AI Glossary

A

• Fundamentals AGI (Artificial General Intelligence) AGI is the idea of an AI that can handle a wide range of intellectual work at human level, rather than being limited to one narrow task. It is the stated goal of many leading AI companies. Read the explanation →
• Fundamentals AI (Artificial Intelligence) AI is an umbrella term for computer technologies that perform tasks associated with human intelligence, such as understanding, prediction, and decision-making. Read the explanation →
• Fundamentals AI and Copyright Copyright questions about AI fall into two separate areas: whether protected works can be used as training data, and who owns AI output and whether it infringes existing works. Read the explanation →
• Fundamentals AI Bias AI bias is skew in a model's judgments or output. It usually originates in the training data, and it matters most where decisions are expected to be fair. Read the explanation →
• Fundamentals AI Coding AI coding covers any development work done with AI assistance, from inline completion to agents that read a repository, write changes, and run tests. Read the explanation →
• Fundamentals AI Credits AI credits are a service-specific unit that measures how much you have used features such as image generation or AI agents. Paid plans usually include a monthly allowance. Read the explanation →
• Fundamentals AI Hallucination An AI hallucination is a fluent, confident answer that contains invented or incorrect facts. Important names, dates, numbers, and sources require verification. Read the explanation →
• Fundamentals AI Literacy AI literacy is the knowledge and judgment needed to understand what AI is good and bad at, and to use it well. It is increasingly treated as a basic skill rather than a specialty. Read the explanation →
• Fundamentals AI Memory AI memory lets an assistant carry your preferences and past context into new conversations. It is distinct from the context of a single chat and from training the underlying model. Read the explanation →
• Fundamentals AI Model An AI model learns patterns from large amounts of data and then predicts or generates output from new input. It is the engine inside a product such as ChatGPT. Read the explanation →
• Fundamentals AI Search AI search looks across the web for you, reads multiple sources, and returns an answer written as prose. Most services show the source links alongside the answer. Read the explanation →
• Fundamentals Alignment Alignment is the research and engineering work of making an AI's goals and behavior match human intent and values. It is the central topic in AI safety. Read the explanation →
• Fundamentals API (Application Programming Interface) An API is a defined way for software to request data or functions from another service. AI APIs let developers add models to their own apps and workflows. Read the explanation →
• Fundamentals API Key An API key is a secret string that identifies which account or project is calling an API, and controls permissions and usage. Treat it with the same care as a password. Read the explanation →
• Fundamentals ASI (Artificial Superintelligence) ASI describes an AI that would far exceed human intelligence across every field. It is discussed as the stage beyond AGI. Read the explanation →

B

• Fundamentals Benchmark A benchmark is a standardized test used to compare AI models. Scores are a useful reference, not a definitive measure of usefulness for your work. Read the explanation →
• Fundamentals Branch A branch is a separate line of development in Git, letting you work on a feature or a fix without disturbing the stable version. Read the explanation →

C

• Fundamentals Chatbot A chatbot is any program that converses with people through chat. Generative AI changed them from scripted responders into systems that hold a natural conversation. Read the explanation →
• Fundamentals CLI (Command Line Interface) A CLI lets you operate a computer by typing commands instead of clicking. You work in a terminal, entering short instructions rather than using buttons and icons. Read the explanation →
• Fundamentals Clone Cloning copies a repository — its files and its history — from a service such as GitHub onto your own machine so you can start working with it. Read the explanation →
• Fundamentals Commit A commit records a selected set of changes into Git's history along with a message describing them. It is saved locally first, and is not sent anywhere until you push. Read the explanation →
• Fundamentals Context Engineering Context engineering is the practice of choosing what information an AI receives — instructions, source material, history, tools, memory — and shaping how it is delivered. Read the explanation →
• Fundamentals Context Window A context window is the amount of conversation, instructions, and source material an AI model can consider at once, measured in tokens. Read the explanation →
• Fundamentals Credentials Credentials are the information used to establish who someone is and whether they may use a service — passwords, API keys, access tokens, and certificates. Read the explanation →
• Fundamentals Custom Instructions Custom instructions store the tone, assumptions, and rules you want an assistant to follow, so they apply across conversations instead of being repeated each time. Read the explanation →

D

• Fundamentals Deepfake A deepfake is a convincing fake image, video, or audio clip produced with AI. As quality improves, judging authenticity by eye is no longer reliable. Read the explanation →
• Fundamentals Deep Learning Deep learning trains many-layered neural networks modeled loosely on the brain. It is the technique behind almost every AI advance since the 2010s. Read the explanation →
• Fundamentals Deploy Deploying puts a site or application onto a server or runtime so that people can actually use it. It is a separate step from pushing code to a repository. Read the explanation →
• Fundamentals Design Harness A design harness is a working setup that gives an AI the design principles, references, procedure, and review method it needs to produce consistent screens. Read the explanation →

F

• Fundamentals Fine-tuning Fine-tuning trains an existing model further on additional data to adapt it to a specific use, rather than building a model from scratch. Read the explanation →
• Fundamentals Foundation Model A foundation model is a large model pre-trained on vast data and reusable as a base for many purposes. GPT, Gemini, and Claude are the best-known examples. Read the explanation →

