Part IThe Story
The first commit came on a Friday night in October, around ten o'clock. Chris Wanstrath supplied that detail himself two years later, unprompted. Development of the platform is dated to October 19, 2007, and that date was a Friday. Wanstrath was twenty-two. By day he was a consultant, building Rails sites for other people's money with his friend P.J. Hyett so that the two of them could eventually afford to build something of their own. By night, and on Saturdays, he was writing the code for a website whose name another man had come up with, to solve a problem that perhaps a few thousand people on earth could describe. Git, the version-control system Linus Torvalds had written in 2005 and named, by his own account, after himself, was fast, distributed, and hostile to newcomers. Wanstrath and Tom Preston-Werner loved it. They also had, as Wanstrath later put it, "no acceptable way to share code with others." So they were going to build one.
Nothing about that hour suggested what it would become. No investors had been pitched. There was no office. The company that would eventually hold the world's source code began as one young man's small entry into a ledger of changes, which Git is built to keep forever. The detail is worth dwelling on. Git does not forget. Every commit carries a timestamp and an author and sits on top of the one before it. A repository is an argument that history is cumulative and recoverable, and that you can always go back and see who changed what, when, and why.
Wanstrath spent the next eleven years building the most important company in that discipline. He was its first chief executive, then not, then chief executive again. He was a self-taught programmer who had dropped out of an English degree, and over those years he wrote apology letters that were read in boardrooms, sold to the company his teenage self had refused to let "touch any of my stuff," and walked away a billionaire at thirty-three. Then he went back to making things for strangers on the internet, which is roughly where he had started.
The question his career keeps raising is an old one in a new dialect: where does a person who found his people among strangers actually belong? In the room with the other builders, or at the head of the table? Wanstrath kept answering it, and then revising the answer. Each revision is on the record.
By the Numbers
The Wanstrath Ledger
22Wanstrath's age at GitHub's first commit, October 2007
4 yearsGitHub ran on customer revenue alone before taking outside capital
$100MAndreessen Horowitz's 2012 investment, at a $750M valuation
100,000GitHub users by July 2009, fifteen months after public launch
$300MAnnualized recurring revenue when he left in 2018
$7.5BMicrosoft's all-stock purchase price, announced June 4, 2018
34,020Copies of Tape to Tape sold in week one, Null Games' first release
Softcoding in the Grandparents' Basement
Begin with the hardware, because the hardware was where it started. The first machine was an IBM PC running OS/2 in his grandparents' basement in Ohio, and the first language was QBasic. In sixth grade, for show-and-tell, he brought in a printout of a web page he had made, "including its source code." Decades later he added, with the dry self-regard that runs through nearly everything he writes: "I like to imagine that everyone was impressed."
The summer after seventh grade, a group he describes only as "rowdy high schoolers" took him to a Linux installfest at the local university and put Red Hat on his family's old PC. That became his main computer for all of high school. "So pretty much from the start," he said, "I was a web-slinging, UNIX-loving hippie." The same high schoolers handed him the O'Reilly camel book and told him to learn Perl. He did not enjoy Perl. They also showed him PHP3, and then, in his words, "suddenly it all came together."
What came together was a sensibility he had picked up in MUDs, the text-based multiplayer worlds of the late 1990s (and the MUSHes and MUXes and MOOs, he notes with a collector's fussiness). The MUD engines were written in C, and players extended them with house scripting languages. Writing C was called "hardcoding"; writing scripts was "softcoding." "I had no idea what I was doing in C," he admitted, "but I really liked the softcoding aspect." PHP felt like softcoding for the web: HTML with logic poured in. He built the MySQL and PHP backend for a high school comedy website himself, "before blogging software was popular," and declines to this day to name it. The games that pulled him in were Meridian 59 and Diablo, both worlds full of strangers.
The pattern is consistent: older kids who show up and hand him a tool, an online world where the fun is in extending the system, and a community that teaches by doing. When he describes how he actually learned to program, the code barely comes up. He talks about the people. "The strangers on the internet," he told an audience at the Computer History Museum in January 2017, "ended up helping me find a sense of a group. They were trying to help share knowledge and build each other up and teach me things." He had abandoned several languages before he settled on PHP, and he settled on it because its digital community made him feel welcome, not because it was elegant.
He graduated from St. Xavier High School in Cincinnati in 2003 and enrolled at the University of Cincinnati as an English major, on reasoning both practical and slightly evasive: "I figured whatever I did in life, I'd have to speak, read, write English." He took one or two programming classes and thought about switching to computer science. Instead he stopped going to class altogether. "I didn't go to class, and I programmed," he told the university's own magazine, laughing. His roommate, Mark Wilson, with whom he played in bands, said: "He was always coding." Wilson also likes to remind him who did the housework that made the coding possible.
The parental reaction gets the best line. "Instead of doing drugs or whatever, I was programming," Wanstrath said. "So, my parents were equally mad at me."
In his first year of college he switched his Linux distribution to Gentoo and became fascinated by its package manager, which was written in Python. He bought the enormous O'Reilly Python book and was working through it when, "randomly," he found Ruby on Rails. "It hit me like a bolt of lightning and suddenly my PHP and Python days were over." Web 2.0 had just been coined. JavaScript, in his phrasing, was saying, "Hey, everyone. I've been here all along."
He has said plainly that the degree never figured in his plans. "I don't ever think I thought that a degree was going to be necessary. I thought skills were necessary. All I wanted to do was start learning a lot." He left after about a year and a half, had a brief and largely unremembered job at a trucking logistics company in New Jersey, and in 2005 accepted an offer from CNET Networks in San Francisco. He had been entirely self-taught until then, and he knew CNET was taking a chance on him. "I didn't think I was going to cut it as a programmer," he said, "but I'm sort of addicted to it." Asked about the risk of moving across the country at twenty without a degree, he answered with a question: "Do 20 year olds think about risks?"
