#Ai

RTX Spark is not just another CPU story. It shows how NVIDIA is pushing ARM, Blackwell GPUs, unified memory, and CUDA into the next phase of AI PCs.

A Codex skill that turns chat decisions into reviewable proposals for AGENTS.md — so project memory stays compact, intentional, and inheritable across AI sessions.

A hands-on record of writing a training loop by hand, persisting the weights, and serving them through a small FastAPI endpoint, with an Colab notebook for running the version yourself.