Part IThe Story
Configure, make, make install
There was one computer in the house in Lanús, and it ran two operating systems that did not get along. The family had a DSL line, 128 kilobits per second, and on Windows it simply worked. The internet service provider had shipped everything Windows needed. On Linux nothing worked. The driver for the point-to-point connection, PPPoE, lived on a CD-ROM as source code, and source code had to be compiled, and compiling it required knowledge the boy did not yet have. The people who had that knowledge were online, in an IRC channel of Argentine Linux enthusiasts, and he could only reach them from the operating system he was trying to leave.
So he made the trip. Boot into Windows. Get online. Ask the channel what to type. Memorize the answer. Reboot into Linux, where there was no connection and no way to check, and type it from memory: configure, make, make install. When it failed, and it usually did, he rebooted into Windows and asked again. Years later, recounting this on a developer podcast, Guillermo Rauch used a word that most engineers would have dropped. He called the commands incantations. That was what they were to a child in the southern suburbs of Buenos Aires, carrying them across a reboot in his head like water in cupped hands.
Every reboot cost several minutes, and every memorized command was a chance to get one character wrong. There was a wall between where the answer lived and where it had to run, and the only way across it was a boy's short-term memory.
Hold that image next to a figure from the summer of 2026. In an interview after Vercel's ShipNYC conference, TechCrunch reported that the company Rauch founded processed six million deployments a day, and that half of them were triggered not by people but by coding agents. More than a trillion tokens flowed through its AI gateway daily. Nobody in that system memorizes anything. The incantation is generated, carried, executed, and verified, and the answer and the place where it runs have nearly become the same thing.
The distance between those two scenes, a dual-boot PC in Lanús and a cloud that deploys itself, is about twenty-five years. Rauch spent that time removing steps, and he has been unusually consistent about which ones. He went to a selective public high school and did not finish it. He wrote an education company's software before he could legally drink. He built a framework, gave it away, and then built a cloud underneath it to sell. By 2025 he had become, depending on which Argentine newspaper you read, the country's fifth or sixth unicorn founder, its Entrepreneur of the Year, and briefly the most boycotted CEO in developer infrastructure. Along the way he kept losing the thing he loved, which was writing code, and kept counting each loss as progress.
By the Numbers
The Vercel Machine
$9.3BVercel valuation after its $300M Series F, September 2025
6MDeployments per day by mid-2026, half triggered by coding agents
1T+Tokens flowing through Vercel's AI gateway daily
500MNext.js downloads in the 12 months to October 2025
1.9%Share of all websites hosted on Vercel, per W3Techs
11Rauch's age when he received his first paycheck for online work
10thHis rank among thousands on the entrance exam to the Carlos Pellegrini
The language of the manuals
The father came first. In every telling, Rauch's father is the one who brings the machine home. He was an industrial engineer who, in his son's words, "was really into engineering in general and Star Trek, so he always wanted to buy new, cool things for the family." Rauch's mother was a chemical engineer. Infobae reports that both parents were trained at the University of Buenos Aires. Rauch was about seven when the computer arrived, and he still remembers watching Windows 95 boot for the first time. He has called that the moment "it all started."
The father did not stop at delivering the box. "My dad challenged my brother and me to explore what we could do with it," Rauch told First Round. "Could we install software? Could we hack it? He had this very experimental mindset about it." When the boy turned out to love video games, the father reframed the hobby in a sentence Rauch repeated to Clarín decades later, in the Buenos Aires slang that never left his speech: Te gusta jugar videojuegos, con la programación podés hacer los tuyos. You like playing video games. With programming you can make your own.
That sentence turns a consumer into a producer. Rauch built a career, and later a company, around performing the same turn for millions of other people.
The materials were in English, and he did not speak English. The early web had very little Spanish on it and less about compilers. So he learned the language from the manuals, as GV and First Round both report, and so his English and his programming came from the same source. He did not translate code from Spanish. He acquired one technical vocabulary that happened to include a natural language. Decades later he would build a product whose central promise is that you can make software by typing English sentences. The Acquired episode about it was titled "Building Web Apps with Just English and AI." For him, English had always been a programming interface.
A Red Hat Linux CD, by his own account to Eric Ries, changed the direction of his life. After it came the Lanús Linux community, which went by the name Lanux. Then the IRC channels, then forums. A photograph that ran with a Between the Wires interview shows the boy beside Richard Stallman, founder of the free software movement, with a caption that is very much a programmer's joke: Guillermo, 12, explaining to Richard Stallman that
vi > emacs. Arguing with Stallman about text editors at twelve is a fair summary of the temperament: respectful of authority's existence, unconvinced by its conclusions.The money started early, and it came from abroad. Rauch told Infobae that at eleven he was building pages for something MercadoLibre then called Mercado Socios, an affiliate program that paid you if your websites sent it traffic. On My First Million he described the opening that followed. A forum user, whose real name Rauch says he'll never know and who went by "Dark Shadow123," noticed how much the boy enjoyed writing tutorials and helping strangers. "There's this website, it's a freelancing website," Dark Shadow123 told him. "You could just sell your services here because you know so many things about Linux and PHP and programming." Rauch got his first check at eleven. By twelve or thirteen he had a client in the Netherlands. He was, he says, "helping my parents with our home finances."
Shaan Puri asked the obvious question: was he pretending to be an adult, or telling clients he was eleven? Rauch's answer trails off in the transcript. The point stands either way. Dark Shadow123 occupies a small but structurally important place in this story, as the first of a series of strangers who looked at the boy's public work and saw an asset. Nobody was mentoring him. The work was simply visible, and someone could find it. Every important door Rauch walked through afterward opened the same way.
Some of this was about craft and some was about something else. Rauch told the DevJourney podcast that what hooked him on programming was how short the path was from writing a file to running GCC to holding a binary. "I could actually extract so much excitement from these few characters on a black screen," he said. Programming is mostly failure, he explained, "so much negative feedback involved in the process, that the victories have to get you extremely excited. It's the little things — like a test passing with a bunch of green dots on the screen."
Notice how much of what he loved was about the feedback loop. Write, run, see. Green dots. He was learning to love a loop, and the rest of his career is an argument for making that loop shorter.
