David Heinemeier Hansson

Hansson built the first version of Basecamp in 380 hours, and he can name the figure to the hour because he was billing for every one of them at fifteen dollars.¹ He was in Copenhagen, his client was in Chicago, and he asked to be paid in an iPod and a MacBook rather than in money.

Hansson is a Danish programmer, the co-owner and chief technology officer of 37signals, and the creator of Ruby on Rails. The tension that organizes him is already inside that first number: scarcity is what made the work good, and he does not trust himself without it. Ten hours a week and no money did the editing on Basecamp, not discipline, and he credits the constraint rather than the effort. Everything downstream follows from taking that seriously, the refusal of investors, the roughly sixty people after twenty-five years in business, the pleasure taken in killing a recurring bill, the flat refusal to sit in an office. The standing caveat runs the same direction as the evidence: every figure here is self-reported inside a friendly interview, unaudited, and told by the man who benefits from it being true.

What he built

Basecamp launched in 2004 on the pitch of "less software," into a market defined by Microsoft Project and the heavy enterprise tools of the era, and the argument was never that it did more but that it did less, better.¹ Two decades on, ease of use is still the first reason customers give in 37signals' own retention surveys, which is the only form of the claim he considers admissible: "it doesn't matter if it comes from your mouth, it's not credible. But if it comes from the customer, it's very credible." Ruby on Rails came out of that build rather than the other way around. He took an obscure Japanese language with no killer application, built the tooling around the thing he actually wanted to make, extracted the framework from the product, and gave it away unmonetized.

By 2007 the company was seven people, running Basecamp while shipping roughly one new product a year.¹ The pattern that makes 37signals unusual is what happened to those products afterward: nothing. Basecamp 1 was pulled from sale in 2010 and sixteen years later still has paying customers, still produces millions of dollars a year in near-pure profit, and still shows them no new features, because a discontinued product carries no development cost and almost no support load once its customers already know exactly how it works. High Rise, the company's second-biggest hit, is frozen and running. Three generations of Basecamp are live at once. The reader is directed to the file on killing the business that made you, where Danny Yeung shuts down an operation that had produced eight hundred million dollars of revenue over three years, while the money was still arriving, and where Hansson's practice is the counter-case rather than the rebuttal: he concedes that a product can get stuck with its initial cohort and age with it, and answers by running several chassis at once instead of sunsetting any of them.

The one check

Around 2005, with Basecamp taking off, roughly forty venture firms came at 37signals and Hansson wanted none of them. He took one anyway, and his reason is the whole man: he did not trust himself to keep refusing. He had arrived in the United States with about fifty thousand dollars and spent nearly all of it immediately, in cash, on an Audi S4 at a dealership Jason Fried took him to. Against a bank balance like that, a twenty-million-dollar check is not a decision, it is a temptation, and he treated his own future weakness as a fact to be engineered around rather than resisted. He agreed to meet Jeff Bezos because Bezos was a builder rather than only a source of money, then wrote a term sheet he calls almost offensively overvalued, expecting it to be turned down. Bezos accepted it, still holds the position, and still receives a dividend check on a company that will almost certainly never have an exit event.¹

What the round bought was not capital. It bought the standing ability to say no to everything after it, which is why the file on raising once to never raise again treats the absence of subsequent rounds as the success criterion rather than the deployment of the proceeds. It bought something else too, which he is unusually candid about: neither owner lacked self-confidence, but there is self-confidence and then there is Bezos telling you that you are right.

The ladder

Pressed to name what actually drives him, since it is plainly neither money nor scale, he answers with a claim about other people rather than about himself: "no one can tell me what to do. No one can tell me no. No one can tell me what to work on."¹ He says he has made what he needs three times over. What the money is spent on is the removal of obligations, one rung at a time, and the reader is directed to the file on the independence ladder for the full descent: no investors and no board first, then customers as a claim he calls tolerable and fair, then the open-source stage where there are no customers at all, and finally a retirement he describes as an exit from commerce rather than from work.

The rung he actually occupies most happily is the third. Omarchy, his Linux desktop project, is omakase defaults on top of Arch and Hyprland, thousands of hours over about a year with no roadmap at all: pick it up, fix five of the seventy-five things that are wrong with it, put it down, repeat. It is also free, which places the project he loves most precisely in the zone that produces the unsolicited direction he objects to, and he leaves that unresolved.

How he operates

He is an introvert who spends at least half his waking hours alone and most of the rest with his family, and says this was always true rather than an effect of age. He needs four uninterrupted hours to do real work; forty-five minutes is worthless and ninety minutes is worthless, because none of the value exists until he is far enough into a problem to hold the whole thing in his head. His career, by his own dating, started when he could close a door, and he would quit rather than work in an office every day of the year even with people he loves and a company he owns outright. The reader is directed to the file on the four-hour block for the three stranger practices around it, including the famously empty desk, which is not an aesthetic but a procrastination defense: "I need to run out of obvious things to procrastinate on before I'll start the real work."

The same rule governs the writing, which is also the marketing. 37signals has never outspent anyone; it has taught, a strategy Hansson credits to Kathy Sierra and has run for two decades through the books and the open source. Getting Real, published in 2006, has chapters of half a page to three paragraphs, and he says that brevity arrived by default then, whereas now "I have to work a lot harder to be that succinct." Rework was submitted at fifty thousand words and cut by the two owners to twenty-five thousand, over the publisher's objection that it would look like a pamphlet, and he credits the cut rather than the writing with making the book work. Remote followed in 2013, years before the argument was fashionable.

