AI / Artificial Intelligence
Technology designed to do things we normally associate with human intelligence, such as understanding language, recognising patterns, creating content or making predictions.
Resources · Guide
ai cheat sheet
The jargon-free guide to the AI words, phrases and internet language everyone suddenly seems to be using.
AI can feel more complicated than it is because the language has moved incredibly quickly. You do not need to know all of this before you start using it, but understanding the words people keep throwing around makes everything feel much less intimidating.
The most important thing is still the same: be curious. We're all learning this.
130 terms
Generative AI
AI that creates something new, including text, images, video, audio, music or code.
Model
The actual AI system doing the work. Different models have different strengths, speeds and abilities.
LLM / Large Language Model
A type of AI model trained on huge amounts of language so it can understand and generate text.
Multimodal AI
AI that can work with more than one kind of input or output, such as text, images, audio and video.
Foundation Model
A large general-purpose AI model that can be adapted or used for lots of different tasks.
Training Data
The information used to teach an AI model patterns.
Inference
The moment when an AI model actually uses what it has learned to generate an answer or prediction.
Parameters
Internal values inside a model that help determine how it responds. You will often hear people talk about models having billions of parameters.
Benchmark
A test used to compare how well different AI models perform on certain tasks.
Open Source Model
An AI model whose code or weights are made available for others to use, inspect or adapt, depending on the licence.
Closed Model
A model controlled by a company where the underlying system is not fully open to the public.
Small Language Model / SLM
A smaller AI model designed to use fewer resources and often run faster or more cheaply.
Reasoning Model
A model designed to spend more effort working through complex problems before answering.
Frontier Model
A term often used for the most advanced AI models currently available.
Prompt
The instruction or question you give an AI.
Prompt Engineering
A fancy term for learning how to ask AI for what you want more clearly and effectively.
System Prompt
A hidden or higher-level instruction that helps define how an AI assistant should behave.
Context
The information an AI has available when answering you.
Context Window
The amount of information an AI model can consider at once.
Token
A small chunk of text used by AI systems. Tokens are often used to measure how much information a model can process.
Memory
A feature that allows an AI system to remember certain information about you or previous conversations over time.
Temperature
A technical setting that can affect how predictable or creative an AI response is.
Zero-shot
Asking AI to do something without giving it an example first.
Few-shot
Giving AI a few examples so it better understands the kind of answer you want.
Chain of Thought
A term used to describe step-by-step reasoning inside an AI system. You may hear this discussed in relation to more advanced reasoning models.
Prompt Injection
When someone tries to manipulate an AI system by giving it instructions designed to override its original rules.
Jailbreak
An attempt to get an AI system to ignore its safety or behaviour restrictions.
Custom Instructions
Standing preferences you give an AI once so it applies them to every chat, such as your tone, role or what to avoid.
Persona
A character or role you ask the AI to take on so its answers fit a particular point of view.
Output Format
The shape you want the answer in, such as a table, bullet points, an email or a short summary.
Iteration
Improving a result through a few rounds of feedback rather than expecting perfection first time.
Regenerate
Asking the AI to have another go at the same request, which usually gives a different answer.
Agent
An AI system that can carry out tasks rather than simply answer a question.
AI Agent
An AI assistant given a role, instructions, tools or access to information so it can complete work.
Agentic AI
AI that can take multiple steps towards a goal with less hand-holding.
Multi-Agent System
A setup where several AI agents work together, often with different roles.
Autonomous Agent
An agent designed to operate with very little human input once it has been given a goal.
Human in the Loop
A process where a person still reviews, approves or guides what the AI is doing.
Automation
A process that runs automatically when something happens.
Workflow
The sequence of steps used to complete a task.
AI Workflow
A workflow where AI handles one or more of the steps.
Orchestration
The process of coordinating multiple tools, agents or systems so they work together.
Trigger
The event that starts an automation or workflow.
Tool Use
When an AI system can access another tool, service or function to complete a task.
Function Calling
A technical method that lets an AI model trigger specific tools or actions.
Integration
A connection between two apps so they can share information or work together.
Webhook
An automatic message one app sends to another the moment something happens.
Automation Platforms
Tools such as Zapier, Make or n8n that connect apps and run workflows without code.
AI Assistant vs AI Agent
An assistant mostly answers and helps when you ask. An agent can go off and complete a task, taking actions along the way.
Vibe Coding
Building software by describing what you want in normal language and letting AI generate much of the code. You focus more on what you want the product to do than on writing every line yourself.
No-Code
Building websites, apps or systems without traditional programming.
Low-Code
Building using visual tools and pre-built components, sometimes with a small amount of code.
AI Coding Assistant
An AI tool that helps write, explain, debug or improve code.
Copilot
A general term for AI that works alongside you while you complete a task. Some companies also use Copilot as a product name.
Wrapper
A product built around an existing AI model or API, usually adding its own interface, workflow or features.
API
A way for different pieces of software to communicate with each other.
MCP / Model Context Protocol
A standard designed to let AI assistants connect to other tools, apps and data in a consistent way, a bit like a universal plug.
Prompt-to-App
Describing an app in words and having an AI builder create a working version of it.
Lovable
An AI app builder where you describe what you want in plain language and it builds a real website or app you can keep improving.
Frontend
The part of a website or app you see and click on.
Backend
The behind-the-scenes part that stores data, handles logins and makes things actually work.
