As of October 22, ForgeKernel, an operator set fully written by AI with no manual line-by-line tuning, ranked first in both NVIDIA's SOL-ExecBench FlashInfer-Bench collection and Ascend's CANNBench A3 hardware leaderboard, with a baseline speedup of 1.835x, 25% ahead of the second place, and outperformed DeepSeek v4.1 Flash operators.