From Zero to GPU: A Guide to Building and Scaling Production-Ready CUDA Kernels

PolyKAN: Effiziente GPU-Operatoren für Polynom-KAN-Varianten

Migrating AI from Nvidia to Huawei: Opportunities and trade-offs

Sigmoidal Scaling Curves Make Reinforcement Learning RL Post-Training Predictable for LLMs

SBVR: Neue Quantisierungsmethode für schnelle LLM-Modelle

Software Frameworks Optimized for GPUs in AI: CUDA, ROCm, Triton, TensorRT—Compiler Paths and Performance Implications

Gemini 3: Google’s neue KI könnte OpenAI und Nvidia ernsthaft bedrohen