Glossary, cheat sheet, and next steps

Your permanent reference for the whole course — plus a clear path forward.

You've reached the final lesson. This page is designed to live in your bookmarks. It has the full glossary you can search, an interactive cheat sheet you can print, a 4-week action plan with checkboxes, and a curated list of where to go next.

How to use this page: Work through the content once. Then bookmark it. Come back when you need to look something up, print the cheat sheet, or check off your weekly progress.

The full glossary

Twenty-plus terms you've met across this course. Search by name or definition, or filter by category.

AI Software doing tasks that normally need human intelligence — understanding language, recognising images, answering questions, writing, planning, and predicting. Concepts
Machine Learning (ML) AI that learns patterns from data instead of hand-written rules. "Here are 10 million emails — figure out what spam looks like." Concepts
Deep Learning ML that uses multi-layer neural networks. Responsible for most of the progress in AI since ~2012. Concepts
LLM Large Language Model — a text-trained model like Claude, ChatGPT, or Gemini. Predicts the next token based on everything it has read. Concepts
Generative AI AI that creates new content — text, images, audio, video, or code. LLMs are one type; image generators (Stable Diffusion, DALL·E) are another. Concepts
Model The trained AI itself — the swappable "engine." Claude Opus, Llama 3, Gemini Flash are all different models. Concepts
Parameters A model's internal knobs — "7B" means 7 billion parameters. More parameters generally means smarter but also heavier and slower. Concepts
Prompt What you type to the AI. The whole craft of getting good results is learning to write better prompts. Concepts
Token A chunk of text — roughly ¾ of a word. AI reads and charges by the token, not by words or characters. Concepts
Context window How much text the model can consider at once. Past the limit, the model "forgets" the beginning of your conversation. Concepts
Multimodal A model that handles more than text — also images, audio, video, or documents. Most modern frontier models are multimodal. Concepts
Agent AI that takes multi-step actions toward a goal rather than just answering once. It can call tools, browse the web, or run code. Concepts
RAG Retrieval-Augmented Generation — feeding the AI relevant snippets of your own documents so it can answer from them instead of guessing. Concepts
AGI Artificial General Intelligence — hypothetical human-level general AI that can do anything a human can. Does not exist yet. Concepts
API Application Programming Interface — the way your programs send prompts to and get answers from an AI service over the internet. Tools
API key Your secret password for using an API. Anyone who gets it can bill to your account. Store it in a .env file and never commit it to git. Tools
CLI / terminal The text window where you type commands directly to your computer. Used to run Ollama, Python scripts, and Claude Code. Tools
Local model A model running entirely on your own computer. No data leaves your machine — useful for privacy, offline work, or just experimenting for free. Tools
Ollama A command-line tool for downloading and running local models. One command to download a model, one to run it. Tools
LM Studio A desktop app (Mac/Windows/Linux) for running local AI models with a friendly GUI — no terminal required. Tools
Claude Code Anthropic's AI coding tool that runs in your terminal and can read, write, and edit files in your project. Great for code generation and refactoring. Tools
Hallucination Confident, false AI output — stated as fact, with the same fluent tone as correct answers. You cannot tell from tone alone. Always verify what matters. Risks
Bias AI inheriting unfair patterns from its training data — over-representing some groups, under-representing others, or reflecting historical discrimination. Risks
Prompt injection An attack where malicious text in input data tricks the AI into following hidden instructions — bypassing your intended instructions. Risks

No terms match “”. Try a different search.

One-page cheat sheet

Everything you actually need to remember, compressed into six cards. Click any card to expand it. Hit Expand all then ⌘P to print a one-pager.

To just use AI ▶

Go to claude.ai and type. That's it.

Free tier is real — you can do useful work without spending anything. Create an account with a Google or email login and start a conversation.

Other options: ChatGPT (chatgpt.com), Gemini (gemini.google.com). They all work similarly from the chat interface.

To get good answers ▶

The recipe: Role + Context + Task + Format.

"You are a friendly lawyer. I have a dispute with a landlord about a broken boiler [context]. Explain what rights I have under UK law [task]. Use bullet points, plain English, no jargon [format]."

Tips:

• Show an example output if you have one.

• If the first answer is wrong, say what was wrong and ask again — don't start over.

• Be specific. Vague question = vague answer.

To run AI privately on your computer ▶

Easy route (GUI): Download LM Studio, open it, search for a model (try Llama-3.2-3B for speed, Mistral-7B for quality), download, run.

Terminal route: Install Ollama, then:

ollama pull llama3.2:3b

ollama run llama3.2:3b

Nothing leaves your computer. No API key. No costs.

To let AI write & edit code ▶

Install Claude Code (requires Node.js 18+):

curl -fsSL https://claude.ai/install.sh | bash

Then navigate to your project folder and run claude. It reads your files and can edit them directly.

Alternative: Cursor (cursor.sh) — a VS Code fork with AI built in; good for beginners who prefer a traditional editor.

To automate with code ▶

1. Get an API key at console.anthropic.com.

2. Store it in a .env file: ANTHROPIC_API_KEY=sk-ant-...

3. Install the SDK: pip install anthropic python-dotenv

4. Start with a cheap model (claude-haiku-4-5) and set max_tokens to cap costs.

5. Only upgrade to a larger model when you've confirmed the logic works.

Always (the three rules) ▶

1. Verify what matters. AI is a confident guesser, not a fact database. Check anything consequential against a real source.

2. Protect private data. Don't paste passwords, patient records, or confidential company data into a cloud chat box.

3. Keep a human in the loop. For anything with real-world consequences — medical, legal, financial, safety-critical — a human makes the final call.

A suggested 4-week path

This is not a course. It's a practice plan. Each week adds one new capability. Check off tasks as you go — your progress is saved in your browser.

Where to go next

A curated selection of resources, grouped by what you want to do. All open in a new tab.

The whole secret: You don't need to understand how an engine works to drive a car. The same is true here. Start with Chat, build the verify-everything habit, and let curiosity pull you toward the next thing. That's it.