How this book was made and checked

Who it is for. A senior software developer moving into ML and AI engineering: fluent in systems, new to models. Each essay explains one mechanism until you can recompute its worked example, then gives you an exercise with an expected result. Read them in any order.

Where it comes from. The author is enrolled in seven Udemy courses and owns the books named on each essay's source line. The essays are study notes made rigorous: the mechanisms are re-derived in the book's own words and arithmetic, the courses and books are credited by name, and nothing from them is reproduced beyond a phrase. Where a number is a lecturer's, the essay says "as reported in the lecture". The book is not a substitute for the courses and does not pretend to be.

What is checked by machine. Every lecture an essay cites exists as a transcript file on the author's machine; every book page cited is within the book's page count; no twelve-word run of any cited transcript appears in any essay; every page builds with one title, one exercise and no broken link.

What is checked by reading. Two independent model reviewers and a consolidator look for missing steps, unreceipted numbers, wrong arithmetic, close paraphrase and framework-as-mechanism; four reader personas report where they got lost. Nothing they say is applied automatically. The author reads every essay before it is published.

What is not checked, and cannot be. That the lecturer is right: mechanism claims are re-derived on the page, the rest is reported. That any code snippet runs: snippets are illustrative unless the essay says a run log exists. That the mechanism supports the lesson drawn from it: only reading catches that. Whether a tool or API still behaves as described: product claims carry an "as of". Whether you learned anything: the exercise has an expected result, and that is all the book claims.

Nothing is sold. No newsletter, no course, no consulting. The book ends and that is all it does.

Written by Claude, reviewed by Gemini and ChatGPT models, edited and published by Anthony Kot.