temple carpentry, ceramics, lacquer. He would tell The Verge in 2025 what he took from its craft culture. In 2015 he was mostly taking advantage of its rents.
What emerges from his descriptions of those months is not suffering but something closer to rapture. "That's some of the happiest moments," he says. "Birthday every day." He and Last could read each other's thinking: "You know what each other other people are thinking and you can just cross through the problem space really quickly. The technical product space, design space, and just non-stop of shaping stuff." Two people, one paper wall, no meetings. The interpersonal overhead he would spend the next decade trying to engineer out of a growing company was zero.
The new codebase got a name that tells you how its authors felt. "Hence our main git repo is 'notion-next,'" Zhao wrote in 2024. The repository that runs a company serving more than 100 million people is still named for the version that was supposed to come after the failure.
Kyoto gave him something more durable than a codebase: a theory of resets. "I think it's not just a self-help way to say don't be afraid to reset," he says, almost impatient with the bumper-sticker version. "The more interesting point here, it's like you can create progress through better abstractions. And that thing compounds faster, can catch up to all the things you build much quicker than you ever thought." Humans think linearly, he argues; abstractions compound. Smalltalk, he notes, had a tiny kernel inspired by Lisp. "The kernel of things could be really small, but just like math. It can compound." If you find the right primitive, three years of sunk cost can be recovered in months. That is not a feeling. It is a bet on mathematics.
Notion 1.0 launched in March 2016: a clean, block-based tool for notes, documents, wikis, and tasks. A paragraph was a block. An image was a block. A checkbox was a block. You could drag them around. The broccoli was in there, hidden.
The Small Bus
Traction did not arrive as a moment. Zhao is almost stubborn on this point, resistant to the narrative every interviewer wants from him. "It never hit us as a binary state," he says. "Just like, 'Oh, good. We have people who care about this thing we make now. Oh, good. People reach out to us who are paying us.' And it's a very gradual ramp." By his own account, it was not until 2018, with the Notion 2.0 launch on Product Hunt, that the company saw real signs of traction. Notion 2.0 introduced databases, the brick that would let users assemble their own project trackers, CRMs, and content calendars. That March, the Wall Street Journal reviewed it as perhaps the only app you need for work and life.
The sign that something had changed was not revenue. It was a box of dog treats. Notion had brought a dog into the office and someone had posted about it on Twitter. One day treats arrived for the dog, addressed to an office whose location wasn't public. An investor had tracked the company down, the dog included. "Why do people want this so much?" Zhao remembers wondering. "So that was a moment I paused a little bit and I guess there's enough attraction for investors." Did they take that investor's money? "Not the first time."
What happened next is the most underrated episode in the company's story, mostly because nothing happened. Zhao started taking the meetings. Venture capitalists were knocking; it felt good to be wanted after years of indifference. Then one of his existing investors stopped him. Her name is Shana Fisher, a New York investor whom Zhao describes, affectionately, as "another therapist." Her intervention, as he remembers it, was brutally simple:
— Shana Fisher, as recalled by Ivan ZhaoIvan, what are you doing? You clearly don't need money. You're just trying to feel good to do external validation about this.
"And I said, 'Oh, that's so true,'" Zhao says. "It doesn't help us make a better product and the truth is with what customer tell us. Then we just went back to building. I went back to hardcore building, no meeting modes."
That is the origin of what became one of Notion's defining characteristics, an almost perverse leanness at a moment when the industry regarded headcount as proof of ambition. Notion did not hire a salesperson until it had passed $10 million in annual recurring revenue. It did not hire its first product manager until it had fifty employees. By 2025 it had been profitable for years, and when Rachitsky suggested that most of the money it had raised was still sitting in the bank, Zhao nodded. The company tracks what it calls talent density: revenue per employee rather than raw headcount. "People want to work with either more talented people," he says. "So it's a positive company group."
Asked whether this was trauma, the scar tissue of a company that nearly died of running out of money, he rejects the premise. "I wouldn't say we're, Notion, is a cost saving-first company. I like fancy chairs. I like furniture." The leanness, he argues, comes from taste in problem-solving: "I just believe better system is much better than brute force through people." Notion is itself a tool for building tools, and the company runs almost entirely on Notion. "We use the same mindset to build our company," he says. "And accidentally, that keep our headcount low."
The internal metaphor is a bus. "The smaller the bus, it's easier to turn corners, easier to accelerate, easier to maneuver," he explains. "As a leader in the company, you decide who sit around you on the bus seats." It is a mundane image, the kind of thing you could find on an inspirational poster, until you remember the original bus held two people separated by a paper screen.