A pibe of seventeen
The Colegio Carlos Pellegrini is one of Argentina's elite public secondary schools, free and selective, with an entrance exam that sorts thousands of applicants. Rauch prepared hard. "I worked so hard to get in," he told My First Million. "Entered in position number 10 out of thousands of students." He bristles at the label that later attached to him. "I've never been a fan of the high school dropout moniker because I actually really loved the high school that I went to." He told Clarín that the Pellegrini exposed him to things "that maybe in another universe I would have learned at university," and that when he got to the United States he was already "well rounded." He said that phrase in English, inside a Spanish sentence.
He loved the school, and he left it. "I had two competing interests," he said, and the other one was winning.
That other interest started with a small, ugly problem. He was building a website for a music catalog. It had rows of entries, each with three buttons on the right: Inspect, Edit, Remove. He wanted Remove to delete the row on the client side without reloading the whole page, which in the mid-2000s was harder than it sounds. His solution was to post to a hidden iframe and listen for its callback event. "Later on I discovered that this iFrame thing is hack," he told Between the Wires. The search for a better way led him to MooTools, a JavaScript library for intermediate and advanced developers. At sixteen he was named a core developer.
The same year, he told Refactoring's Luca Rossi, he released a small open-source project called Fancy Menu, and a popular blog featured it. It was his first taste of global attention. More followed. "I would create these demos that would make it to the front page of Digg and Reddit," he told First Round. "I was becoming a five-minute internet celebrity every time I would create a really cool demo using these frontend technologies."
Then came the Swiss episode. A startup in Lausanne had bet its entire front end on MooTools and wanted experts. Aaron Newton, another MooTools core developer who was consulting for the firm, recommended Rauch. Newton never met the boy in any way that mattered. He knew him through commits and forum threads, the way open-source contributors know each other. "I think this is why it's so important to have people that bet on you early on," Rauch said. A week later the company flew him to Switzerland. He remembers the CEO meeting him at the train station, and the transcript catches the CEO's first reaction mid-sentence: "wait." To Clarín, Rauch supplied the rest of the joke: they had no idea that one of the co-authors "was a seventeen-year-old pibe from Argentina."
That "wait" sums up the arrangement Rauch would exploit for the next decade. His work arrived before he did. When the person showed up, he was younger, browner, more foreign, and less credentialed than the work had suggested. By then the work had already made the argument.
First Round reports that Facebook came calling around then, interested in hiring the young phenom. He did not finish the secundario. He spent time in Europe, then landed in San Francisco at eighteen to take his first full-time job as a frontend engineer. On his personal site he lists the sequence plainly: joined the MooTools core team, got a job at eighteen, moved to San Francisco. He owes much of his career, he writes, "to the Web and Open Source."
— Guillermo Rauch, to First Round ReviewOpen source really was my path out of Argentina in the end.
He arrived with no network. "As a Latin American founder in the U.S., Guillermo didn't have a built-in network in Silicon Valley," GV's Tyson Clark wrote in 2020. Rauch did not dwell on what he lacked. He described what replaced it. "One thing that helped me was open source. People were happy to talk to me because I'd built things that piqued their curiosity." And on the city itself: "I really fell in love with the Bay Area's openness to entrepreneurship and new ideas. I was here to start something new. I made a lot of connections with immigrant entrepreneurs who showed me it was possible." Speaking to a startup newsletter years later, he called himself "someone from Argentina who came to the U.S. with virtually nothing" and gave San Francisco credit for being "willing to bet on me."
That word "bet" appears again and again in his accounts: Aaron Newton bet on him, San Francisco bet on him. He never describes himself as having earned things. He describes being wagered on, which is how a person talks when he knows he could have been passed over.
An education company, founded by a dropout
The first company was LearnBoost, which built software for teachers, including gradebooks and classroom tools that ran in the browser. Its founder had not finished high school and had not gone to college. Rauch noticed the irony himself. "I was really passionate about building an education product," he told Between the Wires, "but as I just narrated my own story, I didn't finish high school and I didn't go to college."
The product mattered less than the stack it ran on. LearnBoost, Rauch's site says, "was one of the earliest adopters of Node.js in production," at a time when running JavaScript on a server struck most engineers as a stunt. The team contributed to projects that became infrastructure for a generation of web developers: Express.js, Connect, Jade, Stylus. Rauch himself wrote Mongoose, which his site says "remains the most popular way of accessing MongoDB data in the JavaScript ecosystem." He also wrote Socket.IO.
Socket.IO deserves a pause, because it shows where Rauch's instincts pointed. It is a library for real-time communication. Before it, most of the web worked on request and response. After it, a browser and a server could hold a live conversation. InfoWorld covered its readiness for real-time apps in June 2014. Its underlying engine, Rauch's site notes, "powers Notion's realtime sync and the first Coinbase trading product." The boy who used a hidden iframe to delete a single row without reloading the page had grown up and built the plumbing that made reloading obsolete.
In 2012 he published Smashing Node.js: JavaScript Everywhere with Wiley, one of the first books on the platform. Its premise, in Rauch's own description, came from "identifying the potential for JS to run on both server and client." The subtitle reads like a prophecy. It also describes his own life. A person who learned one language from manuals and decided it would do for everything is well suited to argue that one programming language should run everywhere.
LearnBoost led to Cloudup, a file-sharing product. Rauch's site says it was built on "realtime infrastructure for MongoDB I authored inspired by Firebase and a proprietary predecessor to Electron for cross-platform Web-based desktop apps." Later accounts describe a progression from the sprawl of an education platform to the single gesture of file sharing. Drop a file, get a link. Automattic, the company behind WordPress, acquired Cloudup in 2013, according to Rauch's site, "to power their editing and site building technology." When he spoke at WordCamp San Francisco in 2014, the program listed him as "former CTO and co-founder of LearnBoost and CloudUp," a speaker at O'Reilly OSCON, QCon and NDC, and a mentor in an Open Source Engineering class "organized and pioneered by Stanford, with students from Harvard, MIT, Carnegie Mellon, UPenn, Columbia and others."
So the dropout was teaching Stanford's students after all, through the one channel that had never asked him for a diploma.
He described the Automattic exit modestly on Jack Altman's Uncapped: "it's nice to have an exit." He spent time inside a company that had built a sustainable business on open-source software, and he watched how it worked. That lesson stuck with him.
