In Open Source: Miguel Paraz voted up an answer.
My first job out of university saw me writing educational web platforms at the University of Edinburgh, in a room that used to be a broom closet with a window that didn't close properly. (If you've ever been to Edinburgh, you understand the freezing implications of that.) Incredibly, I had to share this tiny space with someone else. I moved in second, and for a long time, he resented my presence. Eventually, though, the room began to thaw (figuratively and literally) and we started chatting about our respective work.
I'd been publishing on the web and running online communities for nine years by that point (this was 2003). He was an e-learning PhD student who was focusing on eportfolios (when he wasn't playing online tennis): personal spaces where people represent themselves and their learning journeys online.
Existing educational platforms were expensive, and terrible. Nobody enjoyed using them, from the people who had to administer them down to the students. But people were learning and representing themselves in amazing ways on the web - this was towards the beginning of the social media movement, and sites like Livejournal, Blogger and Flickr were heavily used - and the open source movement was beginning to pick up momentum. It seemed to us that there should be no reason why tax-funded institutions should pay hundreds of thousands of pounds for lousy, closed software that locked their data away from them. It was immoral.
So we wrote a couple of papers describing what an open, social, educational platform might look like. It was completely unofficial, but we both used our university email addresses at the top to lend it some gravitas. And sure enough, they were both widely read.
We decided to begin writing it, and initially offered it to the university: it wasn't something we planned on running ourselves, and certainly not as a business. They, however, felt that blogs and social software were used by (here I'll paraphrase) "teenage girls crying in their bedrooms".
So I quit my job, burned through the savings I had, and built the first releasable version. It was called Elgg, because I already owned the domain for use as a respectable-sounding email address, but hadn't uploaded a website. (Saved us having to buy another one.) In November, 2004, we put a version online for people to use; five months later, in March, 2005, we uploaded the code to EduForge under the GPL v2. I was so terrified - other people examining my code! - that I actually went on holiday to Granada, Spain, the next day.
I'd run out of money by that point, and ended up going to work as a web administrator at the Saïd Business School in Oxford, where I continued to work on Elgg in the evenings (sometimes until 2 or 3am). Dave, my colleague, moved to Vancouver to try and ramp up momentum for it, as well as work at the university there. We were beginning to get more and more contributions - and certainly more interest - but the bulk of the work still came from us.
Sometime during my stint at the Saïd, I paid the £50 to start a limited company that would provide support services for Elgg. Dave had wanted to start a band called Curverider sometime before, so that's what we called it. (It was only fair; I had named the software, after all.) We were invited to speak around the world at educational events, which I spent most of my vacation days on attending. Charities from Columbia and schools in Bangladesh were using the software.
Finally, a world-class American university (which I won't name) running a world-class e-learning project (which I won't name) invited us to build the software to power it, and I found myself quitting my job once again. This was it! We'd managed to build up momentum around the project while writing most of the code ourselves, and we worked out that we could support ourselves doing it. So we did: we bootstrapped a company, and continued to evolve the software based on feedback and testing in a way that resembled an open source version of the lean startup methodology.
We rented ourselves an office in Oxford, and for the next year, put together some of the best software of my life. We began to see more and more contributions from countries like the Netherlands, Columbia, Argentina, New Zealand (including a major contribution from one of the big companies behind Moodle) and Australia. We were featured in global news outlets, like Wired and the Guardian. It was exciting times - although we were still writing most of the code ourselves. I was described by a member of the community as a "benevolent dictator", which is about right.
We acquired some investment, and grew our team, which included a very talented designer (we were very wary of the "open source look" of the time). My biggest regret is that, during this period, I didn't move to California (I'm a dual national) and attempt to raise more funding in Silicon Valley. Years later, I moved here, and I finally understand how different it really is - even for a startup providing open source services. We did have one trip as a team, for a week, after which a well-known industry luminary that I deeply respect joined our advisory board.
Ultimately, though, we found it hard to scale open source software as a business. We were constantly told that it had an accounting value of 0, and after a while we lost sight of the fact that although our base market (education) didn't really pay us much, they brought all kinds of other benefits. We tried to shut down our largest site, EduSpaces, which served as a great way to gauge how people actually used the software - I still think it's impossible to run an end-user open source web platform without also running a hosted version that you can examine. Being able to test user behavior, and get direct feedback from non-technical end-users, is important. As well as it being a dumb decision from a product perspective, we also managed to create a lot of bad blood.
I chose to leave Elgg in 2009. Today, Curverider is no more, but Elgg is going strong, now run by an entirely new team. It's more of an open source project than ever before, and it's still widely used. I do wish it was more of a social network in a box than a tool for developers to build on top of, but that's the community's prerogative - and I only have good things to say about the core developers.
Overall, it was an amazing journey, both in terms of the software itself and from the entrepreneurial and personal perspectives. What's particularly amazing is that, after the company itself imploded, the software is able to continue to help people connect and learn from each other. I'm truly excited about that, and so glad we made the original decision to release it as open source.See question on Quora