A Complete Failure at Being an Employee
At CNET he wrote PHP for GameSpot, mp3.com, and tv.com, which suited a boy who had wanted to make video games. A colleague there, Kim Cuartero, gave him a Logitech MX518 mouse. Years later, asked to describe his setup, he spent more words on that mouse than on any other tool he owned. He had used the same one for five years, and it "still feels futuristic." It "fits my hand perfectly." "Every other mouse feels like I'm trying to pull the OS's cursor dingus through a pool of mud." It sat on a free Ratpadz mouse pad. "So basically I got really lucky. Thanks Kim!" In the same interview he admits that he "lose[s] every umbrella I touch," that he listens to metal, that he plays a great deal of Carcassonne, and that he photographs himself in Photo Booth each month wearing the newest pair of sunglasses from a subscription service called Stunners of the Month. "Not because I'm vain (though there is that)."
The interview reads like a minor document, but it says a good deal about him. He values tools that disappear into the hand and gifts from coworkers. He jokes at his own expense in parentheses, and he cares about small details in a way that some people would call obsessive.
In 2006 he moved within CNET to Chowhound, where the company was starting to use Rails, which he had been writing on the side for over a year. There he met P.J. Hyett, and the two worked on the relaunch of Chowhound and then built chow.com. Hyett, three years his senior, is the quietest of GitHub's founders in the historical record. He is a Rails developer with no loud public persona. He wrote the July 2009 blog post announcing GitHub's hundred-thousandth user and the dry January 2010 post explaining that the company had become GitHub, Inc.: "What does that mean for you? In a word: nothing." If Wanstrath was the coder and Preston-Werner the designer and evangelist, Hyett played the steadying friend, the person you start a consulting company with when you can't stand an office.
That is exactly what happened. In April 2007, Wanstrath and Hyett left CNET and founded Err Free, a Rails consultancy. The reason Wanstrath gave, in a 2009 keynote in Atlanta, was a failing he presented without shame:
"See, when it comes to being an employee, working in a cubicle, showing up at 9 and leaving at 5, I'm a complete failure. I just can't do it. It sucks. I'm not saying working in a cubicle is inherently bad. I'm just not good at it. Some people are good at sports. Not me."
The plan was patient and modest. They would consult for a few months, save money, and then work full-time on their own website, self-funded, with full ownership. "It was our path to freedom. Being our own bosses." The website was FamSpam, launched in December 2007, which he described as "like Google Groups but simpler, with an emphasis on photo sharing. For families." They pitched it to people they met in restaurants and bars, and the people loved it. "Oh, my whole family would use that!" they said. "I would pay good money for that."
— Chris Wanstrath, Startup Riot keynote, 2009They didn't. Pay attention to what people do, not what they say. There's a big difference there. My mom thought it was a great idea. And so did we. But what we didn't know at the time is how little the idea matters.
So the company that became GitHub's direct predecessor was a family photo-sharing site that nobody paid for. Wanstrath has told that failure many times, and almost always as the start of the story rather than a footnote to it. "GitHub's story is one of many, many failures," he told the Startup Riot audience. "My part in it even begins with one."
FamSpam also served as GitHub's first real testing ground. Once the basics of GitHub worked, Wanstrath and Hyett used it every day on FamSpam, "really getting a feel for what was missing and what was working." The failed product was where the successful one learned to walk.
Saturdays at Zeke's
Tom Preston-Werner grew up in Dubuque, Iowa. His father died when he was a child; his mother was a teacher and his stepfather an engineer. He spent two years at Harvey Mudd College in Claremont, California, and left. In 2004 he founded Gravatar, a service for avatars that follow you from site to site, and sold it to Automattic in 2007. He came to San Francisco to work at Powerset, a natural-language search engine that Microsoft bought. When Microsoft offered him a $300,000 bonus plus stock options to stay, he turned it down to work on GitHub full-time, and wrote a blog post about it under a title that left no doubt: "How I Turned Down $300,000 from Microsoft to go Full-Time on GitHub."
So the two founders had matching backgrounds. Both were Midwestern, both had left college, both were self-taught, and both were drawn to Ruby, a language whose community Preston-Werner once called "my model for how it should be done." Both had left larger technology companies because they could not tolerate being employees. Preston-Werner had already lived through the small-founder version of the story once. He had built something useful, discovered how expensive it was to give away, and sold it.
They met through code before they met in person. "Tom I knew through the open source community," Wanstrath said. "He had released some pretty awesome Ruby code that I was using." Preston-Werner returned the compliment: "I'd always admired his work. He was doing consulting and had put out a bunch of open-source libraries that a lot of people used." They were regulars at a San Francisco Ruby group that met every two weeks in someone's office for an hour-long technical talk. Preston-Werner later said the meetings eventually filled up with venture capitalists hunting for talent, which drove him toward smaller gatherings.
The founding moment, in Wanstrath's telling, was at Zeke's, a sports bar about a block from where GitHub's headquarters would later stand. After a meetup, Preston-Werner pitched him the idea of a Git hosting site, "a place to easily share code and learn about git, a git hub." Wanstrath was in immediately. They began meeting on Saturdays. "We'd grab brunch, talk about ideas we had for the site, then get to work. Tom would design pages or features and I'd implement them." Preston-Werner remembers the next stretch less calmly: "For the next three months Chris and I spent ridiculous hours planning and coding GitHub."
From the beginning the work was split between vision and implementation, between the man who pitched and the man who built. That split would eventually show up in titles, then in a reversal of titles, then in a crisis.
Their first beta user was Yehuda Katz, whom Wanstrath called "a prolific hacker" and who could "give us tons of good feedback and invite other high quality developers." The private beta opened in January 2008. Wanstrath wrote the first post on the company blog, "The Blog Arrives," on February 22. The legal entity was named Logical Awesome LLC, as if to confirm that adults had not yet entered the room. Hyett came aboard as the third founder. The public launch came on April 10, 2008. "Hosting git repositories wasn't something we'd ever done before," Wanstrath said. "We had to rewrite a few basic systems because our assumptions were wrong."
Even the mascot came cheaply. The Octocat, the five-armed cat-octopus that would become one of the most recognizable logos in software, was clip art. A graphic designer named Simon Oxley had drawn it for iStock under the name "Octopuss." GitHub negotiated exclusive rights, renamed it, and trademarked it. Asked about it years later, Oxley said: "I don't remember drawing it."
