I'm not sure how much it really says about feature creep. I mean, sure, features increased, but it was really more about how the people used it than anything.
You can look at it in (at least) two ways: On one hand, the rapid feature implementation on Caffeinator was a natural (and predictable) outcome of having a project where the person working on it was passionate about the project and its impact. On the other hand, it's worth noting that I probably saved us a whole 5-10 minutes on each coordinated coffee trip at the cost of hours and hours of (admittedly extra-curricular -- much like a current project I own at Netflix, at the time Caffeinator was explicitly not what I was getting paid to work on) work. It's quite likely I could have invested some of that time in other projects I was just as passionate about which may have been more relevant to the business (but, well, I didn't and I'm OK with that).
That said, I'd argue the debt tracking system built into Caffeinator was a totally unnecessary bolt-on that made it unnecessarily complex. If I had it to do over again, I'd have kept Caffeinator purely coffee-based (which means, for example, less concern around authentication and validation because we didn't have a big potential for someone to play a practical joke by ordering someone else a drink) and created some other system for tracking debt and transactions. Call it some catchy name like "Bitcoin"[0] or something and ship it.
