Zero GPU, zero dollars: a Linux rookie's crew of free AIs

A few weeks ago, the author had no knowledge of the command 'sudo'. Now, their laptop runs a variety of AI agents that manage their backups, home lab, and gadgets, reporting daily in character. These agents include Rick Sanchez and a British butler named Joshua, both of which do not cost any money and the laptop itself does not have a graphics card.

The author did not have programming experience and built everything by describing their needs to Claude, testing them, and refining them as needed. This post shares their learning experience for others starting from zero. The author follows four rules: the item must be free, prefer free and open-source software, use local solutions over cloud services, prototype before scaling, and only run things with explicit permission.

The setup includes a laptop with 15GB of RAM and no GPU, using Arch Linux + Hyprland as the operating system. AI models like Qwen3, Llama 3.2, Gemma 3, and Mistral run on the CPU through Ollama, with requests sent to free tiers of Groq, Gemini, Mistral, OpenRouter, and Cloudflare when online and local models otherwise. Each AI agent is a scheduled job in n8n, gathering information, writing reports in character, and adding them to the author's daily notes.

The setup includes a Rick and Morty themed dashboard, a British accent-enabled AI butler named Joshua, and a collection of models. All projects are public, free, and MIT licensed, with a script to clean and scan for personal information before publishing. For beginners, the author emphasizes building a safety net first and encourages describing what they want in plain words for AI coding tools to turn into working code.