The fourth name usually listed among the founders is Scott Chacon, the company's "resident git wizard," as Wanstrath called him. GitHub sent him "all over the world" to teach Git. He later wrote the standard text on the tool, Pro Git. In an industry that rewards founders who sell, Chacon was paid to explain. That GitHub kept him on the road was an early sign of what the company actually was: not a code locker but a school.
The business model took longer to settle than it looks in hindsight. Preston-Werner's Gravatar experience had taught him that "offering a resource intensive service for free was a losing proposition," and storing and serving code would mean large server bills. The founders' first instinct was to charge for everything. "There was nothing like GitHub before GitHub," Wanstrath said, "so we were thinking this is a real business, we have to make it sustainable, so we'll just charge for projects. It feels weird to charge for public projects, but it also feels weird to give things away for free. It took a lot of iterations until we decided that private repositories would cost money and public repositories would be free. It's simple in retrospect."
That rule, free in public and paid in private, did more than pay the server bills. It made openness the default and secrecy the premium tier. Any developer who shared code freely was advertising the product, and any company that hid code paid for the privilege. Wanstrath had once explained the business to an imagined cocktail-party guest in one sentence: "If you want to share your source code with everyone and make it public, you don't pay anything. If you need to hide your source code because it's private and runs your business, you pay." The model stayed in place for more than a decade.
Profitable, Small, and Smug About It
By February 2009, when Wanstrath stood before the Startup Riot crowd in Atlanta, GitHub had four full-time employees and one part-time, "thousands of paying customers," and an accelerating growth rate. "January was our best month so far, and February is looking pretty damn good." He offered the comparison the tech press kept making and then took it apart: GitHub, he said, was "Facebook for developers," which "is great when talking about GitHub as a website, but not so great when describing GitHub as a business. In fact, I think we're the polar opposite of Facebook as a business: we're small, never took investment, and actually make money."
He then mocked his own success, listing two criteria (profitability, and being invited to give the keynote), checking both, and declaring victory before retracting it. "Success is about being able to fall asleep at night without your mind racing with new ideas. Success is not worrying if the new feature or bugfix is going to be a huge hit." And then: "Frankly, success sounds pretty boring. I'm much more interested in the f word: failure."
The month before, the Crunchies had named GitHub best bootstrapped startup. By that February, the site held more than 46,000 public repositories, 17,000 of them created in the previous month alone. It passed a hundred thousand users in July 2009 and one million repositories in July 2010. In the first five months of 2011, a ReadWriteWeb analysis found GitHub had logged 1,153,059 commits, against 624,989 for SourceForge and 287,901 for Google Code. In March 2015, Google announced it was closing Google Code.
None of this was obvious to everyone at the time. When Wanstrath's Startup Riot keynote reached Hacker News on February 19, 2009, the top comment was a long complaint. The commenter argued that GitHub had built a closed-source business on top of an open-source tool, that Wanstrath was "so arrogant about it," and, worse, that the whole thing was a contradiction. "The whole point of distributed version control is that it is distributed. Github is the antithesis of this! … It's not Facebook for programmers, it's Geocities for programmers." The comment ended with a prediction: "These 'hubs' never last. Just try to name one that has."
It was a wrong prediction built on a correct observation. Git had been designed so that no central server was needed. Every developer's copy is a complete repository, so the architecture had no center. GitHub put a center back. It offered a canonical place to push, a public profile, a dashboard, a social graph, a pull request button. In engineering terms it was a step backward. In human terms it was a town square, and people, it turned out, wanted a town square more than they wanted a perfect peer-to-peer mesh.
Torvalds himself captured both halves. In 2012 he said GitHub "should be commended enormously for making open source project hosting so easy." In the same interview he called the site's pull requests and online commit editing "just pure garbage" and GitHub "useless for these kinds of things." The inventor of Git praised the hosting and condemned the social layer, which was the very part Wanstrath believed had made GitHub matter.
Wanstrath had in fact begun to see the social layer as the product. "As we got more into it," he said in April 2008, the month of the public launch, "we realized how things like the dashboard could change the way we deal with open source contributions. After that, all the social features began to fall into place." Asked about traction, he gave a builder's answer: "From my perspective it had traction right away."
There was another contradiction the Hacker News commenter had spotted, and it did not resolve. GitHub's own website was never open source. In April 2014, a free-software activist named Mike Gerwitz wrote asking the company to release github.com's JavaScript under a free license. Wanstrath replied on April 16, copying Preston-Werner and Hyett, in three sentences: "We have no plans to release github.com's JavaScript as free software at this time, nor do we have plans to remove the site's dependence on JavaScript. Thanks for the interest." The house where open source lived had locked doors. Wanstrath did not apologize for it, and given the company's business model, there was nothing obvious to apologize for.
When Wired visited in early 2012, GitHub had moved into a 14,000-square-foot loft south of Market that once housed the blogging company Six Apart. Two years earlier it had been eight people holding company meetings in San Francisco cafes; by the start of 2011 it was fourteen "hubbernauts"; a year later, fifty-seven. The founders had turned the biggest office into a parody of an executive suite, with a fake fireplace, plush leather chairs, and a wooden globe that slid open to reveal single malt scotch. A painting on the wall showed the Octocat dressed as Napoleon. Wanstrath, then the CEO, had no private office. He worked on the open floor next to the coders while LCD Soundsystem played, loudly.
The organization was officially flat: no middle managers, open allocation, employees choosing their own projects, the chief executive setting salaries. It looked like an engineer's ideal workplace, and the press loved it. Underneath was the question flat companies always face, which is who handles a complaint when nobody is in charge.
The Biggest Check, Handed Over
In July 2012, four years after launch, GitHub took its first outside money: $100 million from Andreessen Horowitz at a $750 million valuation. Peter Levine, the general partner who joined the board, said it was "the biggest investment we've ever made." He disclosed that GitHub had been growing revenue at 300 percent a year. A half percent of the round went to Ron Conway, whom Preston-Werner had met at a Y Combinator event. Preston-Werner later told employees at a weekly "Beer:30" gathering, while wearing an enormous banana-yellow foam cowboy hat, that the Conway connection showed the value of "serendipitous interactions."