A set of side projects rounds out these years and says something about his temperament. He wrote
wifi-password, a bash script that retrieves the password of the network you're on. It has more than 4,600 stars on GitHub. He wrote slackin, which made public Slack communities easy to join and has 6,500 stars. He wrote ms, a tiny utility for converting time formats that, according to his site, is among the smaller packages that "have accumulated billions of downloads." He also co-built what his site calls "a turn-based, multi-player virtual machine running a copy of Windows XP that went viral, inspired by Twitch plays Pokemon."That last one is worth a second look. The boy who had one computer and a household full of competing claims on it grew up to build a single shared Windows machine that thousands of strangers took turns operating. He has never explained it that way, and it was surely just fun. Still, the image fits him: a shared operating system, open to anyone on the internet.
Operations we perform without thinking
In November 2015 Rauch founded a company called ZEIT, German for "time." The name was close to a thesis statement. Ten years later, in an essay on his personal site, he explained the frustration behind it.
"Vercel was born out of my frustration in 2015 that while the cloud enabled this seemingly infinite array of possibilities (and compute), the pixels on the web weren't getting significantly better. And neither was the experience of crafting them."
He had been an early user of Amazon Web Services and remembered the wonder of it. "Computers were now in the cloud. It was magic," he told First Round. "Hardware became software. It's so abstract, the idea that there are robots moving hardware around in racks somewhere." The magic did not last. The cloud had promised to lift the burden of physical data centers, he wrote, "but ultimately we inherited much of that burden in digital form." Then came the list, which reads like a eulogy: "DevOps, K8s, VPCs, CI/CD, IAM, CDNs, IaC..."
— Guillermo Rauch, 'The AI Cloud,' October 17, 2025The time you should have spent on the pixels... was spent on the configs.
Every acronym on that list stood for a step between the idea and the URL. Each one was a reboot into a different operating system to fetch an incantation and carry it back. At a WebExpo conference in Prague in 2018, Rauch's talk was called "Config-less. Server-less. Effort-less." Its abstract quoted the mathematician Alfred North Whitehead: "civilization advances by extending the number of important operations which we can perform without thinking about them." It is the most honest summary of his program he has offered. He did not want to make developers smarter. He wanted them to be able to think less about the parts that should not require thought.
The vehicle was Next.js, a framework built on Facebook's React library. His site says he "designed and co-authored" it. It was released in October 2016 and introduced at the ViennaJS meetup and React Conf. Next.js made the hard parts of building a React website, such as server-side rendering, code splitting, and routing, into defaults. Its organizing principle is what Rauch calls "progressive disclosure of complexity": a project can start with almost no code and grow, as its needs grow, into something capable of running the largest websites in the world. You see complexity only when you go looking for it.
He gave Next.js away. Tara Tan's retelling, in her Founder Stories series, has friends asking him how he could build a business by giving everything away. Whatever the exact conversation, the logic was sound. A framework is a habit. If millions of developers learn to think in Next.js, they will want somewhere to deploy Next.js, and the company that wrote the framework will know better than anyone how to host it. Vercel's answer, he later wrote, was to be "the React of the Cloud." React had established that a webpage is an output of state and data, not something you edit by hand. Vercel would treat cloud infrastructure the same way: "the hyperscaler primitives are outputs not to be directly manipulated, or jQueryed." Turning a JavaScript library's name into a verb meaning "to fiddle with by hand" is a joke only a certain kind of person would make, and he made it in a corporate manifesto.
The look of the company mattered from the start. In 2016 Nicolás Garro, a designer known online as Evil Rabbit, joined ZEIT as its founding designer and head of design. Garro is also Argentine, and like Rauch he is known by his work and his handle before his name. The company's mark was a black triangle. It was clean and austere, pointing up, like a deploy arrow without a stem. The triangle shows up in the old talk listings as "▲ZEIT." When ZEIT renamed itself Vercel in April 2020, the triangle stayed. The Register's headline on the rebrand was "News sure to ex-Zeit." The pun was terrible and the observation was right: the time company had become something else.
Rauch also built a terminal. Hyper, a cross-platform terminal emulator made with web technologies, reached what his site counts as more than 42,000 daily active developers. Asked whether Hyper came from those childhood hours at the black screen, he said: "Definitely. Hyper.app to me is sort of a continuation of that idea." The man who had found excitement in "a few characters on a black screen" built himself a nicer black screen, and then spent the next decade building a company whose purpose was to mean you would rarely need to open one.
The engineering principles he later posted to LinkedIn, "learned over the years at Vercel," read like a working credo. Some are borrowed: "Make it work, make it right, make it fast." Some are blunt: "Most 'flukes' aren't… your tech just sucks." One is surprisingly tender: "Red is not enough to signal 'error' (8% of men have red-green color blindness)." One is close to melancholy: "You're never done. Software ages like milk, not wine." The obsessions in that list are feedback, latency, copywriting, and the user staring at the screen. It is the list of someone who once waited through a reboot to learn whether he had typed a command correctly, and who has resolved that nobody should have to wait like that again.
Who bets on a kid from Lanús
The capital came from people who already knew the work. Business Insider's 2020 profile carried a long headline: "The 29-year-old founder of Vercel used this pitch deck to raise $21 million from investors like Accel and GitHub's CEO to build faster websites." The individual investors on Vercel's site form a revealing list. Nat Friedman ran GitHub, the website where Rauch's reputation had been built one commit at a time. Jordan Walke created React, the library Next.js was built on. Brendan Eich created JavaScript itself, the language Rauch taught himself in Lanús and then bet his career on.
So the man who wrote the framework on top of React, in the language Eich invented, hosted on the platform Friedman ran, was backed by all three of them. Each of those investors had once built an abstraction layer. Each was now paying the next person to build one above theirs. The cap table works as a family tree.
The list continues: Naval Ravikant of AngelList, Karri Saarinen of Linear, Rasmus Andersson, who designed the Inter typeface, Benjamin De Cock, an early designer at Stripe, Alexandr Wang of Scale. Rauch had a heavy concentration of designers among his angels. The company sold infrastructure to engineers and recruited its earliest believers from people who cared about how pixels looked.
