We considered this with NewsTilt, but never got far enough along to actually try it. The aim wasn't actually A/B testing them, but to use machine learning to pr

We considered this with NewsTilt, but never got far enough along to actually try it. The aim wasn't actually A/B testing them, but to use machine learning to predict how well the article would do given certain criteria: title, what time of day it was released, etc, to allow journalists to tweak those criteria. I think that was actually PG's idea.