RTX 5080 and RTX 3090 Setup: 80 Tok/s on Qwen 3.6 27B Q8

Zero-Copy GPU Inference from WebAssembly on Apple Silicon

Taking on CUDA with ROCm: 'One Step After Another'

MegaTrain: Full Precision Training of 100B+ Parameter LLMs on a Single GPU

Scaling Karpathy's Autoresearch: What Happens When the Agent Gets a GPU Cluster

A CPU that runs entirely on GPU