Use Your Mac for AI Agents: Self-Host Gemma 4 12 B with Pulumi and Tailscale

Open-weight models now run well on consumer hardware, allowing data to stay local and inference to work offline. The Gemma 4 12 B model can be run on a modern Mac, providing high-quality language processing capabilities. This setup uses llama.cpp for host-native inference, k3d for a local Kubernetes cluster, Pulumi for infrastructure as code, and Tailscale for secure access.