An AMD developer acceleration team lead presents 'playbooks' — open-source, step-by-step developer recipes for setting up AI development environments on Windows and AMD hardware. Central to the talk is Lemonade, an open-source tool in the ROCm stack that abstracts GPU/NPU/CPU routing and provides an OpenAI-compatible inference endpoint. Key themes include: running local AI coding agents (e.g., Codex, Claude) on AMD AI PCs without cloud connectivity, dynamic routing of inference workloads between local and cloud based on cost, privacy, or task complexity, and omnimodal model collections that intelligently route queries to text, image generation, or speech models. A headless 'Halo box' appliance for always-on agentic tasks is previewed, and developers are directed to developer.amd.com for playbooks, workshops, and cloud credits.