You know systems. You have read logs at 3 AM, traced a request through services you did not write, and formed opinions about other people's abstractions. What you may not have is the machine learning background that most writing about AI models quietly assumes. This course removes that assumption.
Foundations is a short, engineer-first course: eight chapters, roughly ten minutes each, that take a working cybersecurity engineer or software engineer from no ML background to the point where the machinery inside modern language models is legible. Analogies come from systems and security work, and each one states where it breaks. Every chapter ends with a self-check (answers collapsed, no peeking required) and a Go deeper list of at most five resources, each one verified live.
The fast path: Chapters 1, 3, 5, and 6, in order, unlock the whole tome. Chapters 2 and 4 are detours to take when Part 5 of the masterclass starts spending compute and the bills need explaining.
Model-Making Masterclass: a stage-by-stage comparison of how modern open models are actually made, from data and tokenizers through pretraining, post-training, evaluation, and serving, ending in the cyber read. Foundations exists so you can read it without getting lost.
The unlock map
Which chapters a reader needs before which parts of the masterclass. Each part link goes straight to that part on the masterclass site.
| Prereq chapter | Unlocks masterclass parts |
|---|---|
| Ch 0: The Map | Orientation for all parts |
| Ch 1: Models and Tokens | All parts |
| Ch 2: Just Enough Math | Part 5, Part 6, Part 8 |
| Ch 3: Transformers | Part 3, Part 4, Part 13 |
| Ch 4: Training at Scale | Part 5, Part 10 |
| Ch 5: Post-Training | Part 7, Part 8, Part 11 |
| Ch 6: Reading the Artifacts | Part 1, Part 9, Part 12, Part 13 |
| Ch 7: Second Gear | Part 6, Part 8, Part 9, Part 11 |
Start with Chapter 0, or open any chapter. Search loads a plain-text index of all eight chapters on first use.