A Chapter-by-Chapter Walkthrough of LLM Primer I

Series introduction and index for the twelve-part walkthrough of LLM Primer I: How Generative AI Works — published one post per day, February 18 through March 1, 2026.

Why this series exists

I wrote LLM Primer I to do something I couldn't find anywhere else: explain how large language models actually work, mechanism by mechanism, in a way that a curious general reader can follow but that an engineer would still respect.

Most introductions to LLMs fall into one of two camps. They're either marketing narratives that gesture vaguely at "AI magic" while telling you nothing useful, or they're research papers that throw matrix algebra at you in the first paragraph. The first kind leaves you no more informed than you started. The second kind doesn't actually help you reason about the technology — it just makes you feel intimidated.

The book takes a middle path. It explains every important concept clearly, with the technical precision intact, while staying readable. The 2026 edition adds "In Plain English" sidebars throughout, so anyone — regardless of math background — can follow the entire book.

This twelve-post series is a daily walkthrough. Each post previews one chapter, shares the central ideas, and explains why I structured the chapter the way I did. The posts work either as a standalone tour of the field or as a guide for deciding whether the book itself is right for you.

Who this is for

If you've used a chatbot and wondered what's actually happening underneath, this series is for you. If you're an engineer who wants a durable understanding of LLMs rather than another batch of prompt tips, this series is for you. If you're a manager or executive who has to make decisions about AI tools you didn't build, this series is for you. If you're a student looking for an accessible on-ramp to a serious technical field, this series is for you.

What you don't need: a math background, a programming background, or any prior exposure to machine learning. The book itself is written at two levels — a technically careful main text and a plain-English sidebar that runs alongside it — and this series leans toward the plain-English side.

The twelve chapters

Part I — Concepts & Foundations

Chapter 1 — What Is a Large Language Model? (Feb 18) — What the words "large," "language," and "model" actually mean, how we got here historically, and the three biggest myths about LLMs that lead to bad decisions.

Chapter 2 — Probability, Tokens, and Text (Feb 19) — How LLMs convert text into tokens, why language modeling is fundamentally a probability problem, and how next-token prediction becomes everything else.

Chapter 3 — Neural Networks for Language (Feb 20) — Why feedforward networks couldn't handle language, how RNNs hit a wall, and what attention changed.

Part II — How LLMs Work

Chapter 4 — The Transformer Architecture (Feb 21) — Inside the engine of modern AI. Self-attention, positional encoding, encoder/decoder topologies, and scaling laws.

Chapter 5 — Training Large Models (Feb 22) — Where the data comes from, what the loss function does, how distributed training works, and why training a frontier model costs hundreds of millions of dollars.

Chapter 6 — Fine-Tuning & Adaptation (Feb 23) — From prompt tricks to instruction tuning to parameter-efficient methods like LoRA. And the alignment techniques — RLHF and its modern successors — that turn raw models into helpful assistants.

Chapter 7 — Beyond Next-Token Prediction (Feb 24) — Embeddings, semantic search, retrieval-augmented generation, and the move into multimodal inputs. How RAG actually keeps a model grounded in real documents.

Part III — Practical Perspectives

Chapter 8 — Using LLMs in Applications (Feb 25) — Chatbots, summarization, code generation, knowledge extraction, evaluation, and the rise of agentic systems where the model drives a tool-use loop.

Chapter 9 — Performance, Scaling, and Costs (Feb 26) — Model size vs capability, the latency–throughput trade-off, quantization, edge deployment, and why frontier-tier models are often the wrong choice even when you can afford them.

Chapter 10 — Safety, Ethics, & Trust (Feb 27) — Why hallucinations happen mechanistically, where bias really lives, how layered guardrails work, and why governance is the institutional layer that technical controls can't replace.

Part IV — Advanced Topics

Chapter 11 — Cutting-Edge Research (Feb 28) — Mixture-of-experts, retrieval memory mechanisms, native multimodality, continual learning, and the inference-time scaling paradigm that produced today's reasoning models. The 2026 edition's biggest content addition.

Chapter 12 — Building Your Own LLM System (March 1) — The final chapter. Datasets and licensing, training pipelines, evaluation frameworks, the integrated application stack, and the case-study patterns that distinguish successful deployments from failed pilots.

About the book and the series

The book is LLM Primer I: How Generative AI Works — A Clear and Practical Guide to the Foundations of Large Language Models by Sho Shimoda.

It's the first volume in the LLM Primer series. Other volumes in the series go deeper into specific topics — retrieval-augmented generation, context design with MCP, building real-world LLM applications, scaling AI systems, and AI security. Each volume stands on its own, but together they cover the full engineering picture of working with large language models.

If LLM Primer I leaves you wanting more depth on a specific topic, the companion volumes are where to go next.