What Are Large Language Models?

Everyone is using large language models. Very few people can say what one actually is. This book closes that gap — in plain English, without the maths, and without pretending the technology is either magic or a trick.

It starts where the confusion starts: a model does not look anything up. It reads your words as tokens, turns them into numbers, and predicts what comes next, one piece at a time. From that single idea, everything else follows — why models are fluent, why they are fast, why they hallucinate, and why a longer conversation starts to forget its own beginning.

What's inside

• How a model works — tokens, embeddings, and the transformer's attention breakthrough, explained without a single equation.
• How a model is made — pretraining, fine-tuning, and alignment, and what each stage actually changes about the model's behaviour.
• What happens when you hit enter — inference, sampling, and the context window as working memory.
• Why models make things up — hallucination as a property of the design, not a bug in it, and what retrieval-augmented generation does about it.
• Tools, agents, and multimodality — models that do things, not just say things.
• The major models in 2026 — how they differ, and how to choose between them.
• How to write a good prompt — the habits that reliably change the output.
• How LLMs are reshaping search — why being cited now matters more than being ranked, and what GEO means for anyone publishing on the web.

It closes with a glossary of the terms you'll meet everywhere, and straight answers to the questions people actually ask.

Who it's for

Anyone who has to make decisions around this technology and would rather understand it than be sold it — founders, designers, engineers, marketers, and the merely curious. No background in machine learning is assumed.

---

Based on the Troiana guide, What Are Large Language Models (LLMs)?