In 2018 the bus took on a passenger who had been trying to board for seven years. Akshay Kothari, who had tried to hire Zhao in 2011 and invested in Notion's 2013 angel round, sold Pulse to LinkedIn, rose there to run LinkedIn India, and then joined Notion as COO. He became the third co-founder. The detail that his story always comes back to is humility: he started, by one account, answering support tickets, thousands of them. Kothari's arc rhymes with the others in this story. An immigrant who built a consumer app on the strength of an elegant interface, sold it, and then chose to start again at the bottom of someone else's queue. The trio settled into recognizable roles: Zhao the vision and the eye, Last the engineering co-creator, Kothari the business builder.
There is a photograph, which Zhao posted in 2024, of Kothari painting a wall in the company's third office. The color is Benjamin Moore OC-97, "Cream Froth." It was chosen after, as Zhao puts it, "we tried 30 shades of warm white paints for our office wall." The COO of a company soon to be valued at billions, holding a roller.
Thirty Shades of Warm White
The paint is not a quirk. It is the argument.
Notion's offices are famous in the way that restaurants are famous: for an atmosphere that turns out, on inspection, to be the product of fanatical specification. For a long time there was a no-shoes rule. Everyone ate around a single table. The furniture is mostly home furniture, not office furniture. The ceilings have no overhead lighting, because, as Zhao puts it with the flat certainty of a theologian, "top light is evil." Most offices, he says, look like hospitals: "The white is so pale and the floor is so dark." The goal is a space that feels like an artist's studio or a home.
He offers two justifications, and their ordering is revealing. The first is pragmatic: a cozy space makes people stay longer and feel more creative. The second is not pragmatic at all. "At least personally for me, it hurts the eyes if you just see ugly things," he says. We accept that a bad chair hurts your back, he argues; why not accept that an ugly chair hurts your eyes? "If the chair looks ugly, the wall looks ugly, it hurts. So it's better to not have thing that hurts." On another podcast he compressed the same idea: "You don't want to hear noise, but you also don't want to see ugly things." An Eames chair costs more, he concedes. It is comfortable on your body and comfortable on your eyes.
Talk to him long enough and the sensibility shows up everywhere, sometimes comically. On the Kleiner Perkins podcast Grit he described touching tables in other companies' offices to identify the material: laminate, decent, not the fanciest. He describes himself as "a burrito kind of guy" who would rather have a burrito than a Michelin tasting menu, which sounds like a contradiction of the fancy-chairs man until you see that it is the same preference: simplicity, clarity, everything in one place. He prefers offices with a patina of use to shiny ones, and has been summarized as warning that when an office is too nice, the business is doomed. He says he can feel a bug before he can name it. Shortly before that interview, he noticed that the page title field in Notion responded more slowly than the body text, and reported it.
In the 2016 Chinese interview, asked to recommend one tool from his life, he skipped software and named a shoe: the Dr. Martens three-eye. It goes with everything, he said, dress up or dress down; one less thing to think about. He owned about a dozen pairs. A decade later, when Brian Halligan asked his opinion of Halligan's outfit on Sequoia's podcast, Zhao answered before the joke could land: "Your pants could be looser." (He liked the sneakers.) Asked where to shop in New York, he named La Garçonne in SoHo-Tribeca. The man who learned English from cartoons is now, in at least one publication's estimation, a candidate for best-dressed in tech.
His wife, he says, refers to him as a wood cabinet builder. He seems to treat it as the highest compliment available. "Can I make this wood cabinet more beautiful and more useful and feels nicer on your hand?" he says, describing his mindset. "You have aesthetic direction towards it and you have your technical know-how to actually make things happen. Then you need to do permutation and trade-off in your head." That, to him, is craft, and Notion's internal philosophy is literally called "craft and values." Craft is skill and taste. Values are how you see the world.
The conference rooms are where the craft becomes a curriculum. They are named after timeless tools: the iPhone, the original Macintosh, various chairs, the Lamy 2000 fountain pen, the Sony transistor radio, the Toshiba rice cooker. The rice cooker is the one he likes to explain. It "changed how people eat rice in Asia for a hundred million, tens of a hundred million people." These objects lasted for decades, and software, he notes, cycles every six or twelve months. "I care creating something that at least the form factor lasts longer than 18 month." That is the modesty of a cabinetmaker, not a visionary: not forever, just longer than the industry's attention span.
