Learn AI from scratch, completely free.

Fixed answer keys, approved by people.

  Try a whole lesson: Tokens and subwords, with four tasks →

Each level is a record of the tests you passed.

Every level gives a diploma that shows which tests you have passed, with the number of tasks, the test version and the date.

1. Level 1

   First Light

   Pass condition: pass the First Light test: at least 40 of 50 tasks, including all 10 essential ones.

2. Level 2

   Navigator

   Pass condition: pass all four course tests in one track.

3. Level 3

   Expedition

   Pass condition: pass three tracks and show a piece of your own work in your portfolio. The work is shown, never graded.

4. Level 4

   Zenith

   Pass condition: solve a hard task that today’s AI models could not, checked by fixed automated tests, and publish the code so anyone can run it.

Take Google’s free courses first, then continue at Polaris.

Our courses open later. These courses are open today and teach the basics ours build on. The first four take 3 hours 15 minutes in total, as video and short quizzes.

1. Introduction to Generative AI45 min · FreeIntroduction to Large Language Models1 h · Free
   Then at PolarisF01How Language Models WorkOpens later
2. Introduction to Image Generation30 min · Free
   Then at PolarisR06Image, Sound and VideoOpens later
3. Introduction to Security in the World of AI1 h · Free
   Then at PolarisR04SecurityOpens later
4. Machine Learning Crash CourseOpen page, no price shown
   Then at PolarisR09Under the HoodOpens later

Polaris is independent and not affiliated with Google. Lengths and prices as shown on Google’s pages on 1 November 2026. Google may ask you to sign in. The security course was last updated over a year ago.

Seven foundation courses for everyone who starts.

Estimated 70 study hours, self-paced. The courses are being written. Course texts and tests open later.

1. F01How Language Models Work
2. F02300 Key Terms in AI, Code and Operations
3. F03Structured Prompting
4. F04Checking AI Output
5. F05Computer Craft: Files, Terminal, Git and Editor
6. F06Money, Tokens and Quotas
7. F07Law, Privacy and Responsibility Basics

Choose a track. Each track has four courses.

1. R01Prompt and Language
2. R02Code
3. R03Operations
4. R04Security
5. R05Web Design and Interfaces
6. R06Image, Sound and Video
7. R07Agents and Automation
8. R08Local AI and Own Models
9. R09Under the Hood
10. R10AI as a Business

Course contents for the tracks are being written. They open later.

Start today with the prompt course.

The Prompt Course · 5 lessons · Free. Lessons 1–3 are open directly; lessons 4–5 are sent by email.

1. 01Stop prompting like it's a Google searchOpen
2. 02The Power of ContextOpen
3. 03Iterate, iterate, iterateOpen
4. 04System prompts & personaRequires email
5. 05Advanced techniquesRequires email

How to become a RAG and Agentic Expert

A planned course series about RAG (letting a model search your own texts before it answers) and simple agents. Part 1, “Build your first searchable assistant and agent”, is drafted: you build one small assistant over your own documents that answers with a source or says “I don’t know”, is measured with your own test questions, and can use tools as a simple agent. It runs locally on your own computer.

1. 1Ask without RAG and watch the model guess
2. 2Split the text and search by keyword
3. 3Embeddings: search by meaning
4. 4Answer with a source, or “I don’t know”
5. 5Measure with ten test questions of your own
6. 6Improve: chunk size and re-ranking
7. 7An agent that chooses tools
8. 8Errors and security

Draft outline. Nothing in this course is published yet. Part 1 covers the basics; later parts go deeper.

A teacher that explains and gives hints.