AI and Claude glossary.
Lightbridge Labs maintains this AI and Claude glossary as an independent, plain-language reference for teams adopting artificial intelligence. Each entry defines one term in one to three sentences, grouped into core AI concepts, working with models, the Claude product family, and AI governance, so a reader can settle on the right meaning fast.
Core artificial intelligence terms, defined by Lightbridge Labs
These are the foundational concepts behind every modern AI system. Lightbridge Labs uses these definitions across its consulting and training work.
Artificial intelligence (AI)
Artificial intelligence is the field of building computer systems that perform tasks normally associated with human intelligence, such as understanding language, recognizing patterns, and making decisions. Modern AI is dominated by machine learning, where systems learn behavior from data rather than from hand-written rules.
Large language model (LLM)
A large language model is an AI system trained on very large amounts of text to predict and generate language. Given some input text, it produces the most likely continuation, which is how it answers questions, writes, summarizes, and reasons. Claude is a large language model.
Generative AI
Generative AI refers to systems that produce new content, including text, images, audio, code, and video, rather than only classifying or scoring existing data. Large language models are the text and code branch of generative AI.
Foundation model
A foundation model is a large model trained on broad data that can be adapted to many downstream tasks. The same base model can power chat, summarization, coding, and search, which is why one model serves many products.
Machine learning
Machine learning is the branch of AI where a system improves at a task by learning patterns from data instead of following explicit programmed rules. It is the technique underneath nearly all current AI products.
Inference
Inference is the act of running a trained model to produce an output for a given input, as opposed to training, which builds the model in the first place. Each time you send a prompt and get a response, you are paying for one inference call.
Token
A token is the unit of text a language model reads and writes, roughly a short word or part of a word. Models process and are usually priced by the number of input and output tokens, so token count is a practical measure of length and cost.
Context window
The context window is the maximum amount of text, measured in tokens, that a model can consider at once, including the prompt and its own response. Information outside the current context window is not available to the model unless it is supplied again.
Embedding
An embedding is a list of numbers that represents the meaning of a piece of text, image, or other data, so that similar items sit close together in a mathematical space. Embeddings power semantic search and are the retrieval mechanism behind RAG.
Multimodal AI
Multimodal AI describes models that work across more than one kind of input or output, such as text together with images, documents, or audio. A multimodal model can, for example, read a chart in an image and answer questions about it in words.
Hallucination
A hallucination is a confident but false or unsupported statement produced by an AI model. Because a language model predicts plausible text rather than retrieving verified facts, outputs that matter should be grounded in sources and checked.
Working with AI models: techniques Lightbridge Labs uses in production
These terms describe how practitioners actually build with large language models, from prompting through tool use and retrieval.
Prompt engineering
Prompt engineering is the practice of writing the instructions and context given to an AI model to get reliable, useful results. It covers being specific about the task, supplying examples, and controlling the output format. Lightbridge Labs publishes a free primer on how to prompt Claude.
System prompt
A system prompt is a set of standing instructions given to a model that defines its role, rules, and tone before any user message arrives. It frames the whole conversation and is where persistent behavior is set.
Retrieval-augmented generation (RAG)
Retrieval-augmented generation is a pattern that first searches a knowledge source for relevant text, then supplies that text to the model so its answer is grounded in your own data. RAG reduces hallucination and lets a model answer about information it was never trained on.
Fine-tuning
Fine-tuning further trains an existing model on a focused dataset so it adapts to a specific style, domain, or task. It changes the model's weights, unlike prompting or retrieval, which only change the input at run time.
Tool use (function calling)
Tool use, also called function calling, lets a model invoke external functions or APIs, such as a search, a calculator, or a database query, and then use the results in its answer. It is how a model takes actions in the real world rather than only generating text.
AI agent
An AI agent is a system that pursues a goal across multiple steps, deciding which tools to call and reacting to the results, rather than answering a single prompt. Lightbridge Labs covers building agents in its guide to building with the Claude API.
