These are the highlights IMO:

> Avoiding monoculture preserves the motivation for consensus-based standards processes rather than single-party control

> there shouldn't be a "LLVM libc", or any one library implementation that's "first class" for use with LLVM while others are only "second class".