Theoretical Foundations of LLMs

From token to agent

AI already writes, decides, and advises alongside you, and it does so with the same confidence whether it is right or making things up. What separates those who use it well from those who merely trust it is one thing: understanding what happens inside. This book rebuilds, from a conceptual perspective, the mechanisms that underpin Large Language Models (LLMs) and the systems built on them, from token to agent. It is not a manual of tools, which change every month, but a theoretical foundation of the principles that endure. By understanding why a model answers the way it does, and where it fails, you trade blind faith for judgment: you know when to trust the output, when to verify, and when to intervene. It is that critical reading, not the prompt trick of the moment, that turns AI from a black box into a tool you command.

Who it's for

For people who use AI in decisions that carry weight.

Technical professionals, researchers, consultants, analysts, and managers who apply LLMs to work with real impact, where a wrong answer is costly and the judgment of when to trust, verify, or intervene makes the difference. If you rely on AI to build, decide, or advise, this book gives you the foundations to do it with judgment, not on faith.