Model Context Protocol (MCP)
The Model Context Protocol is an open standard for connecting AI applications to external tools and data sources through a common interface. It lets one connector be reused across many AI clients instead of building a custom integration for each. Lightbridge Labs explains MCP in its Claude API guide.
Reasoning and extended thinking
Reasoning, sometimes exposed as extended thinking, is a mode in which a model works through a problem step by step before giving its final answer. Spending more thinking effort tends to improve accuracy on hard, multi-step tasks at the cost of higher latency.
Structured outputs
Structured outputs constrain a model to return data in a defined shape, such as valid JSON matching a schema, so the result can be consumed reliably by other software. This turns free text into something a program can parse without guesswork.
Prompt caching
Prompt caching stores a large, reused portion of a prompt so it does not have to be reprocessed on every request. For workloads that resend the same context repeatedly, it lowers both latency and token cost.
Claude and Anthropic terms, explained by Lightbridge Labs
These entries cover the Claude product family and the company that builds it. Lightbridge Labs publishes deeper guides on each.
Claude
Claude is the family of AI assistants built by Anthropic. The current model lineup spans tiers from Fable 5 and Opus 4.8 for the hardest work to Sonnet 4.6 and Haiku 4.5 for faster, lighter tasks. Lightbridge Labs publishes a guide explaining what Claude is.
Anthropic
Anthropic, PBC is the AI safety and research company that builds Claude. It is structured as a public benefit corporation, with a stated focus on developing reliable, steerable, and safe AI systems.
Claude Code
Claude Code is Anthropic's AI coding agent. It reads a codebase, writes and edits files, runs commands, and works with git on a developer's behalf, across the terminal, a desktop app, the browser, and code editors. Lightbridge Labs publishes a full explainer on Claude Code.
Claude Cowork
Claude Cowork is Anthropic's agentic desktop app for knowledge workers. It works inside folders you grant it, completes multi-step file tasks, and returns drafts for review. Lightbridge Labs explains Claude Cowork in a dedicated guide.
Claude API
The Claude API is the programmatic interface for sending requests to Claude models and receiving responses, used to build Claude into custom applications. It supports streaming, tool use, and the Model Context Protocol. Lightbridge Labs covers it in its building-with-the-Claude-API guide.
AI governance terms, defined by Lightbridge Labs
These terms describe how organizations control AI responsibly. Lightbridge Labs is an AI governance practice and is pursuing ISO 42001 certification.
AI governance
AI governance is the set of policies, roles, controls, and review processes an organization uses to manage AI systems responsibly across their lifecycle. It covers how AI is selected, built, monitored, and held accountable. Lightbridge Labs delivers AI governance as a consulting practice.
ISO 42001
ISO/IEC 42001 is the international management-system standard for artificial intelligence, published in 2023. It specifies requirements for establishing, operating, and improving an AI management system so AI is governed in a structured, auditable way. Lightbridge Labs is pursuing ISO 42001 certification.
AI management system
An AI management system is the organized framework of policies, processes, roles, and records through which an organization governs its use of AI. It is the central object that ISO 42001 defines requirements for.
Responsible AI
Responsible AI is the practice of designing and operating AI systems so they are fair, transparent, accountable, secure, and aligned with human values and applicable law. It is the goal that AI governance and standards like ISO 42001 work toward.
Vendor note: AI terminology and product capabilities change quickly. These definitions describe general concepts and the Claude product family as Lightbridge Labs understands them today. Verify current model tiers, features, and product specifics at docs.anthropic.com before planning around a particular detail.
This glossary is independent, general educational information published by Lightbridge Labs. Claude, Claude Code, Claude Cowork, and Anthropic are trademarks of Anthropic, PBC. Lightbridge Labs is not affiliated with, endorsed by, or a partner of Anthropic, PBC.
From definitions to a deployment.
When the vocabulary is settled and a real AI project is on the table, Lightbridge Labs helps you build, govern, and train your teams on production AI.