How will you decide which projects to fund? The ones with more code to show, having founders with a more prestigious CV, etc. tend to be the ones building on old, broken (i.e. Unix) foundations of the past.

There's no shortage of crackpots who are convinced that they can turn computing upside down if someone were to pay their living expenses for a year or so (disclaimer: this might include myself: http://www.loper-os.org/?p=8 I started entirely from scratch, and with a conscious rejection of burdensome OS traditions, but have practically no free time to work.)