Before You Touch Anything

Let's start simple.

Claude is an AI assistant made by a company called Anthropic. You type something, Claude responds. It can write, analyse, reason through problems, help with code, summarise long documents, brainstorm ideas, and have extended conversations that actually go somewhere.

That's the surface. Here's what's useful to understand underneath it.

Claude works by predicting what text should come next, based on patterns it learned during training. It's read an enormous amount of human writing — books, articles, code, conversations — and it's gotten very good at producing responses that are relevant, thoughtful, and often genuinely insightful. It's one of the strongest AI models available today for tasks that require careful reasoning and nuanced writing.

But it has real limitations, and we should be upfront about them:

Claude doesn't know what happened today. It was trained on data up to a certain point. It doesn't browse the internet in real time (unless you're using specific features designed for that). So if you ask it about something that happened last week, it may not know.

Claude can be confidently wrong. This is the one that catches people off guard. Claude can produce an answer that sounds completely sure of itself but is factually incorrect. This happens less often than it used to, and it's better at flagging uncertainty now, but it still happens. Treat Claude's factual claims the way you'd treat a smart colleague's — trust but verify when it matters.

Claude doesn't remember you between conversations. Each new conversation starts fresh. Claude doesn't carry context from last Tuesday's chat into today's. (There are ways to work around this, which we'll cover in later modules. For now, just know that the default is a blank slate every time.)

Here's the thing, though: within a single conversation, Claude is remarkably capable. It can hold long, complex threads of reasoning. It can work with documents you share. It can adapt its tone and approach based on what you tell it about yourself and your needs. Most people underuse it because they treat it like a search engine — ask a question, get an answer, leave. Claude is more useful when you treat it as a working session.

We'll get into exactly how to do that in Module 2.

You've probably heard of all three. Let's be honest about each one rather than pretending Claude is the obvious best choice.

ChatGPT (by OpenAI)

The one everyone knows. ChatGPT has the largest user base, the broadest ecosystem (plugins, GPTs, image generation built in), and the most brand recognition. If someone says "AI" in casual conversation, they usually mean ChatGPT.

Where it's strong: Image generation and editing, voice conversations, the plugin/GPT store for specific use cases, and being the "safe default" that most people already have an account for.

Where it's weaker: For longer, more nuanced writing and reasoning tasks, Claude tends to produce more thoughtful output. ChatGPT can sometimes feel like it's optimising for sounding helpful over being genuinely thorough.

Gemini (by Google)

Google's AI assistant. Its biggest advantage is deep integration with Google's ecosystem — Gmail, Docs, Drive, Search.

Where it's strong: If your life runs on Google products, Gemini can pull context from your emails, documents, and calendar in ways the others can't. Its free tier is also quite generous.

Where it's weaker: For standalone tasks that don't involve Google products, Gemini can feel less polished than Claude or ChatGPT. It's improving quickly, but it's younger in its development.

Claude (by Anthropic)

Which is what this guide is about. So let's be honest about the trade-offs too.

Where it's strong: Long, careful reasoning. Nuanced, well-structured writing. Working with large documents (Claude can process very long inputs — up to the equivalent of a short novel in a single conversation). Being transparent about what it doesn't know. And, with Claude Code, the ability to work directly with your files and projects.

Where it's weaker: No built-in image generation (as of this writing). Smaller ecosystem of third-party plugins compared to ChatGPT. Less brand recognition, which means fewer people sharing tips and tricks online.

So which one should you use?

Honestly? The one that fits your actual workflow. If you live in Google's ecosystem, Gemini makes sense. If you want the broadest feature set and don't mind a sometimes-superficial feel, ChatGPT works. If you value careful, thoughtful responses and you're willing to invest a little time in learning the tool properly (which is what this guide is for), Claude is worth it.

Many people use more than one. That's fine too.

This guide focuses on Claude because we think it's genuinely underused — people hear it's good but don't know how to get the most from it. That gap is exactly what we're here to close.

Claude has a free tier. Let's start there, because the honest answer is: a lot of people don't need Pro.

What you get for free

• Access to Claude on the web (claude.ai) and the mobile app
• A limited number of messages per day (the exact number changes, but think of it as "enough for casual use, not enough for heavy work sessions")
• Access to Claude's capable but not top-tier model

What Pro gets you ($20/month)

• Significantly more messages per day
• Access to Claude's most capable models (better reasoning, better writing)
• Priority access when the service is busy
• Access to Claude Code (the command-line tool we'll cover in Module 3)
• Projects, which let you organise your work and give Claude persistent context

Who actually needs Pro?

