Starting From Absolute Zero

Skip this if you have used one of these tools before. Genuinely — it is here so nobody has to pretend, and there is nothing in it you will want if you already have an account and have typed into it. Module 2 is where the work starts.

Still here? Good. Module 1 spent a while arguing that AI is a production department rather than a replacement. That argument assumed you know what the thing is. This module does not assume that.

Everything below is written for someone who has never opened one of these tools and is slightly suspicious of the whole business. That is a reasonable place to be standing.

1. What Claude actually is

A program you type to, which types back.

That is not a simplification for beginners. It is the accurate description. You write something in a box, it writes something underneath, and you carry on from there.

Explain-on-first-use: the thing doing the writing is a model — software trained on an enormous quantity of text until it got good at continuing text sensibly. Claude is Anthropic's model, and also the name of the app you use it in. When this guide says "Claude", it means the app you type into.

Three things it is not, each of which trips up someone new:

• Not a search engine. It is not looking through a list of pages for yours. It can search the web when you ask it to, but that is a thing it does, not what it is.
• Not a database. It does not "have" your files or your client's brief unless you give them to it.
• Not a person, and not trying to be. It has no memory of you unless you turn one on, no opinion it will defend past your pushing back, and no stake in the work.

And one thing it is, which nobody says early enough:

That is not a reason to avoid it. It is the reason the whole of this guide is built around checking, sourcing and showing your work rather than trusting output.

2. Signing up

Go to claude.ai and make an account. That is the whole process, and it costs nothing.

Free is the path this guide is written for

You can complete this entire guide on a free account. Not a stripped-down version of it — the actual thing. Module 2's teaching, the script templates, both generation Skills, the brand-voice workflow, the portfolio builder. All of it.

Verified 2026-08-14, the free plan includes:

• Chat on web, iOS, Android and desktop
• Web search — it can go and look things up when asked
• Cross-conversation memory (see §3 — it's opt-in)
• File creation
• Extended thinking for harder problems
• Connectors, including Google Workspace
• Five Projects (Module 2 §8 explains what those are and why five is enough to work with)

Usage limits apply on every plan, free included. You will meet them if you work hard at it, and they reset.

What Pro adds, and when it's worth it

Pro was $17/month on an annual subscription ($200 billed up front) or $20 billed monthly, verified 2026-08-14. Check the current price — this is a dated figure, not a fixed one.

For a reader of this guide, Pro adds three things that matter:

Do not upgrade yet. You do not know what you would be buying. The honest trigger, and Module 2 §6 says the same: upgrade when you can name the specific job costing you an hour a week. Not before. Paying for a subscription is not a workflow, and a beginner who upgrades on day one usually just has a subscription.

3. What a chat is — and what does not carry between them

A chat (or conversation) is one continuous exchange. You type, it replies, you type again. Everything said so far in that chat is available to it — this is called the context of the conversation.

Explain-on-first-use: context is simply everything the model can currently see: this conversation so far, plus anything you have attached or loaded. If it is not in context, it does not exist as far as the reply is concerned.

The rule that catches everyone

Start a new chat and you are starting from nothing.

The new chat does not know what you agreed in the last one. It does not know your client, your brand rules, your niche, or that you spent an hour yesterday getting the tone right. Each conversation begins empty.

Think of it as a room rather than a person. What you say in the room is in the room. Walk into a different room and you are talking to someone who was not in the first one.

This is the concept Module 2 §8 builds an entire workflow on, so it is worth sitting with for a moment now rather than meeting it later as a nasty surprise.

Memory — the honest version

There is a feature that softens this, and it is worth understanding precisely because half-understanding it is worse than not knowing it exists.

Verified 2026-08-14:

• Memory is opt-in. You turn it on in Settings, under Generate memory from chats. It is off until you do.
• With it on, Claude builds up a synthesis of things it has learned about you and your work, available to new conversations. Available on free, Pro and Max.
• Searching your past chats on demand is a paid feature (Pro, Max, Team, Enterprise).
• Projects each get their own separate memory space, kept apart from your other chats.
• Incognito chats save nothing to memory or history.

Now the part that matters for client work:

So: memory exists, turn it on if you like, and never rely on it for anything you would have to defend.

4. What a prompt is

This is the most important definition in the guide. Everything in Module 2 rests on it, and almost every beginner failure downstream traces back to getting it wrong.

