Knowledge base

AI glossary in pictures

Terms that come up in every AI conversation: a definition, an example, an animation and sources.

Knowledge base

All terms

Each one has a short animation and a link to where you will see it in practice.

Basics

AI agent

An AI agent is a model that plans steps and uses tools by itself until it reaches a goal. I explain agent versus workflow and when an agent pays off.

Context window

The context window is the text a model sees at once. I explain why a larger window is not automatically better and how to manage what goes into it.

Embedding

An embedding is a numeric fingerprint of a text's meaning. I explain how it lets you search by sense, not words, and why it is combined with keyword search.

Evaluation (eval)

An evaluation is a set of tests that checks whether a model-based application works well. I explain success criteria, test cases and a held-out set.

Hallucination

A hallucination is a confident-sounding but false answer from a model. I explain where it comes from and how to limit its effects with sources and checks.

Human in the loop

Human in the loop is a setup where a human approves an agent's key steps. I explain when to ask and when to let it act automatically.

MCP

MCP is an open standard for connecting AI applications to tools and data. I explain host, client and server roles and caution about others' servers.

Price per million tokens

The price per million tokens is the standard API pricing: separately for input and output. I explain how to compute a task's cost and what changes the bill.

Prompt

A prompt is the instruction you give a model. I explain what it consists of, how to improve it and why a good brief beats clever wording.

RAG

RAG is a pattern in which a model first retrieves document fragments and then answers based on them. I explain the idea and when it is needed.

Token

A token is a chunk of text a model splits sentences into. I explain why limits, speed and cost depend on the number of tokens and how to count them.

Workflow

A workflow is a repeatable chain of steps that runs on its own. I explain workflow versus agent and the five patterns from Anthropic's write-up.

Pick one thing. Start doing it.

You do not have to know the whole road yet. Pick the area you want to put in order and check the scope of a workshop. That is where we start.

Choose a workshop See projects