GV led the Series B, announced in December 2020. Tyson Clark's write-up has a sentence that every subsequent account echoes in some form: "Guillermo's energy is off the charts." In the same piece Rauch predicted that the job title "front-end developer" might someday be renamed "product developer," and that designers would increasingly code. He called that broadening "ultimately what will make our society more equitable, more interesting, more artistic, more creative. Giving everybody the opportunity to create products." That sentiment is consistent across a decade. Around the time ZEIT still went by that name, he gave Between the Wires a more specific version. His dream, he said, was that the next Facebook or Snapchat would be created by someone who had not had to go through all the education and gather all the connections. "Really, it can be one girl in Africa. It can be a boy in Bangladesh."
He did not say a boy in Lanús, though he hardly needed to.
In June 2021 Vercel raised $102 million in a Series C. The Argentine press responded with something between pride and bafflement. Rauch told Clarín that Vercel had just raised "US$102 millones con un valor de mercado de US$1.100 millones." El Diario Sur, the local paper of the southern suburbs, called him the creator of the country's fifth unicorn. Clarín called it the sixth. The miscount says a lot about the coverage. Argentina was eager to claim him and not quite sure how. By November 2021 the valuation was $2.5 billion. In May 2022 Infobae caught him on a brief visit home to speak at an Endeavor event and ran its piece under the headline "Unicornio desconocido," the unknown unicorn. "I specialized in front-end engineering," he told the paper, "but I'm not an engineer. I never studied. I'm 100% self-taught." The piece noted, with what reads like a sigh, that he had left the country and never come back.
He supplied Clarín with a sentence that could serve as Vercel's mission statement for the pandemic years: "Vercel transgrede dos narrativas de la modernidad: el rol de la educación mayor o universitaria y el de la oficina centralizada." Vercel transgresses two narratives of modernity: the role of higher education, and the centralized office. Both narratives were his own life. He had no degree, and his first important job came through people who knew him only as an open-source contributor. The company was, in its early years, deliberately distributed. "Open source is a very distributed, decentralized system, especially with git," he told The Big Bet in 2023. "The distributed model has been at the core of everything we do."
One more board appointment closes this circle. In March 2026 Vercel added Mitchell Hashimoto to its board. Hashimoto co-founded HashiCorp and created Terraform, the dominant tool for describing infrastructure as code. In Rauch's essay, "IaC" is one of the acronyms that stole developers' time from the pixels. Hashimoto had also built Ghostty, a fast terminal emulator, which echoes Hyper. Vercel's founder had once listed the configuration tools he hated, and he put on his board the man who had made the best of them. That is either generosity or a quiet admission that the configs never really go away. They just move to a different floor.
Ego death
In 2023 Vercel released v0, a tool that turns natural-language prompts into working web applications, and the AI SDK, a toolkit for building streaming conversational interfaces in JavaScript and TypeScript. v0 won a 2025 Webby Award for developer tools. By the time Sequoia's Training Data podcast recorded him, it had three million users, and Rauch reported that ChatGPT had become Vercel's fastest-growing channel for customer acquisition. The AI was sending people to the platform where AI-written software went to live.
On My First Million in May 2025, the first words out of his mouth were a renunciation.
— Guillermo Rauch, My First Million, May 28, 2025I've been an engineer for a couple of decades now, and I no longer write code. I only prompt.
The Notable Capital newsletter, writing up a closed-door session he held with portfolio founders, said that the thing he "kept coming back to throughout the conversation was ego death. He loved coding and built his career around it. He created frameworks that millions of developers use every day. And he still stood in front of his company to say that the future of coding is generation." The phrase fits him precisely. It is easy to find founders who disrupt other people's crafts. It is rarer to find one who announces, in front of the employees who chose his company because of that craft, that the craft is ending.
He had been circling the subject for a while. On Uncapped he described what programming had given him as a child: "It taught me how to focus. It taught me to be disciplined. It taught me to receive this negative feedback from the compiler and overcome it." Then he added, "I do think, you know, we'll need to find what the next version of that is because I don't think it's going to be programming necessarily." The compiler that made him had been a strict teacher, and he was saying publicly that it might not teach the next generation.
The numbers moved quickly. In May 2024 Vercel raised $250 million at a $3.25 billion valuation. On September 30, 2025, it raised $300 million in a Series F co-led by Accel and Singapore's sovereign fund GIC, at $9.3 billion. Forbes, as relayed by the Argentine press, put Rauch's personal fortune at no less than $2.1 billion. The Argentine site Mosca reported in April 2026 that revenue had grown from $100 million to $340 million in a single year, and that an IPO was in view for 2027. On CNBC in March 2026, Rauch told Jon Fortt that February had been a record growth month, driven by coding agents such as Claude Code that generate deployable software. Forbes, again via the Argentine press, described him gathering his San Francisco team when Anthropic released a new model and calling it "un gran momento para el mundo."
The essay from October 2025, "The AI Cloud," records both the success and the instruction to forget it. Next.js had been downloaded 500 million times in the previous twelve months, "more than in the entire 2016 to 2024 period." Vercel was crossing one billion application deployments. It powered two of the twenty highest-traffic sites on the planet, OpenAI and The Weather Company. "But to win the next decade," he wrote, "Vercel must heed the advice of its hyperscaler predecessor and partner: it's always Day One." He named three shifts: from pages to agents, from problems to solutions, from closed to open. He framed the choice in stark terms: "We can either give up the web, proxy all our data and intent through one closed mega-app, or reinvent it together in the open."
Inside the company, the ego death was being turned into practice. By May 2026, Notable reported, roughly 30 percent of Vercel deployments were made by coding agents, up from the low single digits a few months earlier. Two months later TechCrunch put the figure at half. Teams using Claude were deploying 7.5 times more often than teams that didn't. Vercel's data team no longer answered data questions. It maintained the agent that answered them for anyone who asked in Slack. Support engineers curated an agent that resolved 90 percent of tickets. On Outset Capital's 10 Minutes or Less, Rauch described a sales team shrinking from ten people to one person managing a bot. He has stopped asking his executives "What's the answer?" Now he asks, "What's the agent that can get me this answer?"
