04-AI-C2

2 min read

Module 4: The Force Multiplier (Local AI)

The Instructor's Perspective

In 2026, AI is no longer optional—it’s a teammate. But we don’t want a teammate that reports everything to the cloud. We are building a Local-First AI Stack. This gives you privacy, zero cost-per-token, and the ability to work even when the internet is dark.

🎯 Learning Objectives

By the end of this module, you will:

  • Deploy Ollama as your local model engine.
  • Connect OpenClaw as your Command and Control (C2) orchestrator.
  • Offload a terminal task to a local model (e.g., Llama 3 or Qwen).

🧭 Tactical Overview

Your AI stack has two parts: the Engine (Ollama) and the Dashboard (OpenClaw). Ollama handles the heavy math on your Intel Arc iGPU, while OpenClaw gives that math “hands”—the ability to read your files, run shell commands, and help you manage your lab.

🛠️ Step-by-Step Implementation

1. Hardening the Engine (Ollama)

Deploy Ollama using our Practical AI-Stack (Source: growlf/ai-stack). This repository provides the production-ready logic to ensure Ollama can “see” your GPU hardware. Run ollama run llama3 to verify the signal.

2. Establishing C2 (OpenClaw)

Install the OpenClaw gateway. This is your “Tactical Radio.” Link it to your local Ollama instance and (optionally) a cloud fallback like Gemini for complex tasks.

3. The First Mission

Task your AI to audit your lab: “OpenClaw, list all running LXCs on Proxmox and check if any are low on disk space.” If the AI can answer, your Force Multiplier is online.

⚠️ Common Pitfalls & “Stupid Questions”

  • Pitfall: Running LLMs on the CPU only.
  • Fix: Ensure your Docker compose file passes /dev/dri to the container for GPU acceleration.
  • “Stupid Question”: “Which model should I use?”
  • Answer: Start with llama3:8b for general chat and qwen2.5-coder for help with your Terraform/Ansible scripts.

Knowledge Check

  • What is the advantage of a “Local-First” AI strategy?
  • What role does OpenClaw play in the stack?
  • How do we verify the AI can “see” our hardware?

Next Module: The Tactical Notebook (Documentation) Status: Operational Related: Sparse Page, OpenClaw