Ok, why don't you run this experiment on a large public open source code base? We should be drowning in valuable bug reports right now but all I hear is hype.