AI Glossary

A person or company that earns a commission by referring paying customers to another business, typically through a tracked link.

Coordinating multiple AI models, tools, or agents to work together on a complex goal.

An AI system that can break down goals into steps, use tools, and iterate until a goal is achieved.

A software program that uses an AI model to take autonomous actions — browsing the web, writing files, calling APIs — to complete a goal without step-by-step human instructions.

Making an AI's goals and behavior consistent with human values and intentions.

Application Programming Interface — a defined way for one software application to request data or actions from another, enabling tools to communicate without a human in the loop.

Standalone outputs Claude generates in a side panel — code, HTML apps, React components, SVGs, or documents — that can be shared, iterated on, or embedded.

A neural network technique that lets models weigh which input tokens are most relevant for each output.

Using software to perform tasks without human intervention — triggered by an event, running on a schedule, or responding to conditions.

A standardized test measuring how well an AI model performs on specific tasks.

A prompting technique that asks an AI to explain its reasoning step-by-step before giving a final answer.

A software program designed to simulate conversation with a user, responding to text inputs with pre-defined or AI-generated replies.

The rate at which customers cancel their subscriptions or stop paying — the primary measure of retention health for any recurring-revenue business.

An AI capability that lets the model control a computer — moving the cursor, clicking buttons, typing, and reading the screen — to complete tasks autonomously.

The maximum amount of text (in tokens) an AI model can process in a single conversation.

A product delivered digitally — such as an ebook, template, course, or software download — with zero inventory cost and near-zero cost of distribution.

A numerical representation of text (or other data) that captures meaning, enabling semantic search and comparison.

Teaching a model to perform a task by providing a small number of examples (usually 2–10) in the prompt.

The process of updating a pre-trained model with task-specific or domain-specific data to improve performance.

Graphics Processing Unit — a chip originally built for rendering graphics, now the primary hardware for training and running AI models due to its ability to perform thousands of calculations simultaneously.

Anchoring an AI's responses to factual data to reduce hallucination.

When an AI generates false, made-up, or nonsensical information with confidence.

The process of a trained model generating a response to an input. When you chat with ChatGPT, that's inference.

A connection between two software tools that allows them to share data and trigger actions in each other automatically.

A neural network trained on massive amounts of text to predict and generate human language.

The time it takes for an AI model to generate a response from when you send a prompt.

Model Context Protocol — an open standard that lets AI models connect to external tools, data sources, and services through a unified interface.

Monthly Recurring Revenue — the predictable income a business earns every month from active subscriptions or retainer contracts.

An architecture where multiple AI agents work in parallel or in sequence, each handling a specialised subtask, coordinated by an orchestrator.

An AI model that can process and understand multiple types of input—text, images, video, audio.

A computational system loosely modelled on biological neurons, consisting of layers of mathematical functions that learn patterns from data.

Software tools that let non-programmers build apps, automations, and workflows using visual interfaces instead of writing code.

Software or AI models whose underlying code or weights are publicly available for anyone to inspect, modify, and use.

Anthropic's term for a company or developer that builds a product or service on top of Claude via the API, as distinct from the end user.

The learnable numerical weights inside a neural network — often cited as billions ("7B", "70B") as a rough proxy for model size and capability.

Income that continues to arrive without ongoing active work — generated by assets, products, or systems set up previously.

The initial large-scale training phase where a model learns language patterns from massive text datasets.

The text instruction you send to an AI model asking it to do something.

A technique that stores repeated sections of a prompt server-side so they do not need to be reprocessed on every API call, reducing latency and cost.

The practice of crafting and refining the inputs you give an AI model to get more accurate, useful, or consistent outputs.

A technique that combines document retrieval with AI generation to ground responses in factual data.

A recurring payment arrangement where a client pays a fixed monthly fee for ongoing access to a service or a defined scope of work.

A training technique where human reviewers rate AI outputs, and the model learns to generate outputs that score high.

Software as a Service — software delivered over the internet on a subscription basis, where the provider hosts and maintains the product.

A layer that blocks or modifies AI outputs that violate safety policies (hate speech, explicit content, etc.).

Reusable workflow modules that give Claude a defined set of instructions, scripts, and assets for a specific task.

A mode where the model is constrained to return responses in a specific format — typically JSON — that matches a defined schema.

A special instruction given to an AI model that defines its behavior and personality for all subsequent user interactions.

A parameter that controls how random or deterministic an AI's responses are (0 = deterministic, 1+ = very random).

AI generation of images from a text description, where the model interprets the prompt and synthesises a visual output.

The speed at which a model generates output tokens, measured in tokens per second.

The smallest unit of text an AI processes—usually a word fragment, character, or subword.

Enabling an AI to call external functions or APIs to perform tasks beyond text generation.

A sampling method that limits token choices to the smallest set of high-probability options.

The neural network architecture that powers modern large language models.

The event that starts an automation workflow — such as a form submission, a new email, or a file upload.

A specialized database optimized for storing and searching embeddings by similarity.

AI generation of human-sounding speech from text, producing audio that closely resembles a real human voice.

A way for one application to automatically notify another when something happens, by sending an HTTP request to a URL you specify.

Asking a model to perform a task without providing examples—just an instruction.