Great discussion, but as brush242 says it would be far too messy to require proof-of-human work every single day.  A non-profit verification foundation makes more sense, and to avoid double-dipping/double-claiming it would need to base verification on a derivative of some unique identifying feature that cannot be changed.  I think there are three broad categories of such identifiers, all of them imperfect:

• Commercial identifiers (phone numbers, accounts on social networks)
• Government identifiers (passport metadata, tax/identity number etc)
• Biometrics (see CheapID for a proposal on this front)

The fourth way is to build a web of trust with crypto, check out OpenUDC project who have gone down that road.