ParentFull threadsuperb_dev·My understanding is that MLX is Apple’s CUDA, so a CUDA translation layer would target MLXView on HN