When the company couldn't find merchandise it loved, it designed its own work jackets. Even the swag is a cabinet.
— Ivan Zhao, on Lenny's Podcast, March 2025Too much of yourself, then there's no users. Then you're just doing our project. You're just doing a research project. And too much for a business, you're building a commodity.
Organ Rejection
The hardest lesson in Notion's middle years did not come from a competitor. It came from the company building something that was a little too much like everybody else's.
Project management had become one of Notion's most important uses. The fastest way to compete with dedicated tools was to hard-code their concepts into the product: sprints, milestones, the furniture of agile software development. Notion did. The alternative, the Notion way, was to work out what a sprint actually is in primitive terms ("clusters of a task that group together," Zhao says) and build a new general-purpose brick that would let users assemble sprints themselves. That way is harder and slower. They took the faster road.
It failed on every axis at once. Customers found the feature underpowered compared with specialized competitors and awkward alongside the rest of Notion. Engineers complained that there was now a part of the codebase that felt foreign: "That's your hardcore too much into it." And it simply felt wrong to use. Zhao's description of what happened next borrows from medicine: "There's organ rejection with your employees, with your customers." If you build the Lego way, he says, "the system work for you. If you're building non-Lego way, the system work against you."
What is striking is how long it took him to see it. "That took nine months, a year to realize sometime," he says, and he places the drift at a year to a year and a half of shipping, as he puts it, "not according to our value, not according to my value." In a 2024 LinkedIn post he alluded to "some hard lessons on building at scale (story for another time)" before declaring, with obvious relief, that the product was re-accelerating: seven releases that summer, with "a quality we can be proud of." In 2025 he said that 2024 was the year he could finally say Notion shipped "good stuff at good velocity and good quality and align with our values."
The sprint episode is not the dramatic crisis. That honor belongs to COVID, which delivered Notion the blessing and curse of every collaboration tool in 2020. The company had passed one million users right before the pandemic, a milestone Zhao still remembers by date. Then usage exploded. Simon Last's instinct against premature optimization had served the company well: for years Notion ran on a single instance of a Postgres database, upgraded each time to a beefier machine. Then there were no beefier machines.
"So there's a doomsday clock," Zhao says, "that when we're going to truly run out of this space to store everything in Notion and Notion got a complete shutdown." The company stopped building new features. Nearly every engineer moved to the problem, which was sharding: splitting the database across many machines. The clock ran somewhere between three and six months. Near the end it was weeks, and the database had begun behaving unpredictably as it approached its limits: "You really don't know which day going to hit you." They made it. The lesson, characteristically, came out as a corrected maxim rather than a war story: "Don't do premature optimization but plan ahead a little bit. Don't go late."
Read the two episodes side by side and a pattern emerges. In one, the company optimized too early for the market and violated its own grain. In the other, it optimized too late for its own scale and nearly drowned in success. Both are failures of timing, of knowing which wall to build and when. Kyoto taught Zhao to reset. These episodes taught him that the reset is not a one-time event. It is maintenance.
Bricks and Boxes
He does not recommend what he did. Asked for advice to founders building horizontal software, tools that try to do many things instead of one, Zhao answers immediately: "First, I wouldn't recommend it." Would he do it differently himself? "I wouldn't do it differently myself, but I wouldn't recommend it."
The joy and the suffering of a horizontal product are, in his telling, the same fact viewed from two sides. A vertical product has a natural wall: the market for dental scheduling software is the market of dentists. "The advantage of building horizontal, there's no wall," he says. "But then you need to create a wall yourself." Notion competes, in principle, with all of software. In practice, a customer must be able to understand what Notion is for.
The lesson arrived through his favorite metaphor turned against him. "Only hardcore Lego fans care about Lego bricks," he says. "Most people care about Lego boxes." The box has a picture on the front and instructions inside; it promises a pirate ship. Enterprise customers, he learned, buy boxes. There was a word he had to learn for this: "solutions." "You need to be a solution for enterprise customer," he says, "you need to sit somewhere on a P&L." He admits he stayed too long in what he calls "the Lego brick mindset." At an Australian event in 2025 he reportedly named his biggest lesson as not bringing in sales early enough, in markets like Sydney and Melbourne, where by his estimate about ten percent of people were already using Notion.
