A deep learning textbook that actually runs

What it does

D2L.ai is an open-source textbook—published by Cambridge University Press—that teaches deep learning through Jupyter notebooks. Every chapter mixes exposition, math, figures, and self-contained code you can execute, modify, and break. It is freely available under a Creative Commons license and aims to bridge reading about machine learning and actually doing it.

The interesting bit

Most textbooks treat code as a static illustration; here the manuscript is the repo. Drafting the entire book in notebooks means examples stay runnable as frameworks evolve, and readers can patch a typo or test a hunch without leaving the chapter.

Key highlights

• Entire manuscript drafted in Jupyter notebooks with runnable, self-contained examples.
• Adopted at 500 universities across 70 countries, including Stanford, MIT, Harvard, and Cambridge.
• Multi-framework code coverage alongside math and interactive discussion.
• Published in 2023 by Cambridge University Press; sample code carries a modified MIT license.
• Community contributions are merged directly into the living manuscript.

Verdict

Ideal for students and practitioners who want theory paired with executable homework; skip it if you are looking for a drop-in model zoo or the latest research implementations.