The #1 way to save hackathons is to stop offering cash prizes to winners. That way people who are coming to the events to make money will stop showing up, and people who show up won't worry too much about winning, and the people who are there for fun/exploration/friendship/camaraderie will be able to build and enjoy whatever they came for. Compare the quality of hackathons with massive prizes in cash or funding (i.e. AngelHack) to those without (i.e. StartupWeekend) and you'll see a dramatic difference.

GroupMe is an outlier, it was a great story for Twilio but it was a huge surprise. We never ever ever sponsored or organized hackathons with outcomes like that as a goal, even after it happened.

People alway asked me why Twilio has always supported building "toys" in our hackathons, contests, etc. - and it is simple. Often in their free developers want to build toys, play, explore, be free to make a mess - and we want to make them happy, and associate being happy with using Twilio. And maybe they'll go back to their day jobs and think of us when a problem we can solve crops up. Sponsors should focus less on getting PR hits and shiny hacks, and more on giving what hackathon attendees want, and the events will be good. Organizers should seriously consider whether sponsors are adding value, and whether the prizes they offer are worth the trade-off.