The bricks found a way in anyway, through a strategy Notion calls B2C2B. Individuals adopt Notion for the simplest thing a computer does, writing notes, and discover, slowly, the databases, the relations, the views. Then they bring it to work. Half of Notion's business customers, Zhao says, were previously personal users, mostly people who started with notes. It is broccoli as a go-to-market motion. When Rachitsky surveyed some 6,500 of his readers in 2025, Notion ranked second in project management behind Jira and third in CRM behind Salesforce and HubSpot, a CRM it never intended to build. "What is a CRM is relational database," Zhao says. "That's why we give people that brick." The logic of the top of the funnel also explains the newest products: which categories have a billion users? Documents. Calendars. Email. Notion shipped a calendar in 2024 and was building email in 2025.
He frames all of it through a book he loved as a boy. The classic Chinese novel Romance of the Three Kingdoms opens with a line he quotes from memory: "Empires long united must divide, long divided must unite." Software, he argues, follows the same rhythm. Paper was a universal medium. The PC era of the 1980s splintered it into countless applications. (He notes, with a historian's delight, that dBASE launched as "dBASE II" to seem established.) Microsoft re-bundled everything into a suite in the 1990s. SaaS unbundled it again; at Make with Notion in 2024 he cited companies juggling more than 88 tools. "It's madness," he says. The pendulum, in his reading, is swinging back toward the bundle.
A teenager in Ürümqi reading a fourteenth-century novel about warlords grew up to use its first sentence as a theory of enterprise software procurement. That kind of connection is what he means when he tells founders to look outside tech. "If you go back to history, there's a massive amount of patterns and shapes and trade-offs you can steal from," he says. Asked for a recommendation, he names a 1970s BBC documentary series: James Burke's Connections, which strings together inventions across centuries to show how one domain inspires another. He describes his Enneagram as a 7, "creative, finding connection, see the forest and tree," with an 8, the Challenger: "competitive." The type has a certain neatness to it. A connector with a competitor's wing.
The community completed the circle. At the 2024 conference, Zhao described monks in Thailand running a monastery on Notion and Toyota using its Kanban boards. More important to him are the creators, non-programmers who package their expertise as Notion templates and sell them. Some made more than $1 million in 2023. "We dreamed this in the original pitch deck," he wrote, "but I wasn't sure it would come true." It never gets old, he says, to walk past a coffee shop and see someone using it. The artist friends who couldn't build their own websites in Vancouver now had descendants selling software they built without code on Etsy.
A New Kind of Wood
In 2023, at a company offsite in Cancún, Ivan Zhao got early access to GPT-4. It was, in Sequoia's telling, the second refounding of his career.
He had not been asleep before then. Notion launched an AI writing alpha in 2022, before ChatGPT made the category famous. But GPT-4 was a different kind of shock. Accounts of the Sequoia conversation describe his reaction as near-total conviction: this changed everything, and if the company did not bet on it, nothing else mattered. He announced a reset across an organization that by then numbered roughly 500 people. The host, Brian Halligan, gave him a name: "the refounder." His advice to calcified CEOs, as Sequoia summarized it, was to "feel the AGI first, then trust your body when it tells you to move."
The reset was not triumphant at first. By his account the next eighteen months or so were hard: the underlying models were not yet mature, the team tried nearly every direction, growth slowed, morale suffered. Only as the models improved did the products take off. The first was the writer. The second was Q&A, searching and answering across a company's workspace, which led to a realization he describes as a gift he never expected: "AI is really good with bundled offerings. AI is really good with horizontal tools." All the information in one place, it turns out, is what a language model needs.
The third realization is the one that makes him swear again, the way Engelbart did. The great weakness of Lego, he says, has always been that "not everybody can put together a Lego set from scratch." The bricks need assemblers. "But guess who is really good at piecing things together, assemble things?" Coding models. "So now we're looking at holy shit, we spent the last five, six year building all those Lego blocks for knowledge work. If we're just putting AI coding agent on top of it, you can create any kind of knowledge, customer software, customer agent for whatever your vertical use cases you need." Notion 3.0, released in September 2025, introduced agents that could act inside the workspace.
The broccoli, it turns out, can now cook itself. The dream that ordinary people would build their own software, abandoned in its raw form in 2015 because nobody cared, has returned through a machine that does the part nobody wanted to do.
He thinks about the technology in a cabinetmaker's vocabulary. "I always feels like AI language model feels like a new type of wood," he says. "It feels like aluminum." Mass air travel had to wait for cheap aluminum; computers had to wait for semiconductors. New materials unlock new trade-offs, but the other side of the equation stays fixed. "As a human, ever since we got out of Africa, we're set, right? That's a constraint. It's invariable." In December 2025 he published an essay, "Steam, Steel, and Infinite Minds," extending the material metaphor into a theory of how intelligence becomes cheap and how organizations change around it.
