I think it’s more than that, Julia exists and adoption is still slow. Lua and torch were plenty fast and they were still replaced by pytorch. I think to compete with python you need at least a fraction of the de-facto corporate sponsorship for python in the ML space.