learn.zerwiz.org · now in session

Courses for AI geeks
& freaks

A hands-on course for new people getting into coding with AI. We build real things, break them, fix them, and learn by doing. No fluff, no corporate jargon. Just code, models, and curiosity.

What does the community add? →
git first python next ship for real
~/learnai — zsh
12
tracks
78
lessons
32+
hands-on hrs

// the curriculum

Zero fluff.

Follow them in order, or jump around — your call. Every lesson has code you can run, break, and improve. Start at the top if you’ve never written a line of code.

01
Track 1 · The Ground You Stand On

Your Machine

Every step here was performed on a real machine and the outcome was recorded. Where something went wrong, the failure is written down too, because the failures are the lesson. You end with a working Linux machine with AI coding tools, a local model, and version control — from a bare stick of USB.

6 lessons165m
explore
02
Track 2 · The Machine Listens

Your Voice

Push-to-talk dictation and spoken replies, on the GPU where possible. Because it is the difference between typing and talking, you make the machine listen and answer — and you learn where the GPU is worth it and where it is not.

7 lessons175m
explore
03
Track 3 · The Flow

Your Coding Desk

A coding setup is not a list of programs. It is a flow: you read code, you run it, you ask an agent to change it, you look at what changed, you commit. Every tool below owns exactly one station in that flow. The beginner installs all of them and uses none. The journeyman uses the right one without thinking.

7 lessons180m
explore
04
Track 4 · The Colleague

Working With Coding Agents

A coding agent is the fastest colleague you will ever have, and the least experienced. It never sleeps, never sulks, and will confidently do the wrong thing at 3 AM. Everything in this course is about the two things that make that manageable: how you ask, and how you check.

9 lessons250m
explore
05
Track 5 · The Delivery Chain

GitHub and Getting Code Shipped

Writing code is half the job. The other half is: it lives somewhere, someone looked at it, and it reached production without anyone typing a scary word. That chain is commit → branch → push → pull request → review → merge → deploy. Learn it once and you can ship anything.

7 lessons190m
explore
06
Track 6 · From Folder to Repository

Your First Real Project

Everything so far was about your machine. This is about your work. The beginner's project is code in a folder. The journeyman's project is a repository with a shape: a written contract at the root, a plan on the shelf, an architecture nobody has to guess, a changelog that cannot lie, an agent that knows the house rules, and a registry that says where it lives and how it ships.

9 lessons245m
explore
07
Track 7 · The Machinery

Managing the Work in the Codebase

Most projects do not fail because the code was hard. They fail because the work was invisible: three people in the same branch, an issue nobody closed, a merge that happened at midnight and surprised everyone. This course builds the machinery once, so it costs nothing when the team grows.

9 lessons280m
explore
08
Track 8 · The Foundation

Git & GitHub

Every coder needs this first. Before models, before Python — learn to move code around. Branch, commit, push, and never break main again.

6 lessons69m
explore
09
Track 9 · The Language

Python for AI

Python is the lingua franca of AI. Learn it the right way — environments, data structures, and the real skill: reading error messages.

5 lessons80m
explore
10
Track 10 · The API

Talking to Models

Connect to AI models and make them do things. REST, JSON, prompting, streaming, and error handling — the rail that everything AI rides on.

5 lessons98m
explore
11
Track 11 · The Projects

Building with AI

Put it all together. Real projects, real deployment, real users. CLI tools, web apps, Cloudflare Pages, and auth that actually works.

4 lessons125m
explore
12
Track 12 · The Rabbit Hole

Going Deeper

For the freaks who want more. Run models locally, fine-tune, build agents with tools and MCP, and ship an AI project that's entirely yours.

4 lessons105m
explore
track 06 · coming soon

Your project. Fork this course, add a track, open a PR. Geeks respect contributions.

contribute on github →

// where this goes next

The lessons are free. The room is the thing.

Everything on this site stays free, with no account and no paywall — track 01 in full, plus the first lesson of every track. What we charge for is somewhere else, and we would rather tell you plainly than put a lock on a lesson: AI Geeks & Freaks is $49 per person, per month, and it is a group. A live cohort with real dates, the room where the work gets reviewed, and the self-paced path beside it.

A live cohort

Real dates, a real room, and a human who notices when you have not shipped. That is what the membership is really for.

The self-paced path

Every track, open to you the moment it is published. Miss a week, or join between cohorts, and you are never lost.

Help on your project

The courses are group courses. Hands-on help on your own specific project is quoted separately, per named project — and the teaching still comes with it.

// what runs the work behind all this

These lessons are written from machines that run Ymir, a single-operator agent OS: a fleet of named agents in sealed sandboxes, a memory well, parallel git worktrees, and an anti-hallucination gate. It is a hobby project, open and free — read it, fork it, use it.

https://aigeeksnfreaks.zerwiz.org/courses — course catalogue, and the door when it is open

// the community

Built by geeks, for geeks.

LearnAI is open, free, and a little bit stubborn. Fork it, contribute a track, or just lurk and learn. No certificates, no gatekeeping.

honest over hand-holding

We write for beginners but we don't talk down. Geeks respect honesty — if something is hard, we say so.

code over theory

Every lesson has runnable code. No theory without practice. If you can't run it, it isn't a lesson.

tinkerers welcome

The curious, the stubborn, the 2 AM crowd. If you break things to learn how they work, you belong here.

discussion prompts

  • ›What was the first thing you broke while learning to code?
  • ›Local models vs cloud APIs — when does each win?
  • ›Show your worst commit message. We won't judge (much).
  • ›What AI project would you build if you had a free weekend?

git workflow cheatsheet

The commands every geek keeps within arm’s reach. Save this.

bash
# 1. fork & clone
gh repo fork zerwiz/learnai --clone=true
cd learnai && git remote add upstream [email protected]:zerwiz/learnai.git

# 2. branch (never push to main)
git checkout -b feature/my-idea

# 3. work, commit (small, focused)
git add . && git commit -m "add learning module"

# 4. push & open PR
git push -u origin feature/my-idea
gh pr create --title "Add learning module" --body "one thing at a time"

# 5. sync your fork later
git fetch upstream && git rebase upstream/main && git push origin main
fork → branch → push → PR, the holy loop