The company is being reorganized around the same idea. "We want to be a jazz band, not a marching band," he told Halligan, describing a culture where people improvise around a shared structure instead of executing in formation. Hierarchy, he believes, is human nature and cannot be flattened away; what can change is how the band plays. He offers a formula for talent: capability times taste times agency. Capability, he argues, is depreciating, the way the ability to find information depreciated after Google. Taste and agency are not. By 2026 Notion was running more than 700 AI agents alongside roughly 1,100 employees. He is hiring young people, and he is hiring, as he told Rachitsky, "misfits." "If you're exceptional at many things especially," he said, "then DM me."
In October 2025, a video went online under the title "How to build tools that shape civilizations." It shows a conversation between Ivan Zhao and Alan Kay. The man whose Smalltalk let PARC engineers change a scroll bar's direction while Steve Jobs watched, and the man who has spent a decade insisting Jobs missed the point, finally in the same frame.
— Ivan Zhao, on Lenny's Podcast, March 2025Tools are extensions of us. And once they extend us, once we shape them, once we bring them to world, they can come back to shape us.
What Travels With You
He quotes the line about tools shaping us often, attributing it to Marshall McLuhan, the way most people do. Then he asks the follow-up most people skip: what part of human nature are you extending? Sequoia, he notes, famously talks about investing in the seven deadly sins, because businesses that latch onto them are powerful. "But at least I prefer to amplify creativity and beauty in the domain of software," he says. Lego is creativity. Lego is beauty. "Software to me feels like lacking both."
In 2016, asked which developers he most respected, he said there were too many to choose from, like sometimes wanting Sichuan food and sometimes wanting Cantonese dim sum. Then he named three: Alan Kay, Rembrandt, Newton. Only one of them wrote software. It is the list of someone who reads three scripts imperfectly and refuses to give up any of them.
He has said that he finds his best work environments in coffee shops and in bed. He has said that whenever he feels down, he goes to sleep, and the next day he is reset. He has said the thing he is proudest of is having made "something authentic to myself" that "happen to be also useful for others." His life motto, when Rachitsky asked, was barely a motto at all: "I like to think things as a craft. You just make it better. Make for yourself. If it's unique enough for yourself and useful for others, things will follow."
The office has changed many times. San Francisco, then Kyoto, then San Francisco again, through office after office, each warmer and whiter than the last, painted in cream after thirty samples. Through all of it, one object came along. It was a rug from his hometown in the Muslim part of China, where people use a lot of rugs, red and blue, laid on the floor of the very first Notion office and carried to every one after. It was there through the lost years and the four rebuilds, through Kothari's paint roller and the doomsday clock, through the sprints that the organism rejected and the offsite in Cancún. It was on the floor while the company grew from two people behind a paper screen to eleven hundred people and seven hundred agents.
About six months before he described it on Reid Hoffman's podcast in 2026, Notion moved into a new office in San Francisco. The rug did not move with it. Ivan Zhao rolled it up and took it home.
"Yeah," he said. "It's still clean."
Part IIThe Playbook
What follows is an attempt to extract operating principles from Ivan Zhao's career, not as commandments but as patterns visible in his decisions, reversals, and stated reasoning. Several of them pull against each other. That tension is the point: much of Zhao's skill lies in holding contradictory instincts in the same head and choosing between them deliberately, week by week, rather than settling permanently on one.
The principles
- 1.Hide the vision inside something people already eat.
- 2.Reset toward a better abstraction, not just a fresh start.
- 3.Steal from history, not from the feed.
- 4.Refuse validation you don't need.
- 5.Substitute systems for headcount.
- 6.Treat taste as an operating input, not a decoration.
- 7.Build with the grain of your own architecture.
- 8.Delay optimization, but not past the cliff.
- 9.Build a wall where the market doesn't give you one.
- 10.Make the free user the sales team.
- 11.Treat new technology as a new material.
- 12.Learn the skill each stage of the company demands.
N
The Notion Timeline
Thirteen years, three refoundings, and a pitch deck that stayed true.
2013
Zhao and Simon Last found Notion Labs in San Francisco; raise roughly $2M from angels including Akshay Kothari.
2015
Web Components bet fails; team laid off from five to two; founders relocate to Kyoto and rewrite from scratch.
2016
Notion 1.0 launches in March, block-based notes, docs, wikis, and tasks.
