Concepts
A direct entry point to the ideas behind modern AI.
These guides answer the essential question first and then open up the mechanism. Each concept links to the chapters and engineering notes where it is applied in practice.
Technical foundations
01
What is an LLM?
Tokens, pretraining, generation, alignment and the limits of large language models.
02
How the Transformer works
Attention, representations, residual blocks and the architecture that enabled modern models to scale.
03
Reasoning in LLMs
Chain of thought, search, verifiers, test-time compute and why more steps do not guarantee a better answer.
04
Evaluating AI models
How to move from an isolated benchmark to a valid evaluation of the system and product.
05
What is an AI agent?
Chatbot vs workflow vs agent, tools, memory, state, evaluation, permissions and when agency is actually useful.
06
What is prompt injection?
Why data and instructions can compete inside context, and which architectural boundaries reduce the path from untrusted content to action.
Next level
From concept to system.
The learning series develop each idea progressively. The engineering notes show what changes once latency, state, tools, audio and human interaction become part of the system.
Explore the series →Open the engineering notes →