AI developer @teortaxesTex predicts DeepSeek-V4.1 will receive five to eight times more reinforcement learning compute than GLM-5.2 · Digg