2018
Notion 2.0 adds databases; Product Hunt launch brings first real traction; Kothari joins as COO.
2020
Passes 1M users just before COVID; single-Postgres "doomsday clock" forces an all-hands sharding effort.
2021
$275M Series C at a $10B valuation.
2022
AI writing alpha ships before ChatGPT's public debut.
2023
Early GPT-4 access at a Cancún offsite triggers a company-wide AI reset.
2024
Passes 100M users; first in-person Make with Notion; calendar ships; product craft "re-accelerating."
2025
Notion 3.0 introduces agents; Zhao publishes "Steam, Steel, and Infinite Minds."
2026
More than 700 AI agents work alongside roughly 1,100 employees.
Principle 1
Hide the vision inside something people already eat
Notion's first two years failed because the founders presented their ideology directly: a tool for building software, offered to people who, in Zhao's phrase, "have report due" and "don't care." The vision did not change. The packaging did. Notes and documents were the most ubiquitous activity on a computer, so the programming power went underneath them, discoverable but not required.
The subtlety is that the broccoli must be real. Zhao is explicit that marketing which only deceives "can fool other people once or twice, then there's no future." The productivity suite was not a bait-and-switch. It was a genuinely useful product that happened to contain a larger one. Users who never found the databases still got a good notes app. Users who did found something closer to a programming environment. Both groups got what they came for.
This is also how Zhao himself moved through the world, competing in the contest Beijing measured, learning the humor Canada spoke. The form adapts to the receiver. The substance does not.
Tactic: Write down your vision in one sentence, then identify the most ubiquitous existing behavior it could live inside, and launch as that behavior first.
Principle 2
Reset toward a better abstraction, not just a fresh start
Founders are told not to fear starting over. Zhao finds that advice shallow. His argument is mathematical: progress built on a better abstraction compounds, while humans intuitively expect it to accumulate linearly. A reset that finds a superior primitive can recover years of sunk work in months, which is why throwing away the code in 2015 was rational rather than desperate.
The Kyoto rewrite did not just change frameworks; it committed to blocks as the universal unit. That single abstraction later absorbed databases, a calendar, AI writing, and agents without another full rewrite. The repository name, "notion-next," is a reminder that the reset was meant to be the foundation, not a patch.
Resetting without a better abstraction is just churn. The test is whether the new kernel is smaller and more general than the old one, the way Smalltalk's tiny Lisp-inspired core produced complex behavior.
Tactic: Before any rewrite, name the single primitive the new version will be built on, and check that it can express everything the old system did with less code.
Principle 3
Steal from history, not from the feed
Zhao's founding insight came from a 1962 research paper and a 1970s computing environment, not from a competitor. His strategic framework for bundling comes from the opening line of Romance of the Three Kingdoms. His conference rooms are named after a rice cooker and a transistor radio. His viewing recommendation is a BBC documentary series from the 1970s.
He is explicit that this is a method: tech culture, he says, is "so focused on the now," while history holds "a massive amount of patterns and shapes and trade-offs you can steal from." The advantage is structural. When everyone reads the same feeds, they converge on the same ideas. Ideas imported from other centuries and other fields are, almost by definition, uncrowded.
The Xerox PARC story shows the opposite failure. Jobs took the visible half of a great idea and missed the deeper half. Zhao made a company out of the half that was left on the table.
Tactic: For your next major product decision, find one precedent at least thirty years old or from an unrelated industry, and write down what it would recommend.
Principle 4
Refuse validation you don't need
When investors started sending dog treats, Zhao started taking meetings, until Shana Fisher pointed out that he didn't need money and was seeking "external validation." He went back to "hardcore building, no meeting modes," and only later took outside capital.
This principle is easy to state and hard to practice, because validation feels like progress after years without it. Zhao's corrective was to locate truth in a single place: "the truth is with what customer tell us." Investors' enthusiasm is a lagging signal of the same thing customers tell you directly.
It also helps to have someone positioned to say it. Zhao calls Fisher a "therapist," a word he also applies to Rachitsky. The role matters more than the title: someone trusted enough to name the vanity in a decision.
Tactic: Before accepting any meeting with an investor, journalist, or prospective partner, write one sentence on how it improves the product, and decline it if you cannot.
Principle 5
Substitute systems for headcount
No salesperson until $10 million in ARR. No product manager until fifty employees. Profitable for years with most raised capital unspent. The usual explanation, frugality born of near-death, is one Zhao rejects: he likes "fancy chairs." The real reason is a belief that "better system is much better than brute force through people," and that the main cost of headcount is the overhead of aligning minds.
