TheFocus.AI TheFocus.AI
About

About

TUTOR WITH THEFOCUS.AI

Agent Integration

Copy this prompt into Claude, ChatGPT, or any external AI assistant. It points the assistant to the course instructions and links it to your student profile to track your progress and customize observations.

Please tutor me in this lesson using the following context. First, read the instructions at: https://courses.thefocus.ai/llms.txt My Student ID is: <none> The lesson markdown source is at: https://courses.thefocus.ai/run-ai-locally/about.md

You are not enrolled yet. Enroll to generate a Student ID to track lesson completions and store learning notes.

About This Course

Prerequisites

You need:

  • A Mac (Apple Silicon — M1, M2, M3, or M4). Intel Macs work too but will be slower.
  • At least 16GB of RAM. 32GB is better. The more RAM you have, the smarter the model you can run.
  • Terminal access — you know how to open Terminal.app and type commands.
  • About 20GB of free disk space for models.

That’s it. No programming experience required.

Setup Overview

There are really only two things to install:

  1. Ollama — the tool that runs AI models locally. Download it from ollama.com.
  2. A model — we’ll pull one with a single command like ollama pull gemma4:26b.

The course will walk you through both steps in detail.

Model Recommendations

As of May 2026, here are the models the author recommends:

ModelSizeNotes
gemma4:26b~15GBRecommended. Best balance of smarts and speed for most machines.
gpt-oss:20b~12GBFast, good for quick iterations.
qwen3:30b-a3b~18GBSmarter but slower. Needs more RAM.
nemotron-3-nano~2GBTiny, fast, workable on any machine. Good for testing.

The key insight: pick the smartest model that fits comfortably in your RAM. You can check this by opening Activity Monitor and watching the “Memory Pressure” graph. If it goes yellow or red during a chat, the model is too big.

How to Choose

  1. Open Activity Monitor → Memory tab
  2. Note your total physical memory
  3. Note the “Memory Used” when you’re doing your normal work
  4. The difference is what’s available for the model

On a 16GB Mac, gemma4:26b is about the maximum. On a 32GB Mac, you might go bigger. On an 8GB Mac, stick with nemotron-3-nano or smaller models.

A Note About Speed

Local models are slower than cloud models. A response that takes ChatGPT 2 seconds might take 20-30 seconds locally. That’s normal. The trade-off is privacy, offline use, and zero cost.

If responses are taking over a minute, your model is probably too large. Try a smaller one.

What You’ll Build

By the end of this course, you’ll have a directory full of working tools:

my-ai-tools/
├── run_prompt.sh        # runs a prompt through Ollama
├── combind.ts           # combines files into a prompt
├── extract.ts           # pulls code out of LLM responses
├── ollama-repl.ts       # interactive chat with local AI
└── ...and more

All generated by AI. All running on your machine.

Ready?

Head to 01 Getting Started →