The money changed the job titles. Wanstrath, who had been chief executive and lead developer since 2007, stepped aside after the Series A, and Preston-Werner became CEO. Wanstrath became president and, in practice, head of product for GitHub.com. The man who had built the site went back to the product, and the man who had pitched it took the company.
That fits the founding split exactly, and it also fits what Wanstrath had said about himself in Atlanta: a person who could not stand cubicles, nine-to-five, or being managed, and who had been drawn into programming by the joy of softcoding, of extending a system from inside it. Some founders want the chair. He seems to have wanted to keep building.
The record from that period shows a man writing software. He had already authored and released Resque, a Redis-backed Ruby library for background jobs. He wrote Mustache, which he described as "logic-less Ruby templates." He wrote pjax, whose README summarized it in one line, "pushState + ajax = pjax," and which, he said later, came out of his long affection for REST and hypermedia. He wrote a command-line tool called gist with a modest description: "Potentially the best command line gister." He wrote Emacs modes, including
textmate.el and coffee-mode. In August 2011 he published a blog post about swapping in the Ace editor for in-browser file editing, under a title that is a whole philosophy in four words: "Edit like an Ace."Earlier, at Ruby Hoedown in 2008, he had given a keynote that opened by parodying keynotes. Ruby, he predicted, would become "more popular." There would be "more implementations." Some Rubyists would get dogs, some cats, and some would choose children, "a common mistake, as, in my experience, children are much more expensive than either dogs or cats." He assigned the leading figures of the Ruby community roles from Batman: Matz as a calmer Two-Face, DHH as Batman. "As for me," he said, "I am Iron Man." The copy later forked across GitHub carried the title the talk deserved: Start your side project!
In July 2013, GitHub released the first version of a framework then called Atom Shell, which let developers build desktop applications with web technologies. It was eventually renamed Electron, and the Computer History Museum's official biography of Wanstrath calls him "the inventor of the Electron development platform." The editor Atom Shell existed to serve came seven months later.
Then the titles reversed again.
The Least Transparent Blog Post GitHub Ever Wrote
On January 21, 2014, Preston-Werner announced that he would become president and focus on research and development, and that Wanstrath was CEO again. Recode noted that such role swaps "seem to be all the rage," and pointed to a similar one at Automattic, the company that had bought Gravatar.
Fifty-two days later, Julie Ann Horvath posted on Twitter.
Horvath had joined GitHub in March 2012 as a designer and front-end developer and was the only woman on her team. By late 2012, as GitHub grew rapidly, there were still only three women on the technical side. She started Passion Projects, an all-women talk series held in GitHub's office. She got approval for it by interrupting a company meeting: Preston-Werner had just given a speech about the importance of hiring women coders, and she stood up and pitched her idea on the spot. ReadWrite later called GitHub "a great place for women to work," mostly because of her. When she resigned, she alleged she had been harassed by "leadership" for two years. The harassment, she said, came from a co-founder and his wife, and a male engineer whose romantic advances she had rejected had begun "passive-aggressively ripping out my code from projects we had worked on together." Before she went public, news of her departure had leaked on the anonymous app Secret, where commenters who seemed to be coworkers wrote: "Self proclaimed queen of GitHub is leaving her throne. The masses cheer." And: "Good fucking riddance."
On Sunday, March 16, two days after her posts, Wanstrath published a short note on the company blog titled "Update on Julie Horvath's Departure." The relevant founder and the engineer were on leave. The founder's wife "has never had hiring or firing power at GitHub and will no longer be permitted in the office." He described the company's growth and its recent hire of an HR lead, and then wrote two sentences that a lawyer would likely have cut: "I would like to personally apologize to Julie. It's certain that there were things we could have done differently."
The blogger Ben Brooks called it "about as good as a corporate response can get," and singled out one paragraph as "the only annoying section," the one about rapid growth. The Verge called the apology "rare in the startup industry." Preston-Werner, whose name had not appeared in the note, was quickly identified in the press.
What came next is the most revealing episode in Wanstrath's public career, because he got it wrong and then said so in writing.
On April 21, after what the company described as a "full, independent, third-party investigation," Wanstrath posted the results. The investigation, he wrote, "found no evidence to support the claims against Tom and his wife of sexual or gender-based harassment or retaliation, or of a sexist or hostile work environment," but did find "evidence of mistakes and errors of judgment." Preston-Werner had resigned. "Tom has been a huge part of this company from the very beginning and we appreciate all that he has done for GitHub." Horvath disputed the process publicly: "there was no investigation… There was a series of conversations with a 'mediator' who sought to relieve GitHub of any legal responsibility." A spokesperson told Ars Technica that the specifics were confidential and that "everything we're sharing is in Chris's blog post." Preston-Werner wrote that he and his wife were "prepared to fight any further false claims on this matter to the full extent of the law," and also that "I've made mistakes, and I am deeply sorry to anyone who was hurt by those mistakes."
Exactly one week later, on April 28, Wanstrath posted again.
— Chris Wanstrath, GitHub blog, April 28, 2014Last Monday I published the least open and least transparent blog post GitHub has ever written. We failed to admit and own up to our mistakes, and for that I'm sorry. GitHub has a reputation for being transparent and taking responsibility for our actions, but last week we did neither. There's no excuse. We can do a lot better.
This time he named the investigator, Rhoma Young, and gave the findings that the first post had blurred: "Tom Preston-Werner in his capacity as GitHub's CEO acted inappropriately, including confrontational conduct, disregard of workplace complaints, insensitivity to the impact of his spouse's presence in the workplace, and failure to enforce an agreement that his spouse should not work in the office." He also kept part of the original conclusion: Young had found "nothing to support a sexist or discriminatory environment at GitHub." Recode summarized the shift in one line: "The tone of GitHub's response has gone from defensive to super apologetic."
