It would kind of be like a Facebook which the users had no control over their profiles. All known associates of an individual would be linked together in a giant social network. An algorithm would flag conflicts of interest. This should make it easy to see problems with connections a step away from a public official rather than just high profile associates or close family members.
Automated alerts would trigger when individuals within a public official's group are arrested, convicted of a crime, etc. Much like Google's algorithm identifies "bad neighborhoods" of web site links, you could instantly see if a particular public official was surrounded by questionable or explicitly corrupt characters.
Most importantly this could be used to generate an easy to understand corruption score, making it simple for the public at large to understand how corrupt the individual is. Once a particular politician or official scored past a certain point they may become an outcast, finding potential campaign fundraisers and advocates avoiding them.
The migration of eyeballs away from television and old media is going to have a profound impact on the democratic political space. Imagine if a political candidates advertisements delivered via Google also contained a corruption score next to them, much like how e-commerce stores on Adwords display user ratings.
In addition, their developer section (http://sunlightlabs.com/) is full of APIs and datasets for dealing with government transparency data.
"Influence Networks is an open-source, collaborative directory of relationships between people, institutions and companies. Each relation has its own level of trustworthiness, so that facts can be distinguished from noise."
Bonus points for this progression being driven by technology.
Consider without this kind of system there is considerable amount of opacity that exists between organisational units. Corruption functions well as long as this kind of isolation is maintained for one simple reason: usually in this kind of arrangement, your gain is somebody else's loss.
If the network of corruption is too large you have to spread the gains much too thinly to make it worth the effort. Likewise, the more distant someone in that network is from the originator of said corruption, the more like they are to have very different goals and ideals and therefore disagree with and expose this kind of corruption.
That's why I think this kind of system has a considerable degree of resistance to misreporting of information. Certainly not perfect, but it makes corrupt behaviour much more difficult and risky.