You've been hearing about Claude for months and have never tried it seriously.
This is the guide you needed before starting.
🔖Save it. You'll want to re-read it when you use it seriously.
1. What is Claude?
Basically: an AI you can chat with. You ask it for things, it does them.
It writes, summarizes, analyzes documents, creates Excels with real formulas, searches the internet and, if you use the desktop app, works with your computer files directly.
The difference with ChatGPT is not that one is better.
It's that they have different personalities:
- Claude writes more naturally. Much less "corporate robot."
- It can read a 200-page PDF without losing the thread.
- Well configured, it remembers how you work.
That's what makes it different.

Claude Interface
2. How to get it (and what it costs)
Go to claude.ai
- Free: To browse. Limited messages. Useful to see if it convinces you.
- Pro ($20/month): The sweet spot. You have the best model, Projects, Skills, and Cowork. If you're going to use it more than 3 times a week, it's worth it.
- Max (from $100/month): For people who live inside AI and find Pro insufficient.
My rule for choosing:
Free if you're still in doubt.
Pro if you see that you open it every day.
Max if Pro becomes too small for you.

Tip: pay monthly for the first month. If after 3 weeks you don't open it, cancel. You lose $20, not $240.
3. The 3 "Claudes" that exist
This confuses everyone at first:
1. The usual chat. You enter claude.ai from the browser. You write, it responds. Start here if it's your first time.
2. The desktop app (Mac/Windows). Same as the browser but Claude can see your computer's folders. This activates Cowork.
3. Cowork. You give it a long task, "organize these 40 invoices and make me an Excel," and it gets to work alone for 15 minutes while you do something else. When you return, the work is in your folder.
Only with a paid plan. And the first time you see it work, you won't believe it.

You give it the task and it gets to work alone.
4. Projects: the function that has changed how I work the most
A Project is a folder within Claude where you put context.
Documents. Examples of how you write. Instructions. References.
Claude reads all that before responding to you.
What is it useful for in practice?
You have a blog:
→ you create a Project with your style and your best posts as examples.
You have a client:
→ you create a Project with everything Claude needs to know about them.
You are studying:
→ you create a Project with your notes and work on them.
Without a Project: you repeat context in every conversation.
With a Project: Claude already knows how you work.
The difference in the quality of the responses is huge.

A project for each thing. Claude already knows how you work.
5. Skills: Claude specialized in what you need
Skills are like "special modes" that you activate in Claude.
You teach it to do a specific task with your rules, your format, and your style. And it remembers them.
They work similarly to ChatGPT's GPTs, but Claude detects them automatically and loads them when needed. You don't have to remember to activate them every time.
Examples of what you can create:
- A Skill to edit your texts with your style guide.
- A Skill to structure meeting notes as you organize them.
- A Skill to write content with your voice, not the generic AI one.
Available on all plans, including the free one.
To start: type / in the chat bar and describe what you want it to do.
6. Artifacts: results you can use, not just read
When you ask Claude for something, it usually returns text.
With Artifacts, it returns something that works directly.
Examples:
→ An interactive calculator. → A task tracker with columns and statuses. → A comparative table with clean design. → A visual diagram of a process.
No need to install anything. It works within the chat itself.
Try this right now:
"Create an interactive weekly planner from Monday to Friday with time blocks from 9am to 6pm."
And see what it returns.

This is not text. Claude generated it inside the chat.
7. What it's great at and where it fails
Because nobody is perfect, not even Claude.
It's good for:
- Writing by adapting to your voice.
- Summarizing very long documents without getting lost.
- Thinking with you (as long as you challenge it when it's wrong).
- Creating Excels, documents, and presentations from scratch.
- Working with your files if you use Cowork.
It's bad for:
- Real-time information (activate web search first).
- Very precise mathematics (it's not a calculator).
- Generating images (use ChatGPT or Gemini for that).
⚠️ The most common error: asking for vague things assuming it guesses what you want.
It doesn't guess. It's very good, but it needs direction. You are the boss. It executes.
8. How to talk to it so it doesn't sound like a robot
If you've never written a prompt, this is for you:
Rule 1: Be specific. Don't say "write an email." Say "write an email to Sara, friendly but direct tone, maximum 5 lines, about the delay in delivery."
Rule 2: Give it examples. Paste something of yours and tell it "write to me like this." It learns faster from examples than from instructions.
Rule 3: Say what you want, not just what you don't want. "Don't sound formal" is worse than "write as if you were talking to a friend on WhatsApp."
Rule 4: Start short. Two sentences, see what comes out, and adjust. It's a chat, not a form.
Rule 5: If it starts saying nonsense, open a new chat. Very long chats get full and Claude loses the thread. Solution: new chat, paste the important stuff.
⚠️ Important: Claude will always give you an answer. It doesn't mean it's correct. Treat it as an intelligent partner, not as the absolute truth.
That's all you need to start.
No need to be technical. No need to know about AI. Just understand what functions exist and what each one is for.
Most people open Claude, write a question, and close it thinking it's like Google but slower.
That's using 5% of what it can do.
Save it for when you're going to try it seriously.
If you have questions, ask in the comments. I try to read everything.
Claude for the rest of the world
marcus.