Whether the second post fully answered Horvath's account is not something the record settles, and she plainly thought the first one hadn't. But the structure of what Wanstrath did is clear, and it is unusual. He published a statement, judged it a failure against his company's stated values, and published a correction that called the earlier statement the worst thing the company had written. In effect he made a commit that reverted his own. In version-control terms the earlier commit was not deleted; a new one was added that undid it, with a message explaining why. Both remain in the history.
Seven years earlier he had told a room in Atlanta that he was more interested in failure than success. In April 2014 he had to fail in public, in the company's own voice, and then say so.
The structural consequence came quietly. In July 2014, GitHub added a layer of middle management, ending the flat organization the business press had admired. Julio Avalos, the general counsel and administrative officer, took over business operations and day-to-day management. Wanstrath remained CEO.
Asked by Fortune in 2015 about his biggest management challenge since retaking the role, Wanstrath answered with something other than the scandal. "We have 335 employees and more than half of them, at any one time, aren't in the office. So we use GitHub really heavily to communicate… but it's hard. It's not like you can just grab someone and walk over to the whiteboard. So it's been a struggle for us." When the question turned to the scandal, the edited transcript ran his answer at length. The company that had grown up being praised was now being scrutinized. A 2016 Business Insider headline described "a full-blown overhaul as execs and employees depart."
The Hackable Editor and Its Heir
During the same weeks, GitHub shipped the most personal product of Wanstrath's tenure.
Atom, which GitHub called "a hackable text editor for the 21st Century," was released in beta on February 26, 2014, three weeks before Horvath's posts. On May 6, eight days after the April 28 apology, GitHub released the full editor under the MIT license. Until then, many of its libraries and packages had been open while the core was not. Atom reached version 1.0 on June 25, 2015, with more than 350,000 monthly active users. Wired covered the launch under the headline "GitHub Atom's Code-Editor Nerds Take Over Their Universe." In March 2016 Atom passed one million monthly active users, a figure GitHub knew because the editor shipped with a package called
metrics that reported usage. Electron, the framework extracted from Atom, reached version 1.0 on May 9, 2016.Atom brought together everything Wanstrath had loved since childhood. It was softcoding: an editor whose behavior users could rewrite from inside, the way he had once extended MUDs. It was built from web technologies, the HTML and JavaScript he had been writing since the sixth-grade printout. And it was a tool meant to disappear into the hand, the software equivalent of the MX518.
For Wanstrath, it was also a way back into building while he ran the company. In 2015 GitHub raised $250 million led by Sequoia Capital, at roughly $2 billion, opened its first office outside the United States in Tokyo, and launched GitHub Universe, its developer conference, where Wanstrath gave the opening keynote three years running. Through all of it, the CEO's name stayed on an editor.
Electron spread well beyond Atom. Atom itself did not last. Its repository,
atom/atom, now displays a gray label: Public archive. It still shows 60.7 thousand stars and 17.2 thousand forks, preserved in the state Git guarantees, readable and closed to new commits. Microsoft, meanwhile, had been building its own code editor, Visual Studio Code, and hosting it on GitHub. One retrospective video about Atom carries a title that does the analysis for you: "The Editor That Accidentally Built Its Own Killer."The shape of that story is familiar. Wanstrath built tools that let other people build things, and those people did not always build what he expected. GitHub took a tool designed to have no center and gave it one. Atom produced a framework that outlived it. It is a strange kind of authorship, in which success is measured by how thoroughly your work becomes infrastructure for someone else.
Selling to the Ecosystem He Had Pledged Against
The engineers who founded GitHub distrusted salespeople. Wanstrath admitted this on an Andreessen Horowitz podcast in March 2015, with Peter Levine: companies founded by engineers often have "a deep-seated mistrust of sales — or more precisely, salespeople," and GitHub was one of them. What changed his mind was customers asking for a sales organization to help them buy. He came to see good salespeople, as the podcast summary put it, as "like good teachers." It was the Chacon idea again, now applied to revenue.
The demand had been there from the start. At ZendCon in 2008, Wanstrath had been "overwhelmed by the number of PHP developers who were working on government sites. They came up and said: 'We love GitHub, we're just not allowed to use it. Please help us.'" GitHub Enterprise, a version customers could run on their own servers, finally shipped in November 2011. The enterprise customers grew. By 2016 Wanstrath was naming John Deere, Ford, and Walmart.
He was also naming Microsoft. In a Fortune interview in September 2016, he described his own change of heart:
— Chris Wanstrath, Fortune, September 2016I started programming on Microsoft technologies, but I switched early on to Linux, all open-source. I didn't want Microsoft to touch any of my stuff. They have an ecosystem and I was pledging my allegiance to a completely different ecosystem. To see that all change has been pretty remarkable. It's real.
The boy who had Red Hat installed on his family's PC after seventh grade, who described himself as a "UNIX-loving hippie," was now telling a business magazine that Microsoft had earned his trust. "I think a lot of people were skeptical of Microsoft at first, but they're real." In the same interview he described the test he applied: "Developers are great at detecting BS." Microsoft had joined GitHub in late 2013. By 2016 it had more open-source contributors on GitHub than any other organization. It had moved .NET Core, PowerShell, and Visual Studio Code there, along with most of its documentation. Wanstrath noted that the open-sourced .NET project had more outside contributors than Microsoft employees, "but every one of the changes was approved by Microsoft." He seemed to admire the arrangement: openness at the edges, control at the center, much like GitHub's own.
In August 2017, CNBC reported that the CEO of the $2 billion startup was stepping down. Forbes's headline was more precise: "GitHub CEO Chris Wanstrath To Step Down After Finding His Own Replacement." He had decided to give up the chair a third time and to choose who got it.
The replacement, when it came, came attached to Microsoft. On June 4, 2018, Microsoft announced it would acquire GitHub for $7.5 billion in stock. Satya Nadella wrote that "developers will be at the center of solving the world's most pressing challenges." Nat Friedman, who had run Xamarin, a developer-tools company Microsoft bought in 2016, would become GitHub's CEO, reporting to Scott Guthrie, Microsoft's cloud chief. Friedman fit the pattern: an open-source veteran who had built his own company and then gone inside Microsoft. Wanstrath, by then GitHub's executive chairman, would become a Microsoft technical fellow.
