This article was translated from the original Korean source.
AiBook: A Static Web Book for Relearning AI
AiBook is a static web book project I created to revisit AI from the ground up.
I am structuring it so that people studying AI for the first time, those who once took an introductory AI course but have become hazy on the concepts, and non-specialists who have used AI tools and want to understand their inner workings can follow the same path.
The goal of this project is not to collect a large volume of material.
It is to create a path for relearning: AI Introduction and Landscape -> Rebuilding the Foundations -> Data Modeling -> Machine Learning -> Deep Learning -> LLMs and Generative AI -> Projects.
Why Learn It Again?
AI is changing so quickly these days.
New models, APIs, tools, and frameworks keep appearing. If we only keep up with the surface of each new development, important questions remain:
- Where does this technology fit within the broader development of AI?
- How are machine learning, deep learning, generative AI, and LLMs connected?
- What roles do prompts, embeddings, RAG, and agents play in a service architecture?
- How much mathematics and Python do I need to revisit to understand actual model computations?
I believe that relearning AI requires more than following tutorials for the latest tools.
A feel for using tools matters, but first we need to understand where concepts fit and which concepts depend on others. That makes it possible to judge whether a new technology is entirely new, an extension of an existing structure, or simply a new name.
How the Book Is Organized
AiBook is organized into seven parts.
- Part 1. AI Introduction and Landscape: Establishes AI’s scope and history, the difference between rules and learning, and the basic elements of using LLMs.
- Part 2. Rebuilding the Foundations: Revisits mathematics, Python, data tools, and execution environments to the extent needed to understand AI model computations.
- Part 3. Data Modeling: Organizes raw data into inputs suitable for training and result structures that allow comparisons.
- Part 4. Machine Learning: Examines problem formulation, data splitting, training, evaluation, and representative algorithms within a common structure.
- Part 5. Deep Learning: Covers the progression from neural network computations and backpropagation to CNNs, RNNs, attention, and Transformers.
- Part 6. LLMs and Generative AI: Connects tokens, prompts, generation settings, embeddings, search, RAG, agents, and evaluation.
- Part 7. Projects: Turns the concepts learned earlier into small deliverables and verifies them using execution logs and evaluation criteria.
This sequence reflects learning dependencies, not just a table of contents.
First, readers get an overview of the AI landscape, rebuild the foundations needed to understand model computations, and then transform data into a structure suitable for training. On that basis, they explore the common structures of machine learning and deep learning before moving on to today’s LLM and generative AI service architectures.
Writing for Non-Specialists
AiBook does not assume prior knowledge of university-level mathematics, programming, or systems fundamentals.
At the same time, I try to avoid reducing everything to simple analogies. Wherever possible, I write key terms in both Korean and English, making it easier to find them later in papers, official documentation, international lectures, and API documentation.
For example, terms such as 학습(learning), 모델 실행(inference), 표현(representation), 임베딩(embedding), and RAG(retrieval-augmented generation) may feel unfamiliar at first. But avoiding these words ultimately makes it impossible to read AI documentation.
So instead of removing terminology, AiBook first introduces situations where a term is needed, then repeatedly connects it to its meaning.
Managing Documentation Like a Product
I want to manage AiBook as a documentation product rather than simply a collection of Markdown files.
The main text lives under docs/, and the table of contents is managed through MkDocs’ nav. To keep concepts from becoming scattered, I maintain a concept glossary, and each section has a Section ID and Version.
Outside the main text, I also maintain a separate management/ area. It contains writing guides, release notes, checklists for each part, tools for collecting supporting evidence, and tools for producing translation quality reports.
There is a clear reason for this structure.
An AI book is difficult to complete in one pass. Explanations need ongoing revision, external sources need to be checked for dates and context, and translations need another review to ensure their meaning has not drifted from the original.
Good writing alone is not enough to keep documentation open to continued revision.
Change history, review criteria, terminology standards, and the table-of-contents structure must be managed together.
Building with AI Tools
AiBook is being created with the help of AI tools.
AI helps with curriculum drafts, research, comparisons, diagrams, translation drafts, and document organization. Tools such as Codex can read files within the repository, create drafts, organize the structure, and assist with repeated reviews.
But a sentence generated by AI is something to review, not a final result.
Natural phrasing does not guarantee a correct explanation. A plausible summary does not guarantee sufficient evidence. That is why I try to distinguish between standard explanations, working hypotheses, and verification needed in AiBook.
Using AI while maintaining a process for reviewing what it produces.
This is one of the book’s important writing principles.
Closing Thoughts
AiBook is not a document for quickly consuming AI information.
It is a book for reestablishing where the AI technology I am looking at fits into the broader picture, before learning yet another tool.
For me, AiBook is both a personal learning notebook and an experiment in how to structure and operate a knowledge base. It is a book about studying AI, but it also incorporates questions about documentation, review, translation, release management, and automation.
If you are reading it for the first time, start with these two pages:
Go Home