AI Glossary

53 terms explained in plain English. No jargon to decode the jargon.

AI Concepts Developer Terms Pricing & Limits Image & Video AI Infrastructure

AI Concepts20 terms

AI (Artificial Intelligence): Software that can do tasks that normally require human intelligence — understanding language, recognising images, making decisions, generating text. Modern AI is not a robot with a brain; it's a statistical system trained on enormous amounts of data to spot patterns and produce useful outputs.
LLM (Large Language Model): The type of AI behind tools like ChatGPT, Claude, and Gemini. An LLM is trained on huge amounts of text and learns to predict what words should come next — which turns out to be surprisingly powerful for writing, reasoning, and answering questions.
Model: The trained AI system itself. When people say 'which model are you using?', they mean which specific version of the AI — GPT-4o, Claude 3.5, Gemini 1.5 Pro, etc. Different models have different strengths, speeds, and prices.
Prompt: The message or instruction you send to an AI. The quality of your prompt has a huge effect on the quality of the response. A vague prompt gets a vague answer; a specific, well-structured prompt gets a useful one.
Token: The unit AI models use to measure text. A token is roughly ¾ of a word — 'fantastic' is one token, 'I love pizza' is about 4 tokens. AI pricing is almost always measured in tokens (per 1,000 or per 1 million). Your input and the AI's output both count.
Context Window: How much text an AI can 'see' and remember at once — both what you've sent and what it's replied. Once you exceed the context window, the AI starts forgetting earlier parts of the conversation. Measured in tokens.
Hallucination: When an AI confidently states something that is false. LLMs generate text based on patterns rather than looking things up, so they can produce plausible-sounding information that is simply wrong. Always verify important facts from an AI.
Inference: The process of running an AI model to generate a response. When you send a message to ChatGPT and it replies, that's inference happening — the model is 'inferring' the most appropriate output based on your input.
Fine-tuning: Taking an existing AI model and further training it on a specific dataset to make it better at a particular task. A company might fine-tune a general model on their customer support tickets to make it better at handling their specific queries.
RAG (Retrieval-Augmented Generation): A technique where an AI looks up relevant information from a database before answering, rather than relying solely on what it was trained on. This reduces hallucinations and lets the AI answer questions about things that happened after its training cutoff.
Multimodal: An AI that can process more than one type of input — typically text plus images, audio, or video. A multimodal model can look at a photo and describe it, or listen to audio and transcribe it.
Training Data: The information an AI model learned from. For LLMs, this is typically a massive collection of text from the internet, books, and other sources. The quality and content of training data directly shapes what the model knows and how it behaves.
Training Cutoff: The date after which an AI has no knowledge. If a model's training cutoff is January 2025, it doesn't know about events that happened after that date — unless it has access to web search.
System Prompt: A set of instructions given to an AI before the conversation starts, usually by the developer or platform rather than the user. It shapes how the AI behaves — its tone, what it will and won't do, what persona it takes on.
Agent / AI Agent: An AI that doesn't just answer questions but takes real actions — browsing the web, running code, sending emails, reading and writing files. An agent works through multi-step tasks autonomously rather than waiting for you to guide each step.
Agentic: Describes AI behaviour that is autonomous and action-oriented — where the AI makes decisions, takes steps, and uses tools on its own initiative rather than just responding to a single prompt. An 'agentic workflow' is one where the AI drives the process.
MCP (Model Context Protocol): An open standard created by Anthropic that gives AI assistants a consistent way to connect to external tools, databases, and data sources. Think of it as USB-C for AI integrations — one standard connector that works across many different AI clients. Each 'MCP server' is a small program that bridges the AI to a specific service like GitHub, Notion, or a database.
Open Source: Software whose code is publicly available for anyone to read, use, modify, and distribute. Open-source AI models can be downloaded and run on your own hardware with no usage fees. The opposite is 'closed source' or 'proprietary'.
Parameters: The numbers inside an AI model that were adjusted during training. More parameters generally means a more capable model — but also larger, slower, and more expensive to run. Often stated in billions (7B, 70B, 405B).
Quantisation: A technique to make AI models smaller and faster by reducing the precision of their numbers — like compressing an image. A quantised model uses less memory and runs faster, with a small trade-off in quality.