His blog post, titled "A bright future for GitHub," contained the most melancholy sentence he has published. A decade earlier, he wrote, "Git was a powerful but niche tool, clouds were just things in the sky, and Microsoft was a very different company." He then stated a fact that was also an explanation: "Microsoft is the most active organization on GitHub in the world."
When he left in 2018, GitHub had close to 1,000 employees, more than 20 million users by Wikipedia's count, and $300 million in annualized recurring revenue. Rivals including GitLab, Bitbucket, and SourceForge reported spikes in developers trying to migrate away. Critics cited Microsoft's handling of Skype and Nokia. The Hacker News commenter of 2009 might have taken some satisfaction, though the mass exodus never arrived. The deal closed on October 26, 2018, and Wanstrath left. The Computer History Museum's biography says he "left the company on good terms."
The money was large. Forbes had valued him at $500 million in 2015 and, oddly, at $360 million a year later, a reminder that private-company fortunes are estimates. After the sale, its estimates rose to between $1.8 billion and $2.2 billion. The man who had declared success "pretty boring" in 2009 was now very successful.
Under Microsoft, GitHub made all core features free in April 2020, including private repositories with unlimited collaborators. The rule he had spent so many iterations arriving at, free in public and paid in private, was simply dropped by a company that no longer needed GitHub to support itself. GitHub passed 100 million developers and $1 billion in annual revenue. In June 2025 it passed one billion repositories. The billionth contained one word: shit.
No Loot Boxes
He went back to the games.
In February 2023, Wanstrath announced Null Games, a publisher of independent games for PC, Mac, and console. It helps developers with marketing, development costs, porting, and publishing. Its website lists its values as refusals. Null will not "publish mobile games or games with gambling, loot boxes, or any other player-hostile behavior." Its first release was Tape to Tape, a hockey roguelite from a studio called Excellent Rectangle, which sold 34,020 copies in its first week. The boy who wanted to make video games, who had once filed his job under "web developer, GameSpot," now owns a game company and has written down what it won't do. Later that year he teased Void, a platform for game developers.
In 2024 he joined Andreas Kling to start the Ladybird Browser Initiative, a nonprofit building a new web browser from a codebase Kling had forked out of SerenityOS, the operating system he had been building himself. Wanstrath's opening contribution, reported by the German trade press, was one million dollars. Like Null, the project defines itself by what it refuses: a browser that does not depend on corporate deals or advertising. On the Changelog podcast in August 2024, the two discussed what it would take to reach an alpha, a plan to gradually adopt Swift in place of C++, and their target, the summer of 2026. Wanstrath had first appeared on that show in January 2010, when GitHub had a few dozen employees. Fourteen years later, the episode's first chapter is titled "Starting with OG shirt nostalgia."
Ladybird is the closest he has come to closing a loop. As a teenager he fell for the web as a platform, writing HTML for a decade and absorbing REST and hypermedia "because it felt natural and made sense." As a founder he built the largest company on that platform. As a philanthropist he is paying for an independent way to see it, on the theory that nobody's business model should decide what you see. In January 2025 he gave a long interview to the author of htmx, a hypermedia library that traces its lineage to pjax. He spent it talking about MUDs and Gentoo and a high school comedy site he still won't name. GitHub came up mostly in passing.
He serves on the board of trustees of the Computer History Museum, an institution devoted to the premise that the past of software should be kept. He lives in California with his wife, children, and cats, plays guitar, and reads science fiction. In 2008 he was asked about his dream setup and answered: "Thought-controlled software and a wireless internet connection in my brain, integrated right into my field of vision. I think I would give up my mouse for that."
In July 2020, GitHub took a February snapshot of every active public repository and some dormant but significant ones, 21 terabytes in all, and stored it as two-dimensional barcodes on archival film in an abandoned mine in Svalbard, near the Global Seed Vault, with an expected life of 500 to 1,000 years. The company that began with a commit made on a Friday at 10 p.m. now keeps its record in a mountain.
The more modest record is still online. At
github.com/defunkt, under a username he has used since GitHub's early days, the profile of Chris Wanstrath shows 22.8 thousand followers and a short list of pinned repositories: Resque, hub, gist, jquery-pjax, Mustache. First in the list is Atom, labeled Public archive. His status is a single emoji, a hamburger. Beside his name is the same button every stranger on the site gets, the one he once built for everyone else.It says Follow.
Part IIThe Playbook
What follows is a set of operating principles drawn from Wanstrath's record: the companies he built, the failures he narrated, the titles he took and gave back, and the letters he wrote. Some are counterintuitive. A few are uncomfortable. None amounts to "work hard." They are written for founders and operators building for technical audiences, running organizations past their first breaking point, or deciding what to do with a title they're not sure they want.
The Wanstrath Principles
- 1.Build for the developers in the building.
- 2.Trust wallets, not compliments.
- 3.Price the asymmetry: make openness free and secrecy the premium.
- 4.Bootstrap until capital is a choice.
- 5.Recentralize what was designed to be decentralized.
- 6.Treat side projects as your real R&D department.
- 7.Separate the title from the work.
- 8.When you fail in public, revert in public.
- 9.Structure is a debt that comes due all at once.
- 10.Hire salespeople who teach.
- 11.Let your adversaries become your largest users.
- 12.Write your values as prohibitions.
Principle 1
Build for the developers in the building
GitHub's first customers were its own founders. As soon as the basics worked, Wanstrath and Hyett used it every day on FamSpam, which meant the product was tested under real, daily, mildly frustrating use by the people who could fix it. "We knew it was good," Wanstrath said, "because we were using it every day during development." Years later, at the Computer History Museum, he named the company's early compass outright: "Our North star in the early days was a lot about mak[ing] something great for the developers in the building."
This is more than dogfooding as a slogan. It is a claim about where taste comes from. A founder who is also the target user can tell good from good-enough without a survey. The risk is the one Wanstrath named in the same talk: eventually the people in the building stop being representative, and the North Star has to move toward the people outside it. The principle works when it is followed early and then deliberately outgrown.
