> their best known approximation had pi ≈ 25/8, which is too small

I'm curious if they had a better one that we don't know of yet—their best known approximation of sqrt(2) is significantly more accurate.