It’s a really odd take given that by 2015 Apple was already on Arm v8 for the iPhone and the performance gains over v7 were clear. Calling it ‘hopeless’ seems incredibly wide of the mark.
ARMv8 is a well designed architecture that I don't think is "RISC" because eg it has some complex addressing modes, SIMD, and other good ideas.
It is more RISC than ARMv7 was. And I think if it was targeting smaller CPUs with less wide decoders, being more x86-like ("CISC") and having variable length instructions would be better.
Even accepting the distinction (which I don't agree with) between "non RISC-y" ARM (v7 and earlier) and "RISC-y" (v8 and later), Apple M1 is "RISC-y" by that definition (instruction set ARMv8.5A)[1]. And it's doing well.
[1] Line 256 https://github.com/llvm/llvm-project/blob/main/llvm/include/...