Same :) I started down the rabbit hole of abstracting CUDA for our robotics framework, but it’s not really something I want to maintain right now.