Did one a couple years ago - there were just 4 of us, and I was the only software guy. In ... about 30 hours (including my drive time to/from home a couple times) we had a decent-looking and moderately functional piece of software. It actually did quite a lot of stuff, but not what was originally talked about. But everyone was still impressed. Why? Mostly because I was able to focus on a few core things while saying
no to most ideas/requests. I just kept saying "no", over and over, while working. Once the base was done, it was easier to say 'yes' to some ideas, or at least sketch them out.
We came in second place, out of about 20 teams that started the Friday night. I chalk about 80% of it up to the focus/constraints and ability to adhere to those.