If there is 36% increase in features shipped with half the team - that is a win.
Even if the new (half) team produces the same number of features it is a win.
Why is PR count or number of issues relevant for the overall outcome?. The article is trying to force fit the point their point about AI effectiveness