Two smaller habits explain more than they look like they should. He hates repeating himself enough that 37signals killed its eight-hour "Building a Basecamp" workshops after three runs. And he derives genuine pleasure from killing a two-thousand-dollar-a-month recurring expense, which he reports as a discovery rather than a discipline: "I just love expense reports. I didn't know this about myself." He takes profits out of the business instead of chasing a valuation, and argues that buying the yellow Lamborghini and the Pagani is a moral obligation of successful capitalists, then immediately volunteers the strongest objection to his own position, that it could all be post-rationalization because he likes nice things and would rather not feel guilty about them.¹ The distrust is the consistent thing. It extends past himself to his co-owner: "I don't trust us. I don't trust Jason."

The rupture, and the outsider

By 2018 a small activist minority inside the company had acquired standing to decide what was acceptable to say, and his description of the failure is not ambush but accommodation, being boiled like a frog. In 2021 the company banned political discussion in work channels, and when the announcement blew up it doubled down, offering up to six months of salary to anyone who wanted to leave. Twenty of roughly sixty people took it, about a third of the company, into a job market where they could start somewhere else within days. His verdict is that it was the best money 37signals ever spent on culture, and his reason is the reusable part: "it didn't just cut out surgically that bit of cancer. It cut out a thick ring around it."¹ The reader is directed to the file on depoliticizing the workplace for why the open, well-funded, self-selected exit reaches people a targeted removal never could, and for how close the lights came to going out.

The detail he tells against himself is that the most useful outsider during the episode was Marc Andreessen, a man he had spent years criticizing publicly, introduced by Tobias Lutke while roughly forty thousand people were attacking on social media at once. What Andreessen supplied was mostly company: the knowledge that other firms were living through the same dynamic. Hansson's conclusion is that his old grievances turned out to be minuscule next to what he actually needed in the moment.

The turn

The archetype earns its name at the point where the constraint stops arriving on its own. Ten hours a week and a fifteen-dollar rate rationed the first version of Basecamp automatically; out of the hundred things customers asked for, the team could build three, so selection was forced rather than chosen. AI removes exactly that. Give the same team 380 hours and ten agents, he argues, and they will build a monstrosity, because the bloat that used to require payment now arrives free.¹ Basecamp 5, the company's first AI-accelerated build, is the reported case: designers could take features all the way to completion for the first time, and what showed up at the shipping gate was not bad work but finished, defensible work with nothing left to stop it. The loudest customer feedback on the release was that a button had moved.

This is not ordinary skepticism, and the sequence matters. He spent the early AI era dismissing the tooling as an interruption machine, "the open office on steroids," which is his own oldest wound restated as a technology forecast. Lutke pulled him out of it by never arguing, only saying look at this, look at this, until the status discomfort of being the out-of-date one in the friendship made him install the things himself. The reader is directed to the files on where you look is where you go and you can't read your way to conviction, which hold the racing frame and the persuasion method respectively. What came out the other side is a position with an unusual shape: converted on capability, unconverted on consequence. He believes the agents work, and believes that their working is the problem, and he explicitly includes himself and his co-owner in the population he expects to misuse the new capacity. The remaining job, on his account, is almost entirely editorial: "our task as software builders becomes so much more about distilling, so much more about killing our darlings."

Where things stand

He still holds the technology seat at 37signals, still sits on the Shopify board, still races cars, and left the Mac after more than twenty years on a sentence with no strategy in it at all: Apple annoyed him enough that he could not in good conscience keep spending money there. He gave Windows two weeks, went to Linux, and burned the boats. The competitor he was never afraid of was Microsoft, on the theory that a fifty-thousand-person organization can only produce fifty-thousand-person software, and the file on fearing the team of four records both the doctrine and his own honest answer about whether it still holds now that four people can generate the sprawl that used to need headcount, which is that he is not quite sure.

The classification here is provisional. Hansson is filed under Fanatical Owner-Operator because the plate's definition describes his record almost line for line: an identity fused with hands-on work, one product perfected rather than a menu widened, and a refusal to sell or hand off because the company is his to run. The plate is contradicted in one place worth naming, since it was built on a founder who loves being on the line among people, and Hansson would quit before spending his days in a room with anyone. Whether that is a different plate or the same one seen from inside an introvert is the open question, and it is not the archive's to settle alone.

Key facts

• Built the first version of Basecamp in 380 hours at ten hours a week, billing fifteen dollars an hour, and asked to be paid partly in an iPod and a MacBook.
• Basecamp launched in 2004 on the positioning "less software"; ease of use remains the first reason customers give in 37signals' own retention surveys.
• Extracted Ruby on Rails from Basecamp and open-sourced it, unmonetized.
• 37signals was seven people in 2007 and roughly sixty after twenty-five years in business, fully remote, meeting in person twice a year.
• Basecamp 1 was pulled from sale in 2010 and sixteen years later still produces millions a year in near-pure profit from customers who have seen no new features since.
• Took exactly one outside investment, from Jeff Bezos around 2005, on a term sheet he calls almost offensively overvalued and expected to be refused; there has been no second round and no board.
• Banned political discussion in work channels in 2021 and offered up to six months of salary to anyone who wanted to leave; twenty of roughly sixty people took it.
• Cut the manuscript of Rework from 50,000 words to 25,000 with his co-owner, over the publisher's objection that it looked like a pamphlet, and credits the cut rather than the writing with making the book work.
• Requires four uninterrupted hours to do real work and keeps a deliberately empty desk so there is nothing obvious left to procrastinate on.
• Converted on AI capability after Lutke's prompting, and holds that removing the cost of building removes the rationing that made the software good.