User blog comment:Upquark11111/An Explanation of Loader's Number/@comment-11227630-20171210002438/@comment-11227630-20171210053253

In that paper, simply typed lambda calculus (i.e. λ→) was very weak - it can only hold elementary recursive functions, which growth rates are limited by tetration (or $$f_3$$).

It seems that the only difference between λ→ and λPω is the (s1,s2) rule, but how is λ→ so weak?