Day 2: The Eight Words That Unlock Everything

The Concept

There is a particular kind of frustration that comes from trying to learn something when every other sentence contains a word you do not quite understand.

AI writing is full of that experience. People talk about prompts, hallucinations, context windows, and large language models as though the meaning is obvious. If you have not been immersed in the field, the whole thing can feel like a different language.

Today solves that.

You do not need to understand the engineering behind AI to use it effectively, any more than you need to understand how an engine works to drive a car. But you do need enough vocabulary to follow a conversation, read a tutorial, compare tools, and recognise when someone is making a claim you should question.

Today's goal: learn eight practical AI terms well enough that you can explain them in your own words.

This is not about sounding technical. It is about removing friction. Once the vocabulary becomes familiar, AI stops feeling like a mysterious category and starts feeling like a set of tools you can evaluate calmly.

The eight terms

These eight words will appear again and again throughout this course. Think of them as the basic map legend. You do not need to memorise every detail today. You need to recognise each term, know why it matters, and understand what it means in everyday use.

1. LLM

LLM stands for large language model.

This is the technical name for the type of AI you are using when you chat with tools like ChatGPT, Claude, Gemini, or similar assistants. "Large" refers to the scale of the system. "Language" refers to what it is trained to work with. "Model" means the underlying system that produces the response.

Plain English: An LLM is an AI system trained to generate and transform language.

A useful analogy is a highly trained writing partner. It has seen many examples of how language is used, so it can continue patterns, rewrite ideas, explain concepts, and produce drafts. It does not understand your life automatically. It works from the words and context you provide.

Why this matters: When people say "the model gave me a better answer" or "try a stronger model," they are talking about the underlying AI system, not the app interface around it.

2. Prompt

A prompt is the instruction or message you give to AI.

"Summarise this article" is a prompt. "Help me write a difficult email to my manager" is a prompt. "Turn these notes into a checklist" is a prompt. The prompt is your brief.

In everyday work, the quality of the prompt often determines the quality of the first answer. A vague prompt asks the model to guess. A clear prompt gives it the job, context, constraints, and preferred format.

You do not need to write elaborate prompts for everything. But you should get into the habit of including the basics:

• Context: what the situation is.
• Goal: what you want the output to help you do.
• Constraints: what the answer should avoid or respect.
• Format: what shape you want the response in.

Why this matters: Prompting is not a technical trick. It is the skill of giving a useful brief.

3. Hallucination

A hallucination is when AI generates something false with the same confident fluency it uses for something true.

It might invent a statistic, misstate a policy, describe a source that does not exist, or present a guess as if it were confirmed. The important part is not just that the answer is wrong. The important part is that it may sound right.

This is one of the first terms every AI user should understand because it explains why polished output still needs judgment. AI can be useful and unreliable at the same time, depending on the task.

Use this rule: If the answer contains facts you plan to rely on, verify them.

Why this matters: Hallucination is the reason AI should not be treated as an automatic authority, especially for current information, numbers, legal issues, medical topics, financial decisions, safety concerns, or public claims.

4. Context window

The context window is the amount of text AI can work with in one conversation.

Your messages, the AI's responses, pasted documents, instructions, and uploaded content all count toward that limit. When a conversation gets very long or a document is very large, the tool may not be able to use everything equally well.

An everyday analogy is a desk. A larger desk can hold more papers at once. A smaller desk forces you to choose what stays in front of you. The context window is the AI's working desk.

In normal use, you may not think about this often. But the term becomes helpful when:

• a conversation has gone on for a long time,
• you paste a very large document,
• the AI seems to forget earlier details,
• or you need it to compare several pieces of information at once.

Why this matters: AI works better when the relevant information is present, clear, and not buried inside a long messy conversation.

5. Token

A token is a small unit of text that AI uses to process language.

You can think of tokens as pieces of words. A token is not exactly a word, but for practical purposes, you can treat it as the way AI measures how much text it is reading or writing.

Most everyday users do not need to count tokens manually. Still, the term appears in product pages, pricing, developer documentation, and tool limits. If a tool says it supports more tokens, it usually means it can handle more text in a prompt, a document, a conversation, or an answer.

