User blog comment:Hyp cos/BEAF isn't so strong/@comment-5150073-20130921115609

The article in FGH is also wrong, as I believe. It is better to start with f(n) = {X,X / 2} & n = X & X & X ... X & X & X & n (with X X's). It is comparable to f_LVO(n). Then when we make the recursion around {X,X / 2}, and the same level of recursion will be performed around LVO. For example, {X,X / 2}^^X & n ~ f_epsilon_(LVO+1)(n).

Then it implies for me that {X,2X / 2} & n ~ theta(W^W,1).