Local AI Crash Course: Run Models Locally and Do Real Work

Most people assume running real AI means a monthly subscription and an API bill that grows every time you use it. It doesn't have to. The open models you can download for free today are roughly where the paid models were a year or two ago, and for most everyday work, that is more than enough.

This is a hands-on crash course on running AI models on your own computer. No subscription, no API key, and your data never leaves your machine. By the end, you will have a real model running locally and connected to your actual work, not just a chat window you try once and forget.

We start from nothing and keep it simple. You will learn the four decisions that actually matter:

• Hardware: how to tell exactly what your Mac or PC can run, using a simple memory lookup table.

• Runner: which tool to use (Ollama, LM Studio, or llama.cpp), and why the choice matters less than people fear.

• Model: how to read confusing model names, what quantization and Mixture of Experts mean in plain words, and which models to start with.

• First task: install a runner, run your first model, and put it to work.

Then you will connect your local model to real tools (your code editor, a coding agent, a script, or a small app) through its local API. And because honesty matters, you will also learn when local AI is the wrong choice, and how to route the hard work to the cloud.

This course is for engineers and technical people who want AI running on their own machine and have not set it up yet. No prior local AI experience is needed, and every term is explained in plain words before it is used. If you already run llama.cpp from source, much of this will be familiar.

This is also a living course. The same machine and the same method run more than text, including image generation, speech to text, text to speech, and even video, and new modules are added based on what students ask for.