Database
An organised place where an app keeps its information, like a very well-behaved spreadsheet.
Deploy / Publish
Putting your app live on the internet so other people can use it.
Hosting
The service that keeps your website or app running online, day and night.
Domain
Your web address, such as lucystudio.dev.
Bug
Something in the software that is not working the way it should.
GitHub
A popular place where code is stored, tracked and shared, so changes can be saved and undone.
SaaS / Software as a Service
Software you use online, usually on a subscription, rather than installing it yourself.
Open Source
Software whose code is publicly available for anyone to use, study or improve.
GPU
The powerful computer chips used to train and run AI models. They are expensive and in high demand.
Compute
The raw computing power needed to train or run AI. More compute usually means more cost.
Transformer
The type of AI design behind most modern language models. It is the T in GPT.
Weights
The learned numbers inside a model that hold what it has picked up during training.
Open Weights
A model whose weights are released publicly so others can run or adapt it, even if other parts stay private.
Fine-tuning
Giving an existing model extra training on specific examples so it gets better at a particular job.
Distillation
Training a smaller model to copy a bigger one, so it is cheaper and faster while keeping much of the skill.
RAG / Retrieval-Augmented Generation
Letting an AI look things up in your own documents before it answers, so replies are based on your information.
Embeddings
A way of turning words, images or documents into numbers so an AI can find things with similar meanings.
Vector Database
A database built to store embeddings and quickly find the most similar items.
Diffusion Model
A type of AI used to create images and video by gradually turning random noise into a picture.
Latency
How long you wait for a response. Low latency means quick.
AI Slop
Low-effort, mass-produced AI content that clogs up feeds and search results.
Hallucinate (everyday use)
Now used jokingly for anyone confidently making something up, not just AI.
It's Giving
Short for 'it gives the vibe of'. Used to describe the feeling something has.
Cooked
In serious trouble, or completely exhausted. Can also mean something is over.
Delulu
Short for delusional, often said affectionately about big dreams or wishful thinking.
Rizz
Charm or charisma, especially the persuasive kind.
Touch Grass
A friendly nudge to log off and spend time in the real world.
Doomscrolling
Endlessly scrolling through bad news or negative content even though it makes you feel worse.
The Algorithm
The system a platform uses to decide what you see in your feed.
Going Viral
When something spreads very quickly and widely online.
NPC
From video games, a 'non-player character'. Used to describe someone acting on autopilot.
Main Character Energy
Acting like you are the star of your own film. Can be a compliment or a gentle tease.
Brain Rot
The foggy feeling from consuming too much low-quality online content. Also the content itself.
Ratio
When a reply to a post gets more likes than the original, usually a sign people disagree with it.
FOMO
Fear of missing out. Very common with AI right now.
Lurker
Someone who reads everything online but rarely posts.
Hallucination
When an AI confidently gives an answer that is wrong or completely made up.
Bias
When AI reflects unfair patterns or assumptions from the data it learned from.
Deepfake
Realistic fake video, audio or images of real people, created using AI.
Guardrails
Rules and limits built into an AI system to stop it doing harmful or unwanted things.
Alignment
The work of making sure AI systems behave in line with human goals and values.
Responsible AI
Using and building AI in a way that is fair, safe, transparent and accountable.
Grounding
Tying an AI's answers to real, trusted sources so it is less likely to make things up.
Citations
Links or references showing where an AI got its information. Always worth checking.
Watermarking
Hidden markers added to AI-generated content so it can later be identified as AI-made.
Provenance
The record of where a piece of content came from and how it was made or changed.
Data Privacy
Protecting personal information and being careful about what you share with AI tools.
GDPR
The UK and EU rules about how personal data must be collected, stored and used.
Copyright
Who owns creative work, and a big open question for AI training and AI-made content.
Red Teaming
Deliberately trying to break or trick an AI system to find weaknesses before others do.
EU AI Act
European law that sets rules for AI based on how risky its use is.
AGI / Artificial General Intelligence
A hypothetical AI that could match humans across almost any task. Much discussed, not here yet.
AI Strategy
A plan for how a business will use AI to meet its goals, rather than trying tools at random.
AI Policy
Clear guidance for a team on which AI tools can be used, how and with what data.
Shadow AI
When people use AI tools at work without the company knowing or approving.
AI Literacy
Understanding enough about AI to use it well, question it and spot its limits.
AI Adoption
How far, and how well, people in an organisation are actually using AI.
Upskilling
Helping people learn new skills, such as using AI confidently in their role.
Augmentation vs Replacement
Whether AI helps people do their jobs better or takes over the work entirely. Most real use is augmentation.
AI-first
Designing a product, team or process with AI at its heart from the start.
AI Transformation
Big-picture change in how an organisation works because of AI.
Pilot
A small, time-limited trial of an AI tool or idea before rolling it out more widely.
Proof of Concept
A quick early version built to show an idea can work.
MVP / Minimum Viable Product
The simplest version of a product that real people can use and give feedback on.
ROI / Return on Investment
Whether what you get back from AI, in time, money or quality, is worth what you put in.
Custom GPT
A version of ChatGPT set up with its own instructions and files for a specific job.
Enterprise AI
AI tools built for larger organisations, with extra security, admin controls and data protection.
Data Readiness
Whether a business's information is organised and clean enough for AI to use well.