At last year’s LLVM Developer Conference, we presented our work on loop vectorization and highlighted the lack of certain loop optimisations as a key factor limiting Clang/LLVM’s ability to generate competitive code for scientific applications compared to other compilers. After another year of effort—and following in the footsteps of many before us— we still don't have any of the loop optimisations running by default, so it’s fair to ask whether loop optimisations in LLVM are a mission impossible. In this follow-up talk, we’ll share what we’ve learned this year, what we’ve achieved, and what’s next, as we explore whether this challenge can truly be overcome.