AI Agent
A model given tools, a goal and the ability to loop — deciding its own next action rather than answering one question.
AI glossary
16 terms explained without hand-waving, each with the practical consequence of understanding it — because most AI jargon exists to describe a trade-off you will eventually have to make.
architecture
Embeddings
Numerical representations of meaning that let software compare text by what it means rather than which words it uses.
MCP (Model Context Protocol)
An open standard for connecting AI assistants to external tools and data sources through one consistent interface.
RAG (Retrieval Augmented Generation)
A technique where a model retrieves relevant documents before answering, so its response is grounded in real sources instead of memory.
Vector Database
A database designed to store embeddings and find the most similar ones fast, at scale.
practice
Chain of Thought
Prompting a model to reason step by step before answering, which measurably improves accuracy on multi-step problems.
Prompt Engineering
The practice of structuring instructions and context so a model produces reliable, useful output.
Temperature
A setting that controls how random a model's output is — low for consistency, high for variety.
fundamentals
Context Window
The maximum amount of text, measured in tokens, that a model can consider at once — its working memory for a single request.
Hallucination
When a model states something false with complete confidence — an inherent property of prediction, not a bug to be patched.
Inference
Running a trained model to produce output — the part you pay for every time someone uses your AI feature.
LLM (Large Language Model)
A neural network trained on vast amounts of text to predict the next piece of text, which turns out to enable reasoning, writing and coding.
Multimodal AI
Models that handle more than one type of input or output — text, images, audio and video in a single system.
Token
The unit of text a model reads and generates — roughly three quarters of a word in English — and the unit you are billed for.
Transformer
The neural network architecture behind nearly every modern AI model, built around a mechanism called attention.
training
Fine-Tuning
Further training an existing model on your own examples so it adopts a specific style, format or behaviour.