Last week we were talking in a team meeting about how cool it is that we can fail fast with datajs. The library is by design very focused and avoids entangling itself with multiple components and dependencies, and so we have the ability to experiment with the codebase and try things out.
This gives us the opportunity for example to try something out for a week and see whether we like the "feel" of the API, the performance, the usability, and the value-to-KB tradeoff. If we don't, we can change it, or drop it completely, and we've only spent a week. Not even a wasted week, as we usually end up learning a lot in the process.
The ability to have a rapid cycle and be willing to fail, learn and move on without a major impact is often referred to as "failing fast", and it's a good thing.
The Love Clean London website and Windows Phone App launched recently, and Planky makes a very interesting point that had not come to mind: the elasticitiy of Windows Azure allows you to start small and grow big, and that is a key enabler for failing fast - low cost of failure. Now the city of London is free to try out other approaches and projects, try out services, and learn from actual experience, all without having to place big bets that come with considerable baggage.
That said, of course I hope that the app is a success, because who wouldn't want a cleaner city?