Tactic: Before you write a line of customer-facing code, find a real recurring task in your own week the product must handle, and refuse to ship until it handles that task better than your current workaround.
Principle 2
Trust wallets, not compliments
FamSpam failed in a particular way that is worth remembering: strangers in bars told the founders they would "pay good money" for it, and then didn't. Wanstrath took a lesson from that which he has repeated many times: "Pay attention to what people do, not what they say." He went further, in a line that should bother anyone who treats ideas as assets: "What we didn't know at the time is how little the idea matters."
GitHub's own early signals were behavioral. A prolific hacker like Yehuda Katz joined and invited others. The Ruby community moved its code onto the site. Customers paid for private repositories within months of launch. Compliments cost nothing to give. A
git push to someone else's server takes some trust, and a credit card takes more.Tactic: Throw out any validation signal that costs the person giving it nothing, and track only actions that require time, reputation, or money.
The founders started by planning to charge for everything and arrived, "after a lot of iterations," at a rule that looks obvious only in hindsight: public repositories free, private repositories paid. The rule works because it lines up the business with distribution. Every free public project advertises the product to the next developer. Every paying customer pays to keep its code out of public view. The product's growth and its revenue come from opposite sides of one decision.
The broader point is to find the place in your product where users want to be seen and the place where they need to hide, and charge for the second. Preston-Werner's Gravatar lesson, that giving away a resource-heavy service is a losing proposition, kept the rule honest. Something had to cover the server bill, and in GitHub's case that something was secrecy.
Tactic: List what your users happily make public and what they would pay to keep private, then make the first free and the second your paid tier.
Principle 4
Bootstrap until capital is a choice
GitHub went four years without outside money. In February 2009 it had five people, "thousands of paying customers," and an accelerating growth rate, and Wanstrath could stand on a stage and call his company "the polar opposite of Facebook as a business." When Andreessen Horowitz did invest, in July 2012, it was the firm's largest check ever, $100 million at $750 million, and the leverage was GitHub's: revenue was growing 300 percent a year.
Bootstrapping was not a moral position for Wanstrath. It was where he had started: consult, save, build, keep full ownership. That posture shaped the terms of the eventual raise. A company that doesn't need money can choose its investors and can afford to wait for the right one.
⏱
From Weekend Project to Exit
The pace of GitHub's capital and leadership decisions under Wanstrath.
2007
First commit on Friday, October 19, around 10 p.m. Saturday sessions with Preston-Werner begin.
2008
Private beta in January; public launch April 10. Charging customers within roughly six months of founding.
2009
Crunchies "best bootstrapped startup." Five staff, profitable, 100,000 users by July.
2012
First outside capital: $100M from Andreessen Horowitz at $750M. Wanstrath steps aside as CEO.
2014
Wanstrath returns as CEO in January; harassment crisis March–April; middle management added in July.
2015
$250M Series B led by Sequoia at about $2B. Tokyo office opens.
2018
Microsoft acquisition announced June 4 at $7.5B in stock; closes October 26.
Tactic: Set a revenue or growth threshold at which outside capital becomes optional rather than necessary, and refuse to start fundraising conversations until you've crossed it.
Principle 5
Recentralize what was designed to be decentralized
The sharpest criticism GitHub ever got was technically correct. Git was built to have no center, and GitHub put one back. A 2009 commenter called it "the antithesis" of distributed version control and predicted that "these 'hubs' never last." Torvalds praised the hosting and dismissed the pull requests as "pure garbage."
Wanstrath bet that people want a place more than they want architectural purity. The dashboard, the profile, and the pull request were social features stacked on a decentralized protocol, and they turned a tool into a community. Many large technology businesses follow this pattern: take a protocol made for experts that works without a center, and build the center people actually want to gather in. The protocol stays open, and the gathering place becomes the business.
⇄
The Hub Paradox
How GitHub's choices diverged from the purist reading of its own foundation.
| Purist expectation | GitHub's choice | Outcome |
|---|---|---|
| Distributed tool needs no central host. | Build the canonical host anyway. | Scale |
| Use Git's native email-patch workflow. | Replace it with web pull requests. | Contested |
| A company built on open source should be open source. | Keep github.com's own code closed. | Risk |
| Code hosting is plumbing. | Make it social: profiles, feeds, followers. | Scale |
Tactic: Find a protocol your users already rely on that offers freedom but no gathering place, and build the gathering place without taking the protocol away from them.
Principle 6
Treat side projects as your real R&D department
GitHub was a side project to a side project. Wanstrath built it on nights and weekends while trying to self-fund FamSpam, which itself was a project on the side of consulting. Resque, Mustache, pjax, gist, and a pile of Emacs modes all came out of the same habit. Electron came out of Atom, and Atom came out of a CEO who kept writing software. His 2008 keynote, posted to GitHub, was retitled by its forkers Start your side project!
The insight here is about risk. A side project faces no board review, has no revenue target, and needs no justification beyond the builder's curiosity, so it can explore ideas a roadmap would reject. Electron's later reach shows that the side project can matter more than the main one. Atom's archived repository shows that you rarely get to choose which.
Tactic: Protect a recurring block of time when your best builders work on whatever they find interesting, and judge the program by what it produces over years, not by any single project.
Principle 7
Separate the title from the work
Wanstrath held the CEO title three times in effect, and gave it up twice voluntarily. He led as CEO and lead developer from 2007, stepped aside after the 2012 raise to run product, returned in January 2014, and in 2017 announced he would step down after finding his own replacement. The Startup Riot speech gives the motive: a person who is "a complete failure" at being managed, and who got into programming for the pleasure of building, does not necessarily want to run the institution that building produced.
The strength of this posture is clarity: work goes to whoever does it best, and the founder's ego is not one of the inputs. Its risk is what happened in 2014. When the person with the title is not the person closest to the product and the culture, problems can grow in the gap between them. Separating title from work is healthy only if accountability is as clearly assigned as the title.
Tactic: Once a year, write down which parts of your role you are uniquely good at and which you simply hold by default, and hand off the second list deliberately.
