I design my home lab like a puzzle where all the pieces can be added or removed. When I use AI, I want to run local models when I can, and only use cloud models when it is needed. - Zo Computer, Prototyping. Runs cloud, and local models - MSI EdgeXpert - DGX Spark, AI Supercomputer, Local model serving - Mac Studio, M2 Max 32GB, Agent host, local embeddings - Mac Mini, M4 24GB, Agent host, local embeddings - MacBook Pro, M5 16GB, mobile, connectivity to the lab - HP Z2 G9, i9 64GB, 8GB VRAM, Workstation (dual boot) Windows/Ubuntu - Raspberry Pi, RPi 4 2GB, Always-on network node, VPN exit, subnet routing - Synology NAS - Tailscale, Mesh VPN - one private network across every machine AI Models - DeepSeek V4.1 Flash, Cloud - qwen3.6-35b-a3b, Spark - local model serving - nemotron-3.5-lightning:30b-a3b, Spark - local model serving - text-embedding-nomic-embed-text-v1.5, LM Studio, on the computer that runs Agent Zero Harnesses - Zo Computer, Cloud hosted, Prototyping - Hermes, A self-improving, open-source AI agent that runs persistently on your server or machine and remembers what it learns across sessions. - Agent Zero, A framework for building autonomous AI agents that can operate independently, learn, and execute tasks across various systems. I worked in IT Operations and never used AI specifically, so this year I bought hardware for AI, to add to my home lab. What am I working on? - AI memory including context in chat - Robotics and AI interface - Testing other AI harnesses like OpenHuman and Tiny Humans as long as I can host locally