He was also candid about where he had been weak. Talking to TechCrunch about Eve, Vercel's framework for writing an agent's instructions and skills in natural language, he described a saleswoman who works on growing existing accounts and who used to wait a full quarter for a new dashboard so she could see which customers were adding seats. "On the R&D side, we're the fastest-moving company in the world," he said. "But on the sales engine, the Salesforce engineering [side], I was so incompetent. I had never opened Salesforce in my life when I started." The agent gave him a way into the parts of the company where his skills had never applied. The engineer who could not run a sales organization found he could build one.
He calls the management philosophy "recursive founder mode": "finding people who are founders at heart and making them the CEOs of their own products within the company." He set up a separate unit, Vercel Labs, because the company's reputation for curated, high-conviction products had become, in Notable's words, a constraint, and "by the time you're certain, you're already late." He told Ali Rohde that teams in AI might need to work even more than the proverbial eighty hours, because you have to be "terminally online." "I work all the time because I enjoy it," he said. The same conversation turned to his mustache, which he trims "as thin and clean as possible" and which, he admitted with some amusement, had become his recognizable symbol "kind of overnight." "It is my trademark."
A man who strips everything else away, removing steps, configs, and even his own coding, keeps one crisp line drawn above the lip. It is the last bit of hand-finished detail in a business he has otherwise automated.
No opina de política
In April 2026 the Argentine site Mosca published a profile with a sentence the author meant as praise. "Guillermo Rauch no es un nombre que circule en la conversación pública argentina. No sale en la tele. No opina de política." He doesn't go on TV. He doesn't opine on politics.
Seven months earlier, on Monday, September 29, 2025, Rauch had attended a private technology meeting in New York with Israeli Prime Minister Benjamin Netanyahu and a group of American tech executives and investors. According to the Israeli outlet CTech, the meeting covered artificial intelligence and its applications in Israel's economy and defense. Rauch posted a selfie with Netanyahu on X. Middle East Eye reported that he framed the meeting around AI education and "keeping our free societies ahead." Tech for Palestine quoted the caption as saying he was "optimistic for peace, safety, and greatness for Israel and its neighbors."
The post appeared nearly two years into the war in Gaza. A UN commission of inquiry had characterized Israel's conduct there as genocide, a finding the Israeli government rejects. The International Criminal Court had issued an arrest warrant for Netanyahu. Gaza's health ministry counted roughly 66,000 dead. The backlash was immediate and loud, much of it from developers in the same open-source world that had given Rauch his career. Within hours people were posting that they were migrating their projects to Netlify, Cloudflare, Fly.io, Render, Hetzner.
Amjad Masad, founder of Replit, moved fastest. Masad grew up in Amman and taught himself to program, and he is another immigrant founder who built a company on the promise that anyone can make software. On the same Monday he posted instructions for importing a Next.js project from GitHub into Replit. "For companies moving lots of work," he added, "happy to help + discount." A Vercel engineer posting as @wesamo__ wrote: "I just handed in my resignation at Vercel." A pseudonymous crypto founder known as Loopify, a Vercel customer who had raised more than $2 million in aid for Gaza, said his ventures would leave and called the post "so dumb," accusing Rauch of "purposefully going out of his way to harm his company." Some in the Next.js community argued that the framework should be separated from Vercel's brand. There were defenders. Viktor Bunin, a protocol specialist at Coinbase, wrote that Rauch "did the right thing because his post goes against 'the current thing.'"
Rauch's response, as reported, was brief. Per Tech for Palestine's account of a Bloomberg interview, he said the backlash had not cost Vercel customers or employees and that he was "sorry for anyone who felt hurt." Other trackers note that he disputed the scale of the exodus. Tech for Palestine endorsed the boycott and built a migration guide. The campaign found new material months later in a terms-of-service update taking effect March 31, 2026, under which, the group said, Vercel could use code and agent chats from Hobby-tier users to train AI models unless those users opted out.
It would be easy to turn this episode into a morality tale, and many people did. The more useful observation is structural. For twenty years Rauch had treated his public output as his credential. The Fancy Menu demo, the MooTools commits, the Socket.IO repository, the tweets and essays and conference keynotes all did his introductions for him. He arrived in Switzerland after his reputation had, and he walked into San Francisco through doors his code had opened. A person who has spent his life being judged by what he publishes does not easily develop the reflex of staying quiet. The X post was simply more output. Like everything else he had shipped, people took it as a statement of who he was.
Infrastructure also works differently from consumer products. A developer's choice of hosting platform is, in a small way, a statement about identity. Rauch had spent a decade cultivating that loyalty: Next.js as a habit, Vercel as a home, the triangle as a badge. The same loyalty turned out to be revocable. The lock-in he was accused of building proved, in the moment of the boycott, to be about one GitHub import away from not existing. Masad's instructions took three lines.
Months later, Vercel co-signed an open letter on "Open Weights and American AI Leadership." Rauch posted that "open weights are the logical next frontier." The letter is political in its own way, a position on how AI should be governed and by whom. It drew mostly approving replies.
He doesn't opine on politics. The sentence was true in Argentina, and only in Argentina.
Roblox cheat scripts
On April 19, 2026, Vercel disclosed that unauthorized actors had accessed certain internal systems. The chain of events, as the company and security researchers reconstructed it, reads like a parable about removing steps. The cybersecurity firm Hudson Rock traced the origin to Lumma Stealer, an infostealer that infected a computer when its user tried to download cheat scripts for Roblox. The compromise reached Context.ai, a third-party AI tool. Through it, the attacker got into a Vercel employee's Google Workspace account. Tom's Hardware's headline explained the core problem: the employee had granted the AI tool unrestricted access to Google Workspace. From there the attacker reached, in Vercel's words, "some Vercel environments and environment variables that were not marked as 'sensitive.'" Someone claiming affiliation with ShinyHunters offered the stolen data on BreachForums for $2 million. The post was later deleted, and the group reportedly denied involvement.
Consider the whole sequence. A kid looking for an advantage in a game downloads a script. The script carries malware, the malware steals credentials, and the credentials open an AI integration that someone had connected with broad permissions because broad permissions are frictionless. Every link in the chain saved somebody a step, and together they added up to a breach.
