How AI60 works — the complete guide

AI60 takes you from the fundamentals to a job-ready AI Engineer in 60 days. Each day is one module. A module has three parts, always in the same order:

1. 1Learn — read the in-app lesson and watch the videos (they're embedded right on the page).
2. 2Checkpoint — a short quiz on what you just learned. Score 70% to unlock the next day.
3. 3Ship — do the homework and push it to GitHub. You get instant feedback.

Some days also have a Deep Track — a bonus video series (like “Build an LLM from Scratch”) shown on the day. Watch it alongside the lesson; spread it across a few days if it's long.

Before Day 1, you set up your computer and your GitHub. This takes ~30 minutes and you only do it once. There's a guided checklist at the setup page — here's the summary:

a) Create free accounts

1. 1A GitHub account (most important — this is where your homework lives).
2. 2OpenAI, Anthropic, and Hugging Face accounts (free).

b) Install your tools

1. 1Python 3.11+, VS Code, and Git.
2. 2In VS Code, install the Python extension and GitHub Copilot (free tier).

c) Get the starter code

1. 1Download the AI60 starter kit — a ZIP with every day's code file (each has TODO(Day N) markers you'll complete).
2. 2Unzip it — you now have an ai60 folder on your computer.

d) Put it on your GitHub (your submission repo)

1. 1Create a new public repo named exactly ai60 at github.com/new. Don't add a README (you already have files).
2. 2Open a terminal in your ai60 folder and run these (replace YOUR-USERNAME):

cd ai60
git init
git add .
git commit -m "Day 0: setup"
git branch -M main
git remote add origin https://github.com/YOUR-USERNAME/ai60.git
git push -u origin main

Refresh your GitHub repo page — you should see all the files. That's it, setup is done forever.

A few things that trip up beginners. Read these once — they'll save you hours.

Install Node.js (needed for a few tools)

Some evaluation tools (like promptfoo, used from Day 20) run on Node.js, not Python. Install Node.js LTS once — then commands like npx promptfoo@latest will work.

How you run your code each day

Inside your ai60 folder, use uv (installed in setup): uv add openai to add a library, and uv run python yourfile.py to run your script. That's it — the same two commands every day.

Free GPU for fine-tuning days (Days 38, 41)

Fine-tuning needs a GPU you probably don't have — use free Google Colab. Open the day's notebook, then Runtime → Change runtime type → T4 GPU. Colab's free tier can disconnect after a while, so save your outputs as you go.

Web search for the agent projects (Days 30, 35, 37)

Agents that “search the web” need a search tool. Two beginner-friendly options: Tavily (free tier, made for AI — get an API key) or the duckduckgo-search Python library (no key needed). Pick one and your agent can search.

Embeddings (Day 22 onward)

To turn text into vectors you can either use a free local model — pip install sentence-transformers — or OpenAI's text-embedding-3-small (needs the billing above). The free local model is perfect for learning.

Your cohort lives on Discord— it's where announcements, live-class reminders, and help from instructors and fellow students happen. Joining takes about a minute, and it's where you'll get answers fastest when you're stuck.

How to join

Click the Join Discord button on your dashboard or the Live page. That opens Discord and drops you straight into the AI60 server. You can use it in three ways — pick whatever suits you:

1. 1In your browser — nothing to install. Click the link, choose “Continue in browser”, pick a display name, and you're in. Best for trying it quickly.
2. 2Desktop app (recommended) — the smoothest experience for live classes and screen-sharing. Download for Mac or Windows, install, sign in, and the AI60 server appears in your left sidebar.
3. 3Phone app — install Discord from the App Store / Play Store so you get live-session reminders and replies on the go.

Where to ask what

Post questions in the help channel — an instructor (or a fellow student who's solved it) will answer. For anything about your grade or enrolment, message an instructor directly. Live classes are announced in the community and you join them right from the Live page in this app.

Open today's day from the sidebar. Work top to bottom:

1. 1Lesson — read it and watch the video. Click “I've completed this lesson” when done.
2. 2Checkpoint — answer the quiz. 70% unlocks the next day. You can retry.
3. 3Ship the homework — see the next section.

This is where you actually learn. Every day's “Ship the homework” step tells you exactly what to do. The flow is always the same:

1. 1Read “What to build” and tick off each requirement as you go.
2. 2Open the day's file — it's named on the page (e.g. day03_api_cli/weather_cli.py). Open it in VS Code from your ai60 folder and complete the TODO(Day 3) parts.
3. 3Push it to GitHub — open a terminal in your ai60 folder and run (the page gives you a copy button):

cd ai60
git add .
git commit -m "Day 3: weather CLI"
git push

1. 4Submit — copy your GitHub link (your repo URL, or the specific file) and paste it into the “Submit for review” box, then click Submit.
2. 5Get feedback — within a few seconds you get a score and personal feedback on your actual code. If it says “revisions requested”, improve it and submit again.
3. 6Check the reference solution — it unlocks after you submit, so you can compare your approach.

“Git” is how you save your work to GitHub. You'll use the same 3 commands every single day. Run them in a terminal, inside your ai60 folder:

git add .                       # stage all your changes
git commit -m "Day 3: done"     # save a snapshot with a message
git push                        # upload it to GitHub

The first time you push, GitHub may ask you to sign in — use a Personal Access Token as the password if prompted, or install GitHub CLI and run gh auth login once. Full basics: GitHub's Git guide.

1. 1When you submit, your actual code is reviewed against the day's rubric and you get a score out of 100 with specific feedback.
2. 2Your mentor/instructor also sees every submission and can add or adjust the feedback.
3. 3Below 70 = revisions requested — fix it and re-submit. There's no limit on attempts.

1. 1Sequential unlock — pass each day's checkpoint (70%) to unlock the next.
2. 2Ship every day — homework is how you actually learn; always push it to your ai60 repo.
3. 3Keep your repo public — it's your portfolio and how work is reviewed.
4. 4Explore in notebooks, ship in scripts — deliver graded work as real .py files.
5. 5Pace is flexible — take the time you need; the 60 days is a target, not a deadline.