From vectors to Triton: an open textbook for the stubbornly curious

What it does

Henry Ndubuaku turned years of AI/ML notebooks into an open, 18-chapter textbook that runs from vector spaces and calculus all the way to CUDA kernels, MLOps, and speculative decoding. The material is written for practitioners who want the underlying machinery explained in plain language rather than survived for an exam. It also ships with an MCP server so Claude, Cursor, or VS Code can query the repo directly as a knowledge base.

The interesting bit

The scope is deliberately maximalist: it treats linear algebra, computer vision, NLP, audio processing, graph neural networks, and SIMD/GPU programming as one continuous curriculum rather than isolated specializations. The author claims friends used these exact notes to interview successfully at DeepMind, OpenAI, and Nvidia, which either speaks to the notes or to the friends.

Key highlights

• 18 chapters are live, covering everything from probability and classical ML to Triton, TPUs, and ML systems design.
• Two final chapters—Applied AI and Bleeding Edge AI—are marked as coming soon.
• Includes an MCP server integration, turning the repo into a queryable knowledge base for AI coding assistants.
• The author frames the work as a deliberate alternative to textbooks that “bury good ideas under dense notation” and quickly go stale.
• Entry barrier is intentionally low: elementary maths and basic Python are the only stated prerequisites.

Caveats

• Chapters 19 (Applied AI) and 20 (Bleeding Edge AI) are listed as “Coming” with no delivery date.
• The README’s foreword is an extended personal essay on IQ, childhood grades, and Serbian psychology studies; readers will need to scroll to reach the actual syllabus.
• The claim that friends passed interviews at DeepMind, OpenAI, and Nvidia using these notes is presented without corroborating detail.

Verdict

Worth bookmarking if you are a self-taught engineer or researcher who wants a single, opinionated map from undergraduate maths to production ML systems. Skip it if you need formally peer-reviewed material or a quick reference card; this is a marathon, not a cheat sheet.