Why do you trust any binaries you've got? Where did your first-use/bootstrapping compiler come from?
And even if you wrote your own OS and compiler from the ground up - who wrote your BIOS? Your network card firmware? Your disk controller software? Your CPU microcode?
This is why it's important to look at PRISM as a political issue and not merely a technical one, like I see a ton of people doing now. The best solution to government spying isn't to tell everyone to use Linux and DuckDuckGo, it's to change the spying itself.
Shifting use away from, as Bruce Schneier puts it, feudal architectures, both puts the Government on notice that its methods aren't appreciated, and creates a damaged class (the SAAS feudal lords: Google, Facebook, AWS, Apple, Salesforce, and others) who can petition the government to lay off the tactics as it's hurting business.
https://www.schneier.com/blog/archives/2013/06/more_on_feuda...
Hell, push this hard enough and a sufficiently feasible decentralized VOIP might become sufficiently common enough to put the WiFi carriers out of the voice business, relegated to carrying encrypted bits. They might know your handset location, your data usage, and the Tor entry point you're using, but that's it. It's something I've been giving though to.
Indeed - but the political changes (if we get them at all) will take time - time probably measured in years or political terms.
The "merely technical" solutions are going to be important in the meantime. Duckduckgo, encfs, Tarsnap, GPG, Tor, ForceSSL - things like that will (probably) help in the meantime (especially if we can help convince "regular users" to use them), as will encouraging places like DDG to implement TLS cyphers that use forward secrecy.
