What you can read today

Three pilot modules, each four lessons long, with an interactive to play with:

• M02.4 Matrices as linear maps. See a matrix as a machine that moves the plane, and drag its columns to watch the grid follow.
• ML01.5 Gradient descent from scratch. Derive every gradient, take a step by hand, and find the exact learning rate at which training blows up.
• RL01.6 Bellman equations. Derive the equations under all of reinforcement learning and watch values spread across a gridworld one sweep at a time.

Every lesson has at least three purpose-made figures, a worked example with real numbers, code built from scratch, and questions with full solutions.

What is planned

The curriculum lists 73 courses in ten tracks, from school-level maths to research-level AI. The roadmap explains how it ships in four rings and is honest about the size of the job.

What “in review” means

The lessons are marked in review. Their maths and code are tested automatically on every change, and their sources have been opened and checked. They have not yet been signed off by independent readers. That is the purpose of the pilot.

If you find a mistake, or a step that lost you, please open an issue. If you can review the maths of one lesson, the contributing guide says how.

It is free

Every lesson is free to read and will stay free. No sign-up, no tracking.