Developer Terms14 terms

API (Application Programming Interface): A way for two pieces of software to talk to each other. An API lets your app send a request to another service and get a response back — without needing to know how that service works internally.
API Key: A secret password that identifies who is making an API request. When you call the OpenAI API, you include your API key so they know it's you and can charge your account. Keep API keys private — anyone with your key can use your account.
SDK (Software Development Kit): A package of tools and code that makes it easier to use a particular platform or API. Instead of writing raw API requests yourself, an SDK gives you ready-made functions in your programming language.
CLI (Command-Line Interface): A text-based way of controlling a computer or program by typing commands. The opposite of a GUI (graphical interface with buttons and windows). Many developer tools are CLIs — you open a terminal and type commands.
Terminal / Command Line: A text interface on your computer where you type commands directly. On Mac it's called Terminal (or iTerm). On Windows it's Command Prompt or PowerShell. Many developer tools only work via the terminal.
Repository (Repo): A folder that stores a project's code along with the complete history of every change ever made. Usually hosted on GitHub. Think of it as a project folder with a time machine built in.
Environment Variable: A secret value (like an API key or database password) stored outside your code, in the system environment. Referenced by name in code (e.g. `process.env.API_KEY`) but never written directly into the code file — so it stays private.
Deployment: The process of making your website or app live and accessible to users. When you 'deploy', you take code that's working on your computer and publish it to a server where the public can reach it.
Serverless: A way of running code in the cloud without managing a permanent server. You write a function, upload it, and it runs on demand — you only pay for the milliseconds it actually executes. Vercel and AWS Lambda use this model.
Open Source: Software whose source code is publicly available. Anyone can read it, use it, modify it, and often contribute to it. Most open-source software is free to use.
Git: A version control system that tracks changes to files over time. Every change is saved as a 'commit' with a message, so you can see exactly what changed and revert to any previous state. GitHub is built on top of Git.
Framework: A pre-built structure for building software — a set of tools, conventions, and reusable code that handles common tasks so you can focus on building your specific product.
Frontend / Backend: Frontend is everything the user sees and interacts with — the website in their browser. Backend is everything running on the server — databases, business logic, APIs. Full-stack means working on both.
Markdown: A simple formatting language that uses plain text symbols to indicate bold, headings, links, lists, and code. Widely used for documentation, README files, and content. This guide is written in Markdown.

Pricing & Limits5 terms

Free Tier: A limited version of a paid product that's available at no cost. Most AI tools offer a free tier so you can try before subscribing. Free tiers typically have lower usage limits, older models, or fewer features.
Rate Limit: A cap on how many requests you can make in a given time period. Hit your rate limit and you'll get an error until the window resets. Higher-tier plans usually have higher rate limits.
Usage Cap: A hard limit on total usage — after which the service stops or requires an upgrade. Different from a rate limit: a usage cap is a total ceiling, a rate limit is a speed limit.
Per-token Pricing: Billing based on the number of tokens processed. Most AI APIs charge separately for input tokens (what you send) and output tokens (what the AI generates). Output is typically more expensive than input.
Context Length: In pricing terms, how much text you can include in a single request. Longer context costs more tokens. Models with larger context windows can handle longer documents and conversations.

Image & Video AI7 terms

Diffusion Model: The technology behind most AI image generators (Stable Diffusion, Midjourney, DALL·E). It works by starting with random noise and gradually 'denoising' it into a coherent image guided by your text prompt.
Prompt (Image): The text description you write to generate an image. Image prompts work differently from chat prompts — they're often stylistic and descriptive rather than conversational.
Negative Prompt: Instructions for what you don't want in an AI-generated image. Used to remove unwanted elements like blurry backgrounds, extra fingers, or watermarks.
Inpainting: Editing a specific region of an existing image with AI while leaving the rest untouched. You 'mask' the area you want to change, then describe what should replace it.
Segmentation: The process of identifying and isolating specific objects within an image or video — drawing a precise outline around a person, product, or background element. SAM 3 by Meta is a leading segmentation model.
Text-to-Image: Generating an image from a text description. The core feature of tools like Midjourney, DALL·E, Flux, and Ideogram.
Text-to-Video: Generating a video clip from a text description. Tools like Kling, Runway, Luma, and Veo work this way.

Infrastructure7 terms

Cloud: Running software and storing data on remote servers owned by a third party (like Amazon, Google, or Microsoft) rather than on your own hardware. Most modern apps and AI tools run in the cloud.
Server: A computer that runs continuously and responds to requests from other computers. When you load a website, your browser sends a request to the site's server, which sends back the page.
Database: Organised storage for structured data. When you create an account on a website, your username and password are stored in a database. Common types: PostgreSQL, MySQL (relational), Firebase, MongoDB (document).
Webhook: A way for one system to notify another system that something happened, by sending an automatic HTTP request. Instead of constantly checking 'did anything change?', a webhook says 'here's what just happened'.
Cron Job: A task that runs automatically on a schedule. Named after the Unix 'cron' scheduler. Used to run repetitive background tasks without manual triggers.
DNS (Domain Name System): The internet's address book. DNS translates human-readable domain names (simpleai.club) into IP addresses that computers use to find each other. When you 'point your domain' to a host, you're updating DNS records.
CDN (Content Delivery Network): A network of servers spread around the world that store copies of your website's files. When a user loads your site, they get the files from the nearest server — making it faster regardless of where they are.

Missing a term?

If there's a word confusing you that isn't here, let us know and we'll add it.