CUDA-l2: Surpassing cuBLAS performance for matrix multiplication through RL

(github.com) by dzign | Dec 4, 2025 | 15 comments on HN