# NerdTerms > A plain-English reference for AI: glossary terms (each tagged beginner, intermediate, or advanced), step-by-step tutorials, and copy-ready prompts. Published by Sister Company (https://sistercompany.ai). ## Terms - [Agent](https://nerdterms.com/terms/agent): An AI that does the task instead of describing how. It decides the steps, uses tools such as web search or file access, checks the results, and keeps going until the job is done. A chatbot answers; an agent acts. That is what makes agents useful, and what makes their mistakes expensive. - [Agent graph](https://nerdterms.com/terms/agent-graph): A way of building an AI agent by drawing its work as a graph: boxes for steps and arrows for what happens next. Each step can ask the AI something or run normal software. Arrows can branch on a condition or loop back, and all steps share one running record of where things stand. LangGraph made the approach common. - [Agentic](https://nerdterms.com/terms/agentic): An adjective meaning able to act and get results on its own, now used mostly about AI that does things rather than just answering. An agentic tool searches, books, edits, or sends. The word is on nearly every AI product page, including on features that do none of those things, so ask what it does while you are not there. - [AI slop](https://nerdterms.com/terms/ai-slop): Low-quality content produced with AI, usually in bulk, with little effort or human judgment behind it. The word spread in 2024 by analogy to spam, the name for unwanted email, and covers text, images, video, design, and code. Merriam-Webster named "slop" its 2025 Word of the Year. - [Anthropic](https://nerdterms.com/terms/anthropic): The American AI company behind Claude, Claude Code, and the MCP standard. Former OpenAI staff founded it in 2021 with AI safety as the stated goal. It is a public benefit corporation based in San Francisco, a legal form that commits a company to a public mission alongside profit. - [Claude Code](https://nerdterms.com/terms/claude-code): Anthropic's agent that works on your computer instead of in a chat window. Give it a task and it reads, edits, and creates files, runs commands, and keeps going until done, asking before risky steps. It runs in the terminal, in code editors, as a desktop app, and on the web, with a Claude subscription. - [Codex](https://nerdterms.com/terms/codex): OpenAI's coding agent, the ChatGPT-side counterpart to Claude Code. Released in April 2025, it writes code, fixes bugs, and works through tasks on a project for you, on your machine or in OpenAI's cloud. It is available inside ChatGPT on the web, in the terminal, as a desktop app, and in code editors. - [Context engineering](https://nerdterms.com/terms/context-engineering): The practice of choosing what an AI model sees before you ask it anything: instructions, examples, files, and history, in the right order and nothing extra. Prompting is the question. Context engineering is everything around it, so the first answer is usable instead of generic. It is the skill behind every AI setup that "just works." - [Context graph](https://nerdterms.com/terms/context-graph): A knowledge graph extended for AI agents. Besides things and relationships, it stores decisions, policies, exceptions, precedents, and outcomes as connected nodes, each marked with when it was valid. The point is the "why" behind a decision, not just the "what," so an agent can look up how a similar case was handled and whether that rule still holds. - [Context window](https://nerdterms.com/terms/context-window): The maximum amount of text an AI model can work with at once: your instructions, attached files, the conversation so far, and the reply it is writing. It is measured in tokens, which are chunks of words. When it is full, chat tools quietly drop or summarize the oldest parts, and more room is not automatically better. - [Cursor](https://nerdterms.com/terms/cursor): A code editor with an AI agent built in, made by Anysphere, now part of SpaceX. A code editor is the program developers write software in; Cursor is a modified version of the most common free one, Visual Studio Code. Type code yourself or hand the agent a whole feature. It is for people who want to see code. - [Dots](https://nerdterms.com/terms/dots): OpenAI's always-on AI agents, launched in September 2026. A dot works in the background toward goals you set, without being tied to the ChatGPT window or any one device. You name your first dot, give it work, and message it from ChatGPT, Slack, or Microsoft Teams. At launch it needed one of ChatGPT's top paid plans. - [Generative AI](https://nerdterms.com/terms/generative-ai): Artificial