Rauch had seen this kind of risk coming, at least in outline. Talking to TechCrunch about Vercel Sandbox, a product that "put[s] the agent in a little cage," he described a conversation with the president of Airbus about the decades of specialized C++ aerospace code that could leak into a training set if "someone comes in and installs the wrong developer tool and boom, all the code goes out to the cloud." On CNBC in March he warned of a wave of "shadow IT," employees building tools with agents faster than their IT departments could keep up. He said it would be far bigger than the shift to mobile and cloud. In the same interview he revealed that Vercel had recently shut down a network of North Korean operatives who were using the platform to run fake AI job interviews aimed at engineers. In November 2025 he had tweeted that two of three famous AI products he used that day had full outages, and the third had hung. "You should architect agent products with the assumption that things will fail(over)." One of his engineering principles had always said the rest: "Most 'flukes' aren't… your tech just sucks."
None of this undoes the story. It complicates it. Rauch's career has been one long project of moving thought out of human heads and into systems, of making important operations things we do without thinking about them, as Whitehead put it. Systems that run without thought fail without thought too. The boy carrying commands across a reboot at least knew what he was carrying. An agent with unrestricted access to a Google Workspace does not.
He seems to know this. He calls Vercel's core value "iterative greatness," and the word "iterative" concedes that greatness is never finished. Software ages like milk. The cloud gets rebuilt every year, so it's always Day One.
In 2025 the consulting firm EY named Rauch Argentina's Entrepreneur of the Year and sent him to Monaco to represent the country in its World Entrepreneur of the Year competition, alongside more than fifty other national winners. When he shared a Forbes Argentina interview on LinkedIn, the comments filled with Spanish: Crack. La cabra. Qué orgullo. One commenter, Francisco Cesare, wrote from Lanús Este that what he loved was that Rauch always remembered where he came from: "en tu presentación primero el ORIGEN." First, the origin. Rauch's own site does exactly that. Before Vercel, before Next.js, before Socket.IO, the second sentence reads: "I'm originally from Lanús, Buenos Aires, Argentina."
He loved the school and he left it for Switzerland at seventeen. When a Clarín reporter caught up with him in 2021, the paper recorded, with what reads like affection, that the founder of a billion-dollar company had never received his secundario diploma, because he still had two subjects left to complete.
History. And Portuguese.
Part IIThe Playbook
What follows is an attempt to extract the operating logic from Rauch's career. These are not his stated values, which are well documented. They are patterns visible in his decisions: things he did repeatedly, sometimes against advice, sometimes against his own interests and inclinations. Some are tactical. Some are closer to temperament. A few are cautionary, because a playbook that includes only the wins is a brochure.
The principles
- 1.Collapse the distance between idea and URL.
- 2.Let the work arrive before you do.
- 3.Give away the habit, sell the home.
- 4.Make infrastructure an output, not an input.
- 5.Disclose complexity progressively.
- 6.Narrow the product until it fits in a gesture.
- 7.Sweat the copy, the color, and the error code.
- 8.Perform your own ego death before the market does.
- 9.Ask for the agent, not the answer.
- 10.Manufacture founders inside the company.
- 11.Architect for the failure you've abstracted away.
- 12.Remember that your public output is part of the product.
Principle 1
Collapse the distance between idea and URL
Almost every product Rauch has built aims at one quantity: how long it takes to go from intention to verified result. As a child it was reboot time, the minutes spent crossing between operating systems with a memorized command. Later it was the page reload that Socket.IO eliminated, then the DevOps acronyms Vercel absorbed, and finally the act of writing code, which v0 replaced with a sentence. The My First Million summary distills it into one line: reduce "the time between a concept and a live URL from weeks to seconds."
The insight that makes this more than a slogan is that iteration speed compounds. A developer who can see a result in two seconds runs a hundred experiments while a developer who needs an hour runs two. In Rauch's 2026 data, teams deploying with Claude shipped 7.5 times more frequently. This is not a productivity tweak. It produces a different kind of organization, because a fast loop changes which ideas are worth trying at all.
He also treats the loop as an emotional system, not just a technical one. "The victories have to get you extremely excited," he said of programming, because the rest of the time the compiler is telling you no. A fast loop gives you more of those victories. More green dots keep people working.
Tactic: Measure the time from "I have an idea" to "a stranger can see it" for your core workflow, and treat every minute in that interval as a defect to remove.
▲
The Shrinking Loop
Each phase of Rauch's career removed a step between intention and result.
c. 1998
Windows 95 boots on the family PC in Lanús; the loop is a reboot and a memorized command.
2007
Named MooTools core developer at 16; replaces a hidden-iframe hack with client-side interactivity.
2012
Publishes Smashing Node.js; one language on client and server.
2014
Socket.IO matures; browser and server hold a live conversation, no reload.
2015
Founds ZEIT in November; deployment becomes a single command.
2016
Next.js released in October; rendering, splitting, routing become defaults.
2023
v0 and AI SDK launch; the code itself becomes optional.
2026
Half of 6 million daily deployments triggered by agents; the human leaves the loop.
Principle 2
Let the work arrive before you do
Rauch had no degree, no network, and no American passport. What he had was a public record. Dark Shadow123 found his tutorials. Aaron Newton found his MooTools commits. The Lausanne startup flew him to Switzerland on Newton's word, and the CEO at the train station said "wait" when he saw how young the expert was. By then it did not matter. "People were happy to talk to me because I'd built things that piqued their curiosity," he told GV. "The content I shared with the world always created new opportunities for me."
This is more specific than "build in public." Rauch's public work was useful to the people who found it: a library they could install, a tutorial that answered their question. It made the case for him without his being present, in a format that skipped every gatekeeper he could not have passed. Credentials summarize what you have done. Open-source work lets people inspect it directly, so they don't have to take anyone's summary on trust.
The lesson matters most to people who start out with the least. Rauch's dream of "one girl in Africa" or "a boy in Bangladesh" building the next Facebook is a generalization of his own route: if the gatekeepers will not let you in, ship something they can download.
Tactic: Before you ask anyone for an opportunity, publish one artifact that solves a real problem for the person you're asking, and let it make the introduction.
Principle 3
Give away the habit, sell the home
The question Rauch's friends asked about Next.js, how you build a business by giving everything away, has a precise answer in Vercel's structure. The framework is free, and the platform that runs it best is not. Next.js builds the habit and Vercel provides the home. Every developer who learns to think in Next.js becomes a likely Vercel customer. Every Vercel feature that makes Next.js faster gives that developer a reason to stay. By October 2025 the framework had 500 million downloads in a year, and the platform powered two of the top twenty sites on the internet.