The method has three parts. Hire people who can do several jobs, designers who code and engineers who think about product, so trade-offs happen inside one head instead of across meetings. Run the company on your own tool, so every process becomes a system rather than a role. Measure revenue per employee instead of headcount.
The small bus metaphor gives managers ownership: each leader chooses who sits next to them, and the seat choices determine how fast the bus turns.
B
The Small Bus vs. the Big Bus
How Notion's staffing choices diverge from venture-scale convention.
| Conventional scaling | Notion's approach |
|---|---|
| Hire sales early to prove go-to-market. | No salesperson until $10M ARR. |
| Hire PMs to coordinate engineering. | First PM at fifty employees; multi-skilled builders instead. |
| Headcount signals ambition to investors. | Revenue per employee as the core metric. |
| Raise to extend runway. | Profitable; most raised money unspent. |
| Buy best-of-breed internal tools. | Run the company on the product itself. |
Tactic: Before opening a new role, spend one week trying to replace it with a system or tool, and hire only for what the system demonstrably cannot do.
Principle 6
Treat taste as an operating input, not a decoration
Thirty shades of warm white. No overhead lights. Conference rooms named for timeless tools. A CEO who touches tables to identify their material and feels bugs before naming them. It is easy to read this as personal indulgence. Zhao frames it as two things at once: a pragmatic choice that makes people stay and create, and a values choice. Ugly things "hurt," the way bad chairs hurt your back.
Taste works operationally in three ways. It attracts and retains people who share it. It sets the quality bar for the product, since a company that agonizes over wall paint will agonize over input latency. And it lengthens the time horizon: the rice cooker reminds the team to build forms that last longer than eighteen months.
His talent formula, capability times taste times agency, makes the point explicit. As AI commoditizes capability, taste becomes a larger share of what a person contributes.
Tactic: Pick one physical or digital surface your team touches daily and redesign it to a standard you would be proud to show a craftsperson.
Principle 7
Build with the grain of your own architecture
The hard-coded sprints feature was the fastest path to competing in project management, and it produced what Zhao calls "organ rejection" from customers, engineers, and the product itself. "If you build in a Lego way... the system work for you. If you're building non-Lego way, the system work against you."
The deeper lesson is that a company's architecture encodes its values. Shortcuts that violate the architecture also violate the values, and both employees and customers can feel it even when they cannot name it. Notion took nine months to a year to recognize the problem, and longer to clean it up.
The right question for any feature is not "what does the competitor have?" but "what primitive would let users build this themselves?" It is slower. It compounds.
R
Three Refoundings
Each reset returned the company to its core abstraction.
| Moment | Trigger | Response | Type |
|---|---|---|---|
| Kyoto, 2015 | Unstable stack, no product-market fit | Lay off team; rewrite on blocks | Survival |
| Sprints, ~2023 | Hard-coded features rejected | Return to "Lego way" of building | Values |
| Cancún, 2023 | Early GPT-4 access | Company-wide AI reset | Technology |
Tactic: For every feature on your roadmap, ask whether it can be built as a reusable primitive, and require written justification for any that cannot.
Principle 8
Delay optimization, but not past the cliff
Simon Last's discipline against premature optimization let Notion run on one Postgres instance for years, saving engineering effort for the product. Then COVID turned that saving into a doomsday clock. All feature work stopped, and the company came within weeks of a total outage.
Zhao's correction is precise: "Don't do premature optimization but plan ahead a little bit. Don't go late." The principle is not about choosing between speed and robustness. It is about knowing where the cliff is. Systems near their limits behave "sporadic," so the last weeks are worse than they look on a chart.
The broader pattern is that a startup's virtues become its vulnerabilities at scale. The instinct that saved it at five people can kill it at five hundred.
Tactic: For each critical piece of infrastructure, estimate the date it will hit hard limits at current growth, and begin the migration when that date is two growth-doublings away.
Principle 9
Build a wall where the market doesn't give you one
A horizontal product has no natural market boundary, which is both its promise and its curse. Customers, especially enterprises, buy "Lego boxes," solutions with a picture on the front that "sit somewhere on a P&L," not loose bricks. Zhao admits he stayed in the brick mindset too long and brought in sales too late.
The answer is to build walls deliberately in positioning and go-to-market while continuing to build bricks underneath. "Solutions" became one of his favorite words to rally sales and product. The danger he names is building the box at the expense of the bricks, the sprints mistake again.
