01 · FIRST STEPS
Ollama + Gemma 2
Install Ollama, run a small Gemma 2 model, and try your first prompts.
Open the tutorial →A hands-on learning journal
Think it through. Then test it.
A practical path into local AI, language models, and agents. Start small, try each tool yourself, and build up one step at a time.
First experiment · Ollama + Gemma 2
$ let’s start locally
ollama run gemma2:2b
> One small experiment at a time.
Learn by doing · keep your own notes
Your learning path
Follow the sequence from a first local chat to experimenting with model runners and AI agents.
02 · MAKE IT FRIENDLY
Add a web UI
Connect a ChatGPT-style interface to your local Ollama model.
Open lesson →03 · UNDER THE HOOD
Explore llama.cpp
Learn another way to run local models and what happens beneath the interface.
Open lesson →04 · HARDWARE EXPERIMENT
Try a Qwen MoE model
Explore model size and quantization on a GTX 1060 with 32 GB of RAM.
Open lesson →05 · BUILD A BOT
OpenClaw + Telegram
Make a small bot that can answer questions and give instructions for local files.
Open lesson →06–07 · COMPARE
Hermes, OpenClaw + Codex
Try the tools on the same task and compare how they follow instructions and use tools.
Hermes lesson → · Compare tools →Ideas to practice throughout
These are useful across the whole path, so we’ll revisit them as each project gives us a reason to use them.
Prompt engineeringMake requests clearer, more efficient, and easier to evaluate. Open guide →
Agents working togetherExplore how multiple agents can split and solve a task. Open guide →
Ralph Wiggum workflowsLearn a bounded iterative agent pattern. Open guide →
Testing and reflectionUse the same small task to compare tools and keep notes on the results.
Start here
Ready for your first local model?
Lesson 01 walks through installing Ollama and chatting with Gemma 2.