intelligence that creates new content, such as text, images, video, audio, or code, when you ask for it. Older AI sorted, ranked, or predicted; generative AI produces. It learned patterns from an enormous amount of existing material and uses them to make something new. ChatGPT, Claude, Gemini, and Midjourney are generative AI. - [Grok Bot](https://nerdterms.com/terms/grok-bot): An always-on AI agent from SpaceXSI, the AI company inside SpaceX that was formerly called xAI. Each bot gets its own computer in the cloud, signs into the tools you already use, and works across apps and inboxes around the clock. You message it like a colleague, and it checks back only when something needs your approval. - [Hallucination](https://nerdterms.com/terms/hallucination): When an AI states something false, invented, or unsupported as if it were fact: a fake citation, a wrong date, a feature that does not exist. It happens because the model produces the most plausible next words, not checked ones. Plausible and true overlap most of the time, not always. Every name, number, and link needs checking. - [Harness](https://nerdterms.com/terms/harness): The software wrapped around an AI model that turns it into a working agent. It runs the tools, hands results back to the model, keeps the conversation trimmed so it fits in memory, and keeps the agent going on long tasks. Claude Code, Codex, and Cursor are harnesses. The model is the engine; the harness is the car. - [Knowledge cutoff](https://nerdterms.com/terms/knowledge-cutoff): The date a model's training data stops. It knows nothing newer unless it searches the web or you tell it. Ask about last week without search on and you get a guess or an admission. Model makers publish the date; Anthropic also publishes an earlier "reliable" date, because knowledge of the final months before the cutoff is thinner. - [Knowledge graph](https://nerdterms.com/terms/knowledge-graph): A way of storing information as a network of things and the relationships between them: a customer, the products they bought, the rep who handles them, drawn as connected dots and lines. Each dot is a thing, each line is a named relationship. AI tools use one to answer from checked facts instead of guessing from text. - [Large language model](https://nerdterms.com/terms/large-language-model): The kind of AI model behind ChatGPT, Claude, and Gemini. It was trained on an enormous amount of text and got very good at predicting what words come next, which at scale turns into answers, summaries, translations, and code. It is fluent whether or not it is right, so it needs checking. - [Loop engineering](https://nerdterms.com/terms/loop-engineering): The practice of designing systems that run an AI agent in cycles toward a goal with little human input: the system prompts the agent, checks the result against a test or a rule, and sends it back around until the goal is met or a stop condition is hit. The human designs the loop instead of typing every prompt. - [MCP](https://nerdterms.com/terms/mcp): A shared standard for plugging AI apps into other systems: your files, calendar, email, databases, and business tools. Build a connector once and any AI app that supports MCP can use it, like a USB-C port for AI. Anthropic released it as an open standard in November 2024; Claude, ChatGPT, and others support it. - [Model](https://nerdterms.com/terms/model): The engine behind an AI app. A model is a program trained on a huge amount of data to recognize patterns and produce answers, without being given step-by-step rules. ChatGPT is an app; the models inside it have names like GPT-5. Claude is both the app and the model family. Most apps let you switch models. - [Multimodal](https://nerdterms.com/terms/multimodal): A model that can take in more than one kind of input: text, images, audio, video, and documents. Older models handled text only. With a multimodal model you can drop in a screenshot, a photo of a receipt, or a PDF and ask about it directly. Images and files use context window space just like text. - [OpenAI](https://nerdterms.com/terms/openai): The American AI company behind ChatGPT, the GPT family of models, Codex, and Dots. It was founded in 2015 as a nonprofit, later added a for-profit arm, and is now a public benefit corporation based in San Francisco. Its release of ChatGPT in November 2022 set off the current wave of everyday AI use. - [Pilled](https://nerdterms.com/terms/pilled): Slang for having been converted to something, usually with fresh enthusiasm: "Claude-pilled," "agent-pilled." Stick "-pilled" on anything and it means "I tried it, it clicked, and now I won't shut up about it." It comes