G

• Fundamentals Generative AI Generative AI creates new text, images, audio, video, or code. ChatGPT and modern image generators are familiar examples. Read the explanation →
• Fundamentals GPU (Graphics Processing Unit) A GPU is a chip that performs huge numbers of calculations in parallel. Built originally for graphics, it became essential to training and running AI models. Read the explanation →

I

• Fundamentals Inference Inference is a trained model actually producing an answer. Every time you send a question to an assistant and get a reply, the model is doing inference. Read the explanation →

K

• Fundamentals Knowledge Cutoff The knowledge cutoff is the date beyond which a model's training data does not extend. It has essentially no knowledge of events after that point. Read the explanation →

L

• Fundamentals LLM (Large Language Model) A large language model, or LLM, is an AI model trained on large amounts of text. It is the language-processing engine behind services such as ChatGPT. Read the explanation →
• Fundamentals Local LLM A local LLM runs on your own computer or server instead of a cloud service, so your input never leaves the machine. Read the explanation →

M

• Fundamentals Machine Learning Machine learning is the practice of having computers find patterns in data rather than following rules written by hand. Almost all current AI is built this way. Read the explanation →
• Fundamentals Markdown Markdown is a lightweight plain-text syntax for headings, lists, links, and emphasis. It is widely used in notes, documentation, GitHub, and AI instructions. Read the explanation →
• Fundamentals Merge Merging combines the changes from one Git branch into another. When both sides changed the same lines, you get a conflict that a person has to resolve. Read the explanation →
• Fundamentals Multimodal AI Multimodal AI can work with more than one type of information, such as text, images, audio, and video, within the same task. Read the explanation →

N

• Fundamentals Neural Network A neural network is a computing structure loosely modeled on connections between brain cells. It is the foundation under deep learning and today's language models. Read the explanation →

O

• Fundamentals OAuth OAuth lets you grant an application limited access to a service without handing over your password. You approve a specific scope, and the app receives a token instead of your credentials. Read the explanation →
• Fundamentals On-device AI On-device AI runs the processing on your phone or computer rather than in the cloud. It offers privacy, lower latency, and offline use, within the limits of the hardware. Read the explanation →
• Fundamentals Open Source Model An open source model is one whose weights are published, so anyone can download, run, and modify it. Because licenses vary, the term open weights is often more accurate. Read the explanation →

P

• Fundamentals Parameter Parameters are the adjustable numbers inside a model. Parameter count is the most common shorthand for how large a model is. Read the explanation →
• Fundamentals Pre-training Pre-training is the first stage of building a model, where broad capability is acquired from a very large dataset. The P in GPT stands for it. Read the explanation →
• Fundamentals Prompt A prompt is the question or instruction given to an AI. Clear goals, context, constraints, and output formats usually produce better results. Read the explanation →
• Fundamentals Pull Pulling fetches new commits from a remote such as GitHub and integrates them into your current local branch. Read the explanation →
• Fundamentals Pull Request A pull request proposes merging one branch into another and provides a place to review, discuss, and approve the change before it lands. Read the explanation →
• Fundamentals Push Pushing sends commits from your local repository to a remote such as GitHub, making them available to everyone else. Read the explanation →

R

• Fundamentals Repository A repository holds a project's files together with its Git history. You work with a local one on your machine and a remote one on a service such as GitHub. Read the explanation →

S

• Fundamentals Second Brain A second brain is an external, digital store for what you want to remember, organized so you can retrieve and reuse it rather than holding it all in your head. Read the explanation →
• Fundamentals Singularity The singularity is the hypothesized point at which AI surpasses human intelligence and technological progress becomes too fast to predict. Read the explanation →
• Fundamentals Spec-Driven Development Spec-driven development puts a written specification at the center, and has AI and people implement, test, and update documentation against it. Read the explanation →
• Fundamentals Staging Staging is choosing which of your working changes go into the next commit. In Git the holding place for that selection is called the staging area, or the index. Read the explanation →
• Fundamentals Summarization Summarization selects the important information from a long text or conversation and restates it briefly. It is one of the most common uses of generative AI. Read the explanation →
• Fundamentals Synthetic Data Synthetic data is generated artificially rather than collected from the real world. It addresses shortages of training data, privacy constraints, and rare-case coverage. Read the explanation →

T

• Fundamentals Token A token is a unit an AI model uses to process text. Context limits and most API pricing are measured in tokens rather than words or characters. Read the explanation →
• Fundamentals Training Data Training data is the material a model learns from. Its quality and coverage largely determine what a model knows and where it is biased. Read the explanation →

U

• Fundamentals Upscaling Upscaling increases the pixel dimensions of an image to produce a larger or higher-resolution version. AI upscalers invent plausible detail rather than recovering the original. Read the explanation →

W

• Fundamentals Workflow A workflow lays out the trigger, the order of steps, the decisions, and the handoffs that carry a job from start to finish. AI can be built into it. Read the explanation →