This model only works if the free thing is genuinely good on its own terms. Developers can tell when open-source software is a funnel wearing a costume. Rauch's credibility came from twenty years of contributions that asked for nothing: MooTools, Mongoose, Socket.IO, and small utilities like
ms and wifi-password with no business attached. When he finally built something with a business behind it, the community extended him trust he had already earned.The model has a weak point, and 2025 exposed it. A habit can be portable. When the boycott came, Replit's founder could post a three-step migration for Next.js projects, because the framework's openness, which is the source of its reach, also means it can be taken elsewhere. The home has to keep earning the occupancy.
Tactic: Open-source the layer where adoption is decided and charge for the layer where operational excellence is decided, then keep investing in the free layer as if it were your only product.
◆
Conventional Wisdom vs. the Rauch Pattern
Where his choices diverged from standard startup advice.
| Conventional wisdom | Rauch's move | Phase |
|---|---|---|
| Protect your core IP. | Give away the framework; monetize the hosting. | Scale |
| Get the credential, then the job. | Ship the library; the job follows the library. | Scan |
| Expose power to power users. | Hide complexity until it's asked for. | Scale |
| Defend the craft your company is known for. | Announce the end of hand-written code to your own engineers. | Try |
| Scale a function by hiring. | Scale a function by curating an agent. | Try |
| Founders stay out of politics. | Post the selfie; absorb the boycott. | Risk |
Principle 4
Make infrastructure an output, not an input
The most original idea in Rauch's thinking is a borrowing. React established that a webpage's DOM is a function of state: you describe what should be shown, and the library works out how to change the page. Rauch applied that inversion to the cloud. "We set out to make Vercel the React of the Cloud," he wrote, "where the hyperscaler primitives are outputs not to be directly manipulated." Developers write a framework application. On deployment, the platform generates the servers, caches, functions, and routes. Nobody configures them by hand.
The strategic consequence is large. If infrastructure is an output of code, then whoever owns the framework that produces the code effectively decides what infrastructure gets built. Vercel does not need to beat Amazon at being Amazon. Vercel's infrastructure runs on Amazon Web Services. It needs to sit one layer above, where intent is expressed, and compile that intent down.
The AI era makes the pattern recursive. If code is an output of a prompt, and infrastructure is an output of code, then infrastructure is the output of a prompt. That is why v0 and the hosting platform belong to the same company.
Tactic: Identify the configuration your customers maintain by hand and ask what higher-level description it could be generated from, then own that description.
Principle 5
Disclose complexity progressively
Rauch's phrase "progressive disclosure of complexity" sounds like interface jargon. In practice it is a theory of how to serve beginners and enterprises with one product. Next.js starts with very little code and grows to run sites at the top of global traffic rankings. A developer meets complexity only when their problem requires it, and never before.
The usual mistake is to build for experts and add a beginner's mode later, or to build for beginners and find out the product can't grow. Rauch reverses the burden. The default must be good enough that most users never leave it, and the depth must be real enough that the few who do leave never hit a wall. His engineering principles say the same thing from another angle: "Minimize the number of concepts & modes."
This principle also explains his distribution. Beginners who are well served become professionals who stay. The hobbyist on the free tier in 2018 is the engineering lead choosing a platform in 2024.
Tactic: Design your product so the first meaningful result requires one concept, and make sure every additional concept is introduced only at the moment a user's real need demands it.
Principle 6
Narrow the product until it fits in a gesture
LearnBoost, an education platform built by someone with no diploma, was ambitious and broad. Cloudup, the product that sold, could be described in one motion: drop a file, get a link. ZEIT began with a similarly compressed promise, deploy with one command. Later accounts of his career describe the lesson as simplify, and Rauch's own advice to founders points the same way. Speaking to a PearX cohort, he framed the early-stage dilemma directly: "I'm two people, how do I build this [into] world class? And so I do think taste and focus are your friends."
A small team cannot be world-class at many things, and it can be world-class at one gesture. The gesture becomes the product's identity, the thing users can repeat to each other without explaining. Rauch cited Steve Jobs on the same point: "people don't have time for BS... you have to be so concrete, so simple. You're almost repetitive."
Breadth comes later, through acquisitions such as Turborepo, Splitbee, Tremor, NuxtLabs, and Better Auth, once the core gesture has earned the right to grow around it.
Tactic: Describe your product as a single physical action a user takes; if you can't, cut until you can.
Principle 7
Sweat the copy, the color, and the error code
Rauch's engineering principles have an unusually concrete core. "Sweat every word of product copy you render." "Errors must have globally unique codes & hyperlinks." "Red is not enough to signal 'error' (8% of men have red-green color blindness)." "Feedback must be given to users instantly." Most founders talk about craft in general terms. Rauch specifies the exact pixels.
The investor list helps explain this. It is full of designers: an early Stripe designer, early Facebook and Dropbox designers, the creator of the Inter typeface. His first design hire was a founding designer, not a contractor. He told PearX that taste is the one thing that cannot be faked: "We're trying to create companies that people will remember... Because otherwise you just open a hot dog stand."
In an AI-saturated market this matters more. When everyone has access to the same models, what sets products apart is judgment about details. Vercel's component system, shadcn/ui, which Rauch has described as raising the visual floor of the internet, is that principle shipped as a product.
Tactic: Write down the five smallest details of your product's most common failure state, such as the message, the color, the code, the link, and the latency, and fix all five this week.
Principle 8
Perform your own ego death before the market does
Rauch built his identity on writing code, and he was among the first prominent engineers to say publicly that he no longer does. "I no longer write code. I only prompt." He told his own company that the future of coding is generation. This is not detachment. He loved programming, and he credits it with teaching him focus, discipline, and the ability to absorb the compiler's negative feedback. He did it because the alternative was worse. A company built on the craft of hand-written web code would be the last to notice that the craft was being automated.
The timing is the point. Vercel shipped v0 in 2023, when AI coding was still widely seen as a toy, and built the AI SDK the same year. By 2026, coding agents triggered half of all deployments on the platform. The Next.js business he built by hand now hosts the output of machines that write Next.js. If he had waited for certainty, someone else would own that layer.