Principle 8
When you fail in public, revert in public
On April 21, 2014, Wanstrath published an investigation summary that cleared his co-founder on the central claims and described his departure in soft terms. On April 28 he called that post "the least open and least transparent blog post GitHub has ever written," apologized, named the investigator, and published findings the first post had blurred. Two posts, seven days apart, signed by the same man.
Most companies, having made a bad statement, let it fade. Wanstrath treated it the way Git treats a bad commit. He didn't delete it. He added a new commit that undid it and explained why. The original stayed visible. The cost was looking wrong in front of everyone. The return was a credibility the first post had spent. His first apology, on March 16, with its unlawyerly "It's certain that there were things we could have done differently," showed the same instinct before the misstep.
Tactic: When an official statement turns out to be wrong or evasive, publish a correction under the same signature, quote the original failure by name, and leave the original up.
Principle 9
Structure is a debt that comes due all at once
GitHub's flat organization was widely admired: no middle managers, open allocation, a CEO with no private office working on the open floor while LCD Soundsystem played. Headcount went from 14 at the start of 2011 to 139 by December 2012 and 234 a year after that. The company hired its first experienced HR lead in January 2014, nearly a year after starting the search. Two months later, the absence of a reliable channel for complaints became a public crisis. Middle management arrived that July, as a reaction rather than a plan.
The lesson is not that flat organizations are wrong. It is that the structure you skip keeps accruing, and it tends to come due at the worst possible time, in the form of the complaint nobody was assigned to hear. Wanstrath's 2015 admission that a remote, half-distributed company was "a struggle" belongs here too. Freedom from process costs something, and someone eventually pays.
Tactic: Before you reach 50 people, name a specific person, outside the founders, who receives and is accountable for employee complaints, even if every other part of your org chart stays flat.
Principle 10
Hire salespeople who teach
GitHub's engineers distrusted salespeople, and Wanstrath said so openly. What changed his mind was customers asking for a sales organization to help them buy, and his realization that good salespeople are "like good teachers." The company had in fact been doing this from its first year. It sent Scott Chacon around the world to teach Git and offered training that Wanstrath admitted "isn't a huge part of our revenue" but helped him "develop a real relationship" with customers.
For any company selling to technical buyers, this redefines the sales job. Developers, in his words, "are great at detecting BS." The way to sell to people who can detect a pitch is not to pitch. It is to make them better at their work, so that buying feels like the next lesson.
Tactic: Rewrite your sales team's job description to measure what customers learned and accomplished, and hire for people who can explain your product's underlying technology as well as your best engineer.
Principle 11
Let your adversaries become your largest users
As a teenager Wanstrath "didn't want Microsoft to touch any of my stuff." By 2016 Microsoft had more open-source contributors on GitHub than any other organization and had moved .NET Core, PowerShell, and Visual Studio Code onto the platform. By 2018 it was the buyer. Wanstrath's own explanation of the sale was the usage statistic: "Microsoft is the most active organization on GitHub in the world."
A neutral platform does not choose its users by tribe. GitHub stayed open to the company its founders had been raised to oppose, and that company became its most committed user and finally its owner. Whether the exit was a triumph or a surrender depends on who you ask, and the developers who rushed to GitLab in June 2018 had their answer. But the mechanism is clear: if a platform is good enough, its former enemies end up depending on it.
Tactic: Find the organization your community most distrusts, make your product work superbly for it anyway, and track its usage as a leading indicator of your strategic options.
Principle 12
Write your values as prohibitions
After GitHub, Wanstrath's projects state their values as things they won't do. Null Games won't publish "mobile games or games with gambling, loot boxes, or any other player-hostile behavior." The Ladybird browser is defined by its refusal to depend on corporate deals or advertising. Neither mission statement is mainly about aspiration. Both are mainly about exclusion.
A positive value ("we put players first") can be stretched to cover almost anything. A prohibition ("no loot boxes") is easy to check and costly to break. Wanstrath spent eleven years watching a business model gradually reshape a product, from the free-public, paid-private rule to the Microsoft era when it no longer applied. His later work reads as if written by someone who learned that the business model is the value system, and that the only durable defense is to rule out certain models in advance.
Tactic: Write down three revenue mechanisms you will never use, publish them, and treat any proposal that edges toward them as a leadership-level decision.
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
— Chris Wanstrath, Computer History Museum, January 17, 2017The strangers on the internet... ended up helping me find a sense of a group. They were trying to help share knowledge and build each other up and teach me things.
— Chris Wanstrath, University of Cincinnati Magazine, April 2014I don't ever think I thought that a degree was going to be necessary. I thought skills were necessary. All I wanted to do was start learning a lot.
— Chris Wanstrath, InfoWorldIt feels weird to charge for public projects, but it also feels weird to give things away for free. It took a lot of iterations until we decided that private repositories would cost money and public repositories would be free. It's simple in retrospect.
— Chris Wanstrath, Startup Riot keynote, 2009Success is about being able to fall asleep at night without your mind racing with new ideas... Frankly, success sounds pretty boring. I'm much more interested in the f word: failure.
— Chris Wanstrath, Fortune, September 2016Developers are great at detecting BS.
Maxims
- The idea matters less than you think. FamSpam was a fine idea that nobody paid for; what people do outweighs what they say.
- Use it before you sell it. A product its builders depend on every day will expose its own weaknesses faster than any focus group.
- Make openness free. Charge for privacy and let every public user advertise for you.
- Earn the right to refuse money. Four years of profitable bootstrapping turned the largest check in a top firm's history into a choice rather than a rescue.
- People want a square, not a mesh. A protocol designed to have no center still benefits from a place where people gather.
- Guard the side project. The things built on nights and weekends, GitHub, Electron, Resque, often outlast the official roadmap.
- A title is a tool, not an identity. Take it when the work needs it and hand it over when someone else can carry it better.
- Revert in public. When your statement fails your values, publish the correction under your own name and leave the original up.
- Flatness runs on credit. The management layer you skip comes due all at once, usually during a crisis.
- Define yourself by what you refuse. "No loot boxes" is harder to bend than "players first."
