Glossary

AI, explained.

The AI terms that actually come up in business, defined in plain English with no hype. Skim it, or dig into any term.

Core concept

AI Agent

A language model placed in a loop where it can plan, call tools to take real actions, observe the results, and repeat until a goal is met.

Context Window

The maximum amount of text a model can consider at once, including your prompt, any documents you provide, and its own reply.

Inference

The act of running a trained model to get an answer; in other words, using the AI rather than training it.

Large Language Model (LLM)

The core AI model behind tools like ChatGPT, trained on huge amounts of text to predict the next word, which lets it read, write, summarize, and reason over language.

Multimodal AI

AI that can work with more than one type of data, such as text, images, audio, and video, in a single model.

Token

The small chunk of text, roughly a few characters or part of a word, that models read and generate, and the unit most AI usage is billed in.

Technique

Embeddings

A way of turning text, images, or other data into lists of numbers that capture meaning, so a computer can measure how similar two things are.

Fine-Tuning

Further training an existing model on your own examples so it adopts a specific style, format, or task without you having to spell it out every time.

Retrieval-Augmented Generation (RAG)

A technique that retrieves relevant passages from your own documents and asks the model to answer using only those, so responses are grounded in your real content.

Structured Output

Getting a model to return data in a strict, predictable format such as JSON, so your software can use the answer directly.

System Prompt

The hidden set of instructions that tells a model who it is, how to behave, and what rules to follow before any user message arrives.

Tool Calling (Function Calling)

A capability that lets a model trigger real actions, like searching a database or sending an email, by returning a structured request your code can run.

Infrastructure

Model Context Protocol (MCP)

An open standard for connecting AI assistants to external tools and data sources through one common interface, instead of a custom integration each time.

Vector Database

A specialized database built to store embeddings and quickly find the items most similar in meaning to a given query.

Risk

Hallucination

When an AI model states something false or made-up with complete confidence, because it is predicting plausible text rather than checking facts.

Prompt Injection

A security risk where hidden instructions in user input or fetched content trick an AI system into ignoring its rules and doing something it shouldn't.

Practice

Evals (Evaluation)

A repeatable test that measures how well an AI system performs on real examples, so you can improve it with evidence instead of vibes.

Guardrails

The checks and limits placed around an AI system to keep its inputs and outputs safe, on-topic, and within policy.

Charleston waterway at sunset with palmetto silhouettes

Get in touch

Past the definitions, what should you build?

Knowing the words is a start. If you want help turning the concepts into something useful, just ask.