The hard part is emotional, not strategic. It means telling employees who joined because they love the craft that their relationship with it is changing, and modeling the change yourself.
Tactic: Identify the skill your company is most proud of and run a serious internal experiment to automate it, with you as the first participant.
Principle 9
Ask for the agent, not the answer
The executive question Rauch says he has nearly stopped asking is "What's the answer?" The replacement is "What's the agent that can get me this answer?" The change sounds small and is not. An answer is used once. An agent keeps producing answers. Asking for the agent forces every leader to build a system instead of an artifact.
At Vercel this has reorganized whole functions. The data team maintains an agent that answers everyone's questions in Slack. Support curates an agent that resolves 90 percent of tickets. A sales team of ten became one person managing a bot. Rauch, who admits he had never opened Salesforce, used the Eve agent framework to get a foothold in parts of the business his engineering background never touched. His standard is "100% agent proficiency," with the meaning varying by level: individual contributors ship with agents, and leaders build the systems that make that possible.
There is also a diagnostic benefit. The question shows quickly whether a team is actually working in the new way or just talking about it.
Tactic: For the next recurring question you'd normally assign to a person, assign instead the building of the agent that answers it, and measure the team by how often that agent is used.
Principle 10
Manufacture founders inside the company
"Traditional founder mode does not scale as you expand your product portfolio," Rauch told CEO Insider. His answer is "recursive founder mode": hire people who are "founders at heart" and make them the CEOs of their own products inside Vercel, with "pools of autonomy" over messaging, marketing, and feature design. Because they build on Vercel's own infrastructure, these internal founders can launch their products directly on the platform, which makes each new product a test of the platform as well.
Vercel Labs extends the idea. Vercel's reputation for curated, high-conviction products had become a constraint in a market where "by the time you're certain, you're already late." So Rauch created a separate unit with a mandate to experiment, protecting the core brand's polish while allowing the mess that discovery requires.
He has also addressed the obvious risk, that talented internal founders leave to become actual founders, as an explicit management problem rather than an afterthought. The acquisitions of NuxtLabs and Better Auth can be read the same way: bringing in people who were already founders and giving them a bigger platform.
Tactic: Give each major product line a single owner with authority over its messaging and roadmap, and build it on your own platform so its success and your platform's improvement are the same work.
Principle 11
Architect for the failure you've abstracted away
Every step Rauch removes from a workflow is a step nobody watches anymore. The April 2026 breach is the case study. Malware that arrived with a Roblox cheat script led, through a compromised third-party AI tool with broad Google Workspace permissions, to Vercel's environment variables. Each link in that chain was a convenience. Abstraction and attack surface come from the same root.
Rauch saw the general shape before the specific incident. He tweeted that agent products should be built "with the assumption that things will fail(over)." He built Vercel Sandbox to put agents "in a little cage." He warned of "shadow IT" on a scale larger than the mobile-cloud transition, and disclosed that Vercel had removed North Korean operatives running fake job interviews on its platform. His old principle, "Most 'flukes' aren't… your tech just sucks," is the right attitude for a post-mortem: an incident reveals how the system actually works.
The broader lesson for anyone building on abstraction is that the stack you no longer think about still exists. Somebody has to think about it, ideally before an attacker does.
Tactic: For every permission you grant an automated tool, write down what an attacker could do with it, and make the narrowest scope the default rather than the exception.
Principle 12
Remember that your public output is part of the product
Rauch's career rested on publishing: demos, libraries, essays, keynotes, posts. That habit made him, and in September 2025 it hurt him. The selfie with Netanyahu was read, by customers who had chosen Vercel partly as a statement of who they were, as a statement about the company. Some left, and at least one engineer resigned publicly. A competitor offered migration discounts within hours. Rauch said the impact was overstated, and Vercel's growth figures for the following months suggest the business absorbed the hit. The episode still produced a persistent organized campaign that later seized on unrelated decisions, such as a terms-of-service update on AI training.
The lesson here is not "never take a position." Founders take positions constantly, including Vercel's co-signing of an open letter on open-weight AI. The lesson is that in developer infrastructure, where switching costs are lower than they appear and community identity is high, a founder's public self is effectively part of the product. Infrastructure built on community loyalty answers to that community. A founder whose reputation was made by publishing will not find it easy to stop. He should at least publish knowing that every post will be read the way his code was, as evidence of what he is.
Tactic: Before posting anything outside your domain, ask how your most loyal and most distant customers will each read it, and decide in advance whether you're willing to defend it on the terms they'll use.
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
— Guillermo Rauch, Between the Wires interviewI could actually extract so much excitement from these few characters on a black screen.
— Guillermo Rauch, Refactoring Podcast, May 2025If it wasn't open source and it wasn't web based, if it wasn't for those two pillars, I would not be where I am. I tried everything else... but there's a longevity and a dopamine hit to being able to create something and share it as a URL.
— Guillermo Rauch, PearX S25 Speaker SeriesWe're trying to create companies that people will remember. That's the game that you get into when you create a venture-backed company. Because otherwise you just open a hot dog stand.
— Guillermo Rauch, 'The AI Cloud,' October 2025We can either give up the web, proxy all our data and intent through one closed mega-app, or reinvent it together in the open.
— Guillermo Rauch, Notable Capital founder session, May 2026The software company (or any company in the future) will fundamentally be an AI factory. That factory can produce code. That factory outputs tokens into the world.
Maxims
- Count the reboots. Every step between an idea and its verified result is a tax on how many ideas you'll ever try.
- Ship before you're introduced. A public artifact that solves someone's problem gets you in where a credential can't.
- Give away the habit. Free software that people learn to think in creates demand for the place where it runs best.
- Generate the infrastructure. Own the description of what should exist, and let the machinery be its output.
- Hide the depth until it's needed. The best products are simple for beginners and still work for the largest customers.
- Software ages like milk. There is no finished product, only the current iteration and the next one.
- Mourn the craft, then automate it. If you won't retire your own favorite skill, the market will retire it for you.
- Ask for the agent. An answer is spent once, and a system keeps answering.
- Every abstraction hides an attack surface. What you no longer think about still needs someone to think about it.
- Origin first. He introduces himself as being from Lanús before he mentions anything else.
