Heres how I see it.. We build the classifier off of trusted sources and then users with low accuracy can be weeded out. Once I do that we can post it to hacker news