from the red pill in The Matrix by way of internet forums, and in AI circles it is usually half a joke. - [Projects](https://nerdterms.com/terms/projects): A feature in Claude that groups chats together with shared files and standing instructions. Upload the documents that matter, write a page of instructions, and every chat you start inside the project already has them. It is a system prompt plus a filing cabinet. Projects are available on every Claude plan, including free. - [Prompt](https://nerdterms.com/terms/prompt): The text you give an AI to get a result: a question, an instruction, a document to work on, or all three. In chat tools every message you send is a prompt, including the pasted file and the "make it shorter." Prompt engineering is the practice of wording that text so you get what you wanted. - [Prompt injection](https://nerdterms.com/terms/prompt-injection): An attack where text the AI reads, such as a web page, email, or document, contains hidden instructions that hijack what it does. The model cannot reliably tell your instructions from instructions buried in content. It is ranked the number one security risk for AI applications, and matters most when an AI reads other people's content and can take actions. - [Reasoning model](https://nerdterms.com/terms/reasoning-model): A model that works through a problem in steps before answering instead of replying immediately. It does better on logic, math, planning, and code than a standard model, and worse on nothing except speed and price. In chat apps it shows up as a "thinking" mode that takes longer and uses more tokens. - [Skills](https://nerdterms.com/terms/skills): Reusable instruction packs an AI agent loads only when a task calls for them. A skill is a folder with a text file inside, named SKILL.md, that explains how to do one kind of task your way, plus any files it needs. A prompt is for one chat. A skill is for every time that task comes up. - [System prompt](https://nerdterms.com/terms/system-prompt): Standing instructions an AI reads before the conversation starts, shaping every reply: its role, tone, rules, and what it knows about you. The app maker writes one you never see. You can add your own on top. In ChatGPT and Claude it appears as custom instructions or project instructions, and one clear sentence changes the output. - [Taste](https://nerdterms.com/terms/taste): In AI conversations, taste means the judgment to tell good output from merely plausible output, and to know what is worth making at all. The dictionary sense is critical judgment or discernment. When a tool can produce five drafts in a minute, picking the right one, or rejecting all five, becomes the human part of the job. - [Token](https://nerdterms.com/terms/token): The unit AI models count text in. A token is a chunk of a word, sometimes a whole short word, sometimes a piece of a long one. Tokens are what you pay for and what the context window counts, which is why limits are stated in tokens and never in words or pages. - [Tool calling](https://nerdterms.com/terms/tool-calling): How an AI model gets things done in the real world. The model cannot act on its own, so it writes a request, such as "search the web for X," and the app running it carries that out and hands back the result. The "Searching the web…" line in a chat is tool calling at work. - [Training data](https://nerdterms.com/terms/training-data): Everything an AI model learned from before anyone used it: an enormous pile of text, images, or other data, studied once and baked in. It shapes what the model knows, how it writes, and which mistakes it repeats. Your company's files are not in it unless they were public, which is why the model knows Shakespeare and not your prices. - [Vibe coding](https://nerdterms.com/terms/vibe-coding): Building software by telling an AI what you want and accepting the code it writes without reading it closely. The term was coined by Andrej Karpathy in February 2025 and named Collins Dictionary's Word of the Year for 2025. It is great for throwaway tools and risky for anything that handles money, customer data, or has to keep working. ## Tutorials - [How to summarize a long PDF with AI](https://nerdterms.com/tutorials/summarize-a-long-pdf): Turn a long report or contract into a structured summary you can check against the original. ## Prompts - [Summarize a long document by section](https://nerdterms.com/prompts/summarize-a-long-document): A prompt that turns a long document into a sectioned summary with page references you can verify. ## Optional - [Glossary as JSON](https://nerdterms.com/terms.json) - [Terms and conditions](https://nerdterms.com/legal/terms)