I'm up at Lake Almanor, and there have been a few fast moving thunder storms hitting the lake in the late afternoon. Since my dad likes to go boating, and getting caught in the middle of a lake in a thunderstorm is incredibly dangerous, we got talking about Dark Sky.
I'm a big proponent of Dark Sky for a few reasons. The first being that their entire API is available to the public, which means pretty much anyone can build an app against their data, so even though I'm on Android, and use Weather Timeline, I'm still using the same dataset.
The second is they started as a Kickstarter campaign. I'm a big supporter of Kickstarters, not entirely sure why, but I back two to three projects a month. I didn't back it at the time, but I love the idea that "1,203 backers pledged $39,376 to help bring this project to life" in 2011, and now it's a largish company helping people stay dry.
The third reason is their dataset's accuracy. I was in SoHo in NYC walking with @jdherg, and he gave me an up to the minute prediction of when it would rain, that was pure magic. How they do it is by compiling a large number of datasets and modifying the data to get a better picture of when rain is coming. Adam Grossman wrote a great technical post back in 2011 about how Dark Sky works, also Kickstarter did a nice interview with him on the high level details.
Anyways, the hilarious thing is a lot of people use this dataset now. Product Hunt's Weather App List has twenty-two weather applications, and from what I can tell (some apps don't list their weather source) at least eight explicitly list Forecast.io (the Dark Sky dataset) as their datasource. Only one explicitly listed another source (8-Bit Weather lists Yahoo).