Explain-on-first-use: a prompt is what you type. That's it — the message you send. The word sounds technical and isn't.

But what kind of thing you type is the whole game:

The difference, concretely

A search query is a handful of keywords you hope will match a document that already exists somewhere. You are fishing. best pan for renters is a search query. Search engines reward short and keyword-shaped.

An instruction describes a job you want done. You are briefing. It rewards the exact opposite: specific, complete, and full of things you don't want.

Watch what happens with the same subject:

As a search query:

hook ideas for pan video

You get something fluent, generic, and usable by anyone selling any pan. It reads fine. It is worth nothing, because nothing in what you typed ruled anything out.

As an instruction:

Write three opening lines for a 22-second testimonial about a
non-stick pan.

Never use "game-changer", "obsessed" or "literally".
Nine words maximum per line.
No question openers.
One concrete sensory detail in each.

Audience: someone in their late twenties who cooks four nights a
week in a rented kitchen and already owns one pan they resent.

Same subject. Different job. Every line after the first eliminates something, and the eliminating is what makes the output yours rather than everyone's.

Four things an instruction can do that a search cannot

1. Say what you don't want. You cannot tell a search engine never use this word. This turns out to be the highest-value thing you can write, and Module 3 builds a whole document out of it.
2. Take material. Paste in a brief, a transcript, your own notes, and ask for something to be done to it.
3. Be argued with. "Line two is the only one with a real detail. Give me five more like it and nothing like the others." There is no equivalent for a search box.
4. Be given a shape. Ten options. Numbered list. Nothing else. You are describing the deliverable, not hoping.

The reframe to carry into Module 2

You are not asking. You are briefing someone very fast, very literal, with no taste of their own, who will not ask you a follow-up question unless you invite one.

If you have ever written a brief for a photographer — or received a bad one from a client — you already know how this goes. Vague brief, generic result, and it is nobody's fault but the brief's.

Module 2 §1 turns this into a working method with four required elements. You do not need those yet. You need to stop typing keywords.

5. The three things that go wrong first

Not the only three. The three that happen in the first week, in roughly this order.

1. Treating it like a search engine

What it looks like: you type four keywords, get something bland, and conclude the tool is overrated.

What actually happened: you asked a generic question and got a generic answer, which is the correct response to what you sent. §4 above is the fix, and it is the whole fix.

The tell you've done it: your message is under about ten words and contains no constraints.

2. Accepting the first answer

What it looks like: you get something serviceable, use it, and quietly decide this is the ceiling.

What actually happened: you stopped at the draft. The first output is almost never the best available, and the conversation is where the value is.

The fix, and it is one line: "Give me ten more, and make them shorter." Or point at the one thing that worked and ask for more of that. Module 2 §1 calls this the highest-leverage move in the entire workflow, and it takes four seconds.

A useful ratio to expect now, so it doesn't feel like failure later: ask for ten, expect one you'd use. Seven in the bin is a normal, healthy session.

3. Expecting it to remember

What it looks like: you spent an hour yesterday getting a client's tone right. Today you open a fresh chat, ask for a script, and get something that sounds like nothing you agreed.

What actually happened: §3. New chat, empty room. Nothing malfunctioned.

The fix: put anything you'll need again into a file, and put that file somewhere it gets loaded automatically. That is exactly what Module 2 §8 is for, and it is the reason Module 3 has you fill in a form rather than just having a conversation about a brand.

6. The glossary

This guide introduces vocabulary and does not avoid it. The reasoning is in blueprint §1 and it is simple: avoiding the words does not make you more capable, it leaves you unable to read anyone else's documentation. So terms get explained where they first appear, then used normally.

This module owns the glossary — one plain-language line per term, with the module that defines it properly.

Everything defined in this module is already in there: Claude, model, chat, context, memory, prompt, project. Everything the rest of the guide introduces gets added there too, on first use.

How to use it: don't read it. Come back to it when a word appears that you half-recognise. That is what it is for.

7What to do next

1. Make the account. Free.
2. Type one instruction, not one search query. Use the pan example above, swapped for something you actually own. Read what comes back out loud.
3. Ask for ten more. Notice that they get better once you say what you didn't like about the first ten.
4. Then go to Module 2. You now know everything it assumes.

You do not need to understand how any of this works underneath, any more than you need to understand a lens to shoot on one. You need to know what it is, what it isn't, and that the quality of what comes out tracks the quality of what you put in.