You probably need Pro if:

• You use Claude regularly for work or study — multiple times a day, not once a week
• You hit the free tier's message limits and it frustrates you
• You want to use Claude Code
• You work with long documents or complex tasks where the top-tier model makes a noticeable difference

You probably don't need Pro if:

• You're still figuring out whether AI tools are useful in your life
• You use Claude occasionally for quick questions
• You're trying it alongside ChatGPT or Gemini and haven't settled on one yet

Our recommendation: Start free. Use it for a week or two. If you keep running into limits — if you find yourself wanting to use it more but can't — that's when Pro pays for itself. Don't pay before you've felt the need.

If you do go Pro, remember: $20/month is roughly the cost of one nice dinner out or a couple of streaming subscriptions. Framed against what it can do for your work, it's genuinely reasonable. But only if you actually use it.

This section matters because once you understand tokens, the anxiety about "am I wasting money?" gets a lot quieter.

What's a token?

A token is roughly 3-4 characters of text, or about three-quarters of a word. When you send Claude a message, that message gets broken into tokens. When Claude responds, its response is also measured in tokens.

You don't need to count tokens yourself. You'll never need to do mental arithmetic about this. We're explaining it because understanding the concept helps you develop good instincts about usage.

How this works on the free tier

On the free tier, you have a daily message limit. You don't pay per token — you just have a cap on how many conversations you can have per day. When you hit it, you wait until tomorrow. Simple.

How this works on Pro

Pro also doesn't charge per token. You pay $20/month and get a generous usage allowance. Think of it like a mobile data plan — you have a monthly quota, and most people never hit it unless they're doing something unusually intensive.

What counts as "intensive"? Processing very long documents, having extended multi-hour conversations, or using Claude Code for complex tasks. Normal daily use — writing help, research, brainstorming, document review — typically stays well within limits.

If you do hit the limit, you're not charged extra. You simply get temporarily slowed down (Claude will use a lighter model until your limit refreshes). No surprise bills.

Token-saving instinct

Here's the useful practical bit. Two habits that naturally keep your usage efficient:

1. Give context upfront. Instead of going back and forth five times to clarify what you need, spend a moment writing a clear first message. One well-written message costs fewer tokens than five rounds of "no, I meant..." More importantly, it gets you a better answer.

2. Don't restart conversations unnecessarily. Each new conversation starts from zero — Claude re-reads your system instructions, re-processes any context. If you're working on the same task, keep it in the same conversation.

That's really it. You don't need to obsess about this. Just know that clear communication and conversation continuity naturally save tokens — and also happen to make Claude more useful. Win-win.

If you take only one thing from this entire guide, make it this.

Most people use Claude like a search engine. They type a question, get an answer, and leave. That works, but it uses about 10% of what Claude can actually do.

Here's a better way to think about it:

Claude is like a very capable colleague who just joined your team today.

They're smart, well-read, and genuinely want to help. But they walked in this morning. They don't know your specific situation, your preferences, your constraints, or what you've already tried. If you just say "fix the report," they'll do their best, but they'll probably miss what you actually needed.

Now imagine you sit down with this colleague for five minutes first. You tell them what the report is for, who's going to read it, what tone you're going for, and what the last version got wrong. Suddenly, their help is dramatically better. Same person, same ability — you just gave them what they needed to actually help you.

That's Claude. The quality of what you get out is directly shaped by the context you put in. And "context" doesn't mean writing a novel before every question. It means:

• Who are you? A quick line about your role or situation. ("I'm a student writing a research paper" vs "I'm a marketing lead preparing a board presentation" changes Claude's entire approach.)
• What are you trying to do? The actual goal, not just the immediate question. ("Help me understand this concept" vs "Help me explain this concept to my team" gets very different responses.)
• What have you tried? If relevant. Claude can build on your existing work instead of starting from scratch.
• What does good look like? The tone, length, format, or style you're hoping for.

You don't need all four every time. Even one or two of these transforms the interaction.

Here's the lovely thing about this mental model: it makes Claude feel less intimidating, not more. You're not learning "prompt engineering." You're doing something you already know how to do — briefing a colleague. The skill transfers naturally.

And once you've experienced the difference between a zero-context question and a well-briefed one, you won't go back.

You can stop here if you'd like. This module was about building understanding, and if you've read this far, you have a solid foundation. You know what Claude is, how it compares to alternatives, what it costs, and how to think about working with it. That's plenty to start using Claude.ai effectively.

When you're ready to go deeper — to learn the specific patterns that make Claude.ai genuinely powerful in your daily work — Module 2 is waiting.