Gemlite: Towards Building Custom Low-Bit Fused CUDA Kernels(mobiusml.github.io)40 points by un_ess 2 years ago | 2 comments
Inside vLLM: Anatomy of a High-Throughput LLM Inference System (2025)(aleksagordic.com)127 points by sebg 2 months ago | 8 comments
LeMario: Training a JEPA World Model on Super Mario Bros(benjamin-bai.com)119 points by kevinjosethomas 3 months ago | 13 comments
LLMs could be, but shouldn't be compilers(alperenkeles.com)104 points by alpaylan 8 months ago | 112 comments