The Rules Are Half the PhysicsWhy world models are hard in regulated industries, and why the rulebook is also the way through25 min read·blog.hammer.aiAug 5
Implementing TurboQuant in llama.cpp: CUDA Scars and What Actually Ships3 weeks of porting TurboQuant to CUDA, 5 scars, and what actually ships for document processing on T4 GPUs11 min read·blog.hammer.aiApr 6