Zhao does not recommend horizontal products at all. If you build one anyway, accept that segmentation is your job, not the market's.
Tactic: Define three named "boxes" for your product, each with a specific buyer and budget line, and make sure every new primitive strengthens at least two of them.
Principle 10
Make the free user the sales team
Notion's B2C2B strategy starts with individuals using it for the simplest computing task there is, notes, and lets them discover its depth until they bring it to work. Half of Notion's business customers were previously personal users. The same logic drives expansion into calendar and email: categories with billion-person top-of-funnel use.
The community extends the effect. Early on, Zhao personally answered users on Twitter and email. Template creators, some earning more than $1 million in a year, turned Notion into an economy. Monks and Toyota engineers became case studies. The product's flexibility makes every power user a potential evangelist, because each one has built something nobody else has.
The model works only if the free experience is genuinely good. The sugar has to taste like sugar.
Tactic: Identify the single most common daily behavior your product supports, make it excellent and free for individuals, and instrument the path from personal use to team adoption.
Principle 11
Treat new technology as a new material
Zhao describes language models as "a new type of wood," like aluminum, which made mass air travel possible once it was cheap. New materials change what trade-offs are available, but human nature, the constraint "ever since we got out of Africa," stays fixed. The product-maker's job is to find the sweet spot between the two.
This framing led to concrete strategy. AI turned out to favor bundled, horizontal tools, because models need context and Notion already held it. It also solved the Lego problem: coding agents can assemble bricks for users who will never assemble them alone. A decade of primitives became, unexpectedly, the ideal substrate for agents.
The material lens also guards against hype. Zhao has said the AI agent "still hasn't arrived" in its full form. A craftsperson learns a new wood's grain before promising furniture.
Tactic: For each new technology you evaluate, list which existing trade-offs it removes and which human constraints remain unchanged, and design for the overlap.
Principle 12
Learn the skill each stage of the company demands
Zhao says Notion becomes "like a new company" every twelve to eighteen months and requires a different skill from him each time. He is good at one-to-one storytelling, but the company needed a one-to-many storyteller, a role he finds draining and rarely performs. He treats it as "a craft," like a skill in a video game "you need to pick up something to unlock something else."
The same humility shows in his willingness to name his mistakes publicly: building what he wanted, not what people wanted; the brick mindset; late sales; shipping against his own values for a year and a half. Each became a skill acquisition rather than an identity crisis.
He calls it "an infinite game," and says infinite games are more fun.
Tactic: Every twelve months, write down the one capability the next stage of your company most needs from you that you currently lack, and schedule deliberate practice in it.
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
— Ivan Zhao, on reading Engelbart's 'Augmenting Human Intellect,' Lenny's Podcast, March 2025When I read that paper, it's like holy shit. If you are making software, if you know how to code or design, this is the highest leverage thing you can do for other people.
— Ivan Zhao, LinkedIn, on Notion passing 100 million usersEventually we learned. But I am glad we didn't cave and change our mission. Notion is built on the 70s' vision that software can "augment human intellect." It's amusing that our original pitch deck still describes the company we are building today!
— Ivan Zhao, via Twitter/XInstead of only focusing on 'build' and 'something', I wish more builders would also think about 'people'. Human nature is the invariable anchoring all businesses, and it hasn't changed since we left the plains of Africa.
— Ivan Zhao, Sequoia CapitalIn the next 5 to 10 years, Notion could be the front-end infrastructure for the world. Just dream of a piece of software, and you should just be able to build that using Notion.
— Ivan Zhao, Lenny's Podcast, March 2025I like to think things as a craft. You just make it better. Make for yourself. If it's unique enough for yourself and useful for others, things will follow.
Maxims
- Sugar-coat the broccoli. Keep the vision intact, but deliver it inside a form people already want.
- Reset toward abstraction. A better primitive compounds fast enough to recover years of sunk work.
- The truth is with the customer. Investor enthusiasm feels like progress; customer behavior is progress.
- Systems before seats. Every role you replace with a tool is overhead you never have to align.
- Ugly things hurt. Treat visual and tactile quality as seriously as ergonomics.
- Build with the grain. Features that violate your architecture will be rejected by the organism.
- Plan a little ahead. Delay optimization, but begin the migration before the doomsday clock starts.
- Sell boxes, build bricks. Customers buy solutions; the primitives underneath are what last.
- Empires unite and divide. Read the bundling cycle before choosing between vertical and horizontal.
- Steal from outside tech. The least crowded ideas come from other centuries and other crafts.
