You Cannot Use a Feature You Do Not Have a Word For: Twenty Claude Terms, Explained
You cannot use a feature you do not have a word for. That is the quiet reason a lot of people plateau with AI assistants – not talent, not effort, but vocabulary. If you do not know that a context window exists, you cannot understand why a long chat starts forgetting things. If you have never heard of a system prompt, you will keep retyping the same instructions forever.
So here are the twenty terms from the reel, grouped by what they are about, defined in plain English – and, because the list mixes very different kinds of words, labelled honestly as product features, general concepts, or community shorthand.
Twenty terms, five families
Learning them as a list is hard. Learning them as five families is easy, because each family answers one question about how the assistant works.
| What it sees | What you set up | How it thinks | What it can do | What it hands back |
|---|---|---|---|---|
| Context Context window Turn System prompt |
About-me file Style file Project Memory Prompt template |
Extended thinking Maximum reasoning Positive prompt Hallucination |
Tool Tool use MCP Agent |
Artifact Slash command Skill |
Most people only ever learn the third column, because it is where the dramatic words live. The second column is where the time is saved.
All twenty, defined
The third column matters. Some of these are names of real product features; some are general ideas that apply to every assistant; a few are shorthand people use online that you will not find in any menu. Knowing which is which saves you hunting for a button that does not exist.
| Term | What it means | Kind |
|---|---|---|
| Context | Everything the model can see when it answers: your message, earlier turns, files and instructions. | General concept |
| Context window | The maximum amount of material the model can consider at once. Past it, older material drops out. | General concept |
| Turn | One exchange: your message and the reply. Long chats are many turns, and every turn carries the earlier ones. | General concept |
| System prompt | Instructions set before the conversation that shape behaviour throughout it. | General concept |
| About-me file | A document describing you – role, goals, preferences – that you give the assistant as standing context. | Community shorthand |
| Style file | A document describing how you write, used to shape tone. Claude also offers styles you can choose or customise. | Shorthand, close to a feature |
| Project | A workspace that holds files and instructions shared across every chat inside it. | Product feature |
| Memory | Information the assistant can carry from one conversation to the next. Availability depends on plan and settings. | Product feature |
| Prompt template | A reusable prompt with blanks you fill in each time. | General concept |
| Extended thinking | A mode where the model reasons step by step before answering, for harder problems. | Product feature |
| Maximum reasoning | Not a menu item – shorthand for giving the model as much thinking room as possible on a hard task. | Community shorthand |
| Positive prompt | Borrowed from image generation. For text, it means stating what you want rather than listing what you do not. | Community shorthand |
| Hallucination | A confident answer that is simply false. | General concept |
| Tool | A capability the model can call on: search, running code, reading a connected app. | General concept |
| Tool use | The model deciding to call a tool and working with what comes back. | General concept |
| MCP | Model Context Protocol – an open standard for connecting AI applications to outside tools and data. | Open standard |
| Agent | An assistant that takes a series of actions toward a goal instead of answering once. | General concept |
| Artifact | A standalone piece of output – a document, code, a small working tool – shown in its own panel so you can iterate on it. | Product feature |
| Slash command | A shortcut typed with a forward slash that triggers a predefined action. | Product feature |
| Skill | A packaged set of instructions and resources the model loads for a specific kind of task. | Product feature |
The post this came from
Watch the reel
All twenty terms on one card. Posted on the BANI Academy page.
Three things change once you know the words
The reel promises three changes. Each one comes from a specific handful of terms, and it is worth knowing which.
| What changes | The words responsible | What you start doing differently |
|---|---|---|
| Your prompts get sharper | Context, positive prompt, prompt template | You hand over the material the answer depends on, say what you want instead of what you fear, and stop writing the same request from scratch. |
| Your workflows get repeatable | Project, system prompt, about-me file, style file, skill | You write your context and rules down once, and every conversation starts from them instead of from zero. |
| You stop being surprised | Hallucination, context window, turn | You check factual claims, start a fresh chat when a long one drifts, and stop assuming the model remembers what scrolled away. |
The first row is the five-slot pattern we set out in vague in, vague out. The second is the ten-minute setup in stop re-explaining yourself. And the third explains the mechanism behind most wasted usage – every turn carries the earlier ones, which we unpacked in you are paying for retries.
The four words worth learning first
If twenty is too many for this week, start with four. They cover most of the gap between people who get generic answers and people who do not.
Context. Every good answer depends on what the model can see. Most bad answers are not the model being weak; they are the model being made to guess.
Context window. Once you know it exists, you understand why very long chats degrade, and you start fresh conversations on purpose instead of by accident.
Hallucination. Naming it makes it checkable. A fluent answer is a draft, not a source – true of every assistant, not one in particular.
Tool use. The line between an assistant that talks and one that acts. It is also the point where access becomes a decision, which is why the rules in stop pasting start with read-only.
Learn those four and the other sixteen stop feeling like jargon. They become names for things you have already noticed. And for the one-layer-down view of why these ideas matter more as the tools get easier, see the tools did not disappear.
Frequently asked questions
Are these terms specific to Claude?
Some are product features with Claude’s names on them. Most are general ideas that apply to every assistant, and the third column of the table says which.
What is the difference between a project and memory?
A project holds files and instructions you set up deliberately for one piece of work. Memory is information carried between conversations more generally. Use projects for anything you want controlled.
Is MCP something I need to learn?
Not to use an assistant. It matters if you want to connect one to your own tools or data – and then mainly as the reason connections are possible at all.
Why do long conversations go wrong?
Because every turn carries the earlier ones, and eventually the window fills. A new chat with a short summary usually fixes it.
One family a week
Pick the family that matches your biggest frustration. Generic answers? Start with what it sees. Retyping the same instructions? What you set up. Confident nonsense? How it thinks. Learn those words this week, and use each one deliberately at least once.
A month from now the whole vocabulary will be yours, and so will most of the capability behind it.
Learning a lot does not make anyone good. Doing a lot does.
If you want the structure that makes that happen on schedule, that is what our free challenge is built for: no barrier to entry, a genuine commitment to action, and a refundable commitment fee you get back when you finish. You are not paying for the knowledge. You are betting on yourself finishing, and we hold the stake.
Which of the twenty did you not know? Tell us in the comments on the original reel.
For more on building channels for English-speaking audiences, that is what we work on at mmoyoutube.com.
2 Comments