Sakana AI and NVIDIA Introduce TwELL with CUDA Kernels for 20.5% Inference and 21.9% Training Speedup in — 2026-05-11
Memori Labs has developed a new persistent memory infrastructure for AI agents, allowing them to retain context across user sessions. Sakana AI and NVIDIA have introduced TwELL, an open-source format and CUDA kernels that improve LLM efficiency by optimizing GPU usage, leading…