Why this matters: Token limits affect how much AI can read, remember within a conversation, and generate in one response.

6. Training data

Training data is the material the AI learned from during development.

That can include books, articles, websites, code, documentation, and many other forms of text or media, depending on the model. Training data shapes what the model is good at, what it has seen often, what it may handle poorly, and what kinds of assumptions or biases can appear in its answers.

This does not mean the model is looking up a specific training example every time it responds. It means its patterns were shaped by the material used to train it.

Why this matters: Training data helps explain both capability and limitation. If the model has learned from broad language patterns, it can write and explain many things. If the training data has gaps, outdated material, or biased patterns, the model can reflect those problems too.

7. Model

A model is the specific AI system producing the answer.

Different tools may offer different models, and the same company may offer several versions. Some are faster and cheaper. Some are stronger at reasoning, writing, coding, or working with long documents. Some are designed for everyday speed. Others are better for harder tasks.

Practical point: Not every AI answer comes from the same underlying system.

If you try a task and get a weak result, the issue may be:

• the prompt was too vague,
• the model was not strong enough for the task,
• the tool lacked important context,
• or the task required verification rather than generation.

Why this matters: Knowing the word "model" helps you understand why the same prompt can perform differently across tools or settings.

8. Generative AI

Generative AI is AI that creates new content.

It can generate text, images, code, audio, video, summaries, outlines, drafts, examples, and ideas. This distinguishes it from older forms of AI that mainly classified, ranked, recommended, or detected patterns.

You have probably used AI for years without calling it AI: spam filters, recommendation systems, route suggestions, search ranking, and fraud detection. Generative AI feels different because it produces something new in response to your request.

Why this matters: Generative AI is powerful because so much everyday work depends on creating, reshaping, or improving information. But because it generates, you still need to review and refine what it gives you.

How these terms work together

The terms are easier to remember when you see how they connect.

When you use AI, you give it a prompt. The prompt goes to a model, often an LLM, which works within a context window. The text is processed in tokens. The model's behaviour has been shaped by training data. Because it is generative AI, it creates a new response. Sometimes that response may contain a hallucination, so you use judgment and verification where it matters.

The whole map: Prompt -> model or LLM -> context window -> tokens -> generated response -> review for hallucinations when facts matter.

If you understand that flow, most AI explanations become easier to follow.

Making the vocabulary stick

Vocabulary does not stick because you read it once. It sticks when you use it.

That is why today's task asks you to choose the two terms that surprised you most and write them in your own words. The surprise is important. A term usually becomes memorable when it changes what you assumed.

Maybe you assumed a hallucination would look obviously wrong, and now you realise it can sound polished. Maybe you assumed a prompt was just a question, and now you see it as a brief. Maybe you assumed AI "knows" things in the way a person does, and now you understand that training data shapes patterns rather than guaranteeing truth.

Those small shifts matter. They change how you use the tool.

Use this today

Run the prompt from today's task. When the answer comes back, do three things:

• Underline the two most useful terms. These are the words that will help you most in future lessons.
• Rewrite them in your own language. Avoid copying the AI's definition directly.
• Attach each term to a real example. Use your job, home life, hobbies, study, or a recent task.

For example:

• "A prompt is the brief I give AI. If my brief is vague, the output will probably be vague."
• "A hallucination is a confident-sounding mistake, so I should check facts before I use them."
• "The context window is the information AI can keep on the desk during a conversation."

This takes only a few minutes, but it turns vocabulary into judgment.

What to notice

As you complete the exercise, pay attention to which terms feel immediately useful and which still feel abstract.

If a term feels abstract, ask the AI for a personal example:

"Use the term context window in an example about preparing for a team meeting."

"Explain hallucination using an example from travel planning."

"Show me the difference between a vague prompt and a strong prompt for writing a customer email."

Personal examples are powerful because they attach the word to something you already understand.

Remember this

If you remember nothing else from Day 2, remember these three ideas:

• A prompt is your brief.
• A model is the system producing the answer.
• A hallucination is a confident-sounding mistake.

These eight terms cover the vast majority of what you will encounter. You do not need to become technical. You need to become fluent enough that AI conversations stop feeling like a foreign language.

That fluency starts today.