BeginnerLesson 1 of 307 min read

What Claude Is and Getting Started

Learn what Claude is and is not, where you can use it, how to set it up in five minutes, and what tokens are, with a program that estimates the size of messages.

In this lesson you will

  • Explain what Claude is and is not
  • Know where you can use it
  • Set up and look around in five minutes
  • Understand tokens roughly

About 20 minutes

Noor has an essay due and no idea where to start

Noor is a first-year student with an essay due on Friday. It is Tuesday night. She has read the chapter twice and still cannot say what her argument is. A friend says, ‘Ask Claude.’ Noor has heard of it, but she has a dozen questions. Is it a search engine? Will it write the essay for her? Will it get things wrong? Is it safe to type her work into it? This lesson answers those questions, one at a time, in plain language, and gets you ready to have your first conversation in the next one.

What Claude is

Claude is an AI assistant made by a company called Anthropic. You talk to it in ordinary language, by typing or, in some apps, by speaking, and it replies in ordinary language. It can explain ideas, draft and edit writing, summarise long text, help plan, answer questions about documents you give it, write and fix code, and work through problems step by step. It is built on a large language model: a program trained on a very large amount of text, which learns patterns in language and uses them to produce a useful next piece of text, over and over, until it has written a whole reply.

That description matters because it explains both Claude’s strengths and its limits. It is very good at language: explaining, rephrasing, organising, brainstorming and drafting. It is not a database that looks up facts, and it is not a person who has lived your life. It can be wrong, and sometimes it is wrong in a confident, fluent way. Anthropic’s own help pages say that Claude can occasionally produce responses that are incorrect or misleading, and that people should check important information. Treat Claude as a quick, capable helper whose work you review, not as an authority.

What Claude is not

  • Not a search engine. A search engine finds existing pages. Claude writes new text. Some versions of Claude can also search the web when the feature is available and switched on, but that is an extra tool, and you should still check the sources it shows.
  • Not a mind reader. It only knows what you tell it in the conversation, plus what it learned in training and anything you provide, such as pasted text or files.
  • Not always up to date. What it learned in training stops at some date. For recent events and changing facts, check a current source.
  • Not a replacement for professional advice. It can help you prepare questions for a doctor, lawyer or accountant, but it does not replace one.

Where you can use it

Anthropic’s help pages list three main ways to use Claude: in a web browser at claude.ai, in a desktop app for Mac or Windows, and in a mobile app for iPhone or Android. They connect to the same account, so you can start on your phone and carry on later on a laptop. Creating an account is free, and a free plan exists, but plans, prices and limits change and differ by country, so this course never quotes them. Check the current plans page when you need to know what you get. The model you are talking to is shown near the text box on web and desktop, or at the top of the screen on mobile, and which models are offered depends on your plan.

Your setup in five minutes

  1. Go to the Claude website and create an account, or sign in with the method it offers.
  2. Open a new conversation. You will see a text box and a button to send.
  3. Look around the screen once: the list of past conversations, the text box, and a button to attach files if your plan offers it.
  4. Open the settings and look at the privacy options. Lesson 9 returns to these, but it is good to know where they are.
  5. Type a simple first message, such as ‘Explain what you can help me with, in five short bullet points.’ and read the reply.

Words and numbers you will meet

You will often hear about tokens. Models do not read words exactly as you do. They break text into small pieces called tokens, which can be a whole short word, part of a longer word, or a bit of punctuation. Tokens matter because tools set limits on how much text a conversation can hold, and the limit is measured in tokens, not words. You do not need to count them, but a rough feeling helps. The short program below estimates the size of three messages using two common rules of thumb: about four characters per token, and about one and a third tokens per word. These are rough guesses, because the real count depends on the tool, so we show both and label them as estimates.

import math, re

messages = {
    "short question": "What is a good name for a bakery?",
    "email help": ("Please write a polite email to my landlord asking when the heating "
                   "will be fixed. It has been broken for two weeks and I have a small child."),
    "pasted notes": ("Meeting notes: budget approved for the autumn fair. Volunteers needed "
                     "for tables, tea and clean-up. Posters due Friday. Printer wants files "
                     "with bleed. Ask Anil about the banner. " * 6),
}

def estimate(text):
    chars = len(text)
    words = len(re.findall(r"\S+", text))
    tokens_a = math.ceil(chars / 4)           # rule of thumb: about 4 characters a token
    tokens_b = round(words * 1.3)             # another rule of thumb: about 1.3 tokens a word
    return chars, words, tokens_a, tokens_b

print("name: characters, words, tokens (two guesses)")
for name, text in messages.items():
    c, w, a, b = estimate(text)
    print(f"{name}: {c}, {w}, {a} or {b}")

c, w, a, b = estimate(messages["pasted notes"])
print()
print(f"reading time of the notes: {w / 200:.1f} min")
print("a rough guess only: real counts depend on the tool")
name: characters, words, tokens (two guesses)
short question: 33, 8, 9 or 10
email help: 140, 28, 35 or 36
pasted notes: 1074, 168, 269 or 218

reading time of the notes: 0.8 min
a rough guess only: real counts depend on the tool

The short question has 33 characters and 8 words, so the two guesses are 9 and 10 tokens, close together. The meeting notes are 1074 characters and 168 words, and the guesses are 269 and 218, further apart. The gap between the two rules widens with long text, which is a reminder that these numbers are estimates. The practical point is that a few pasted pages are a larger load than a short question, and a long conversation keeps growing. A person would read the notes in about 0.8 minutes. The estimate is not what Claude charges or what the limit is; it just helps you picture how much text is going in.

Good first uses

  • Explain something you find confusing, at the level you choose.
  • Turn messy notes into a clear list.
  • Draft a first version of an email, a letter or a plan, which you then improve.
  • Ask for three different ways to say something.
  • Check your own writing for unclear sentences.

Noor tries the fifth idea. She pastes her messy paragraph and asks, ‘Which sentence is the least clear and why?’ Within a minute she knows exactly where her argument goes vague. She did not ask Claude to write the essay. She used it as a patient reader. That is a habit worth keeping: use Claude to help you think and write better, and check your own institution’s rules about AI help in assessed work.

Common mistake: asking without context

Expecting Claude to know your situation without being told. If you ask ‘Is this a good plan?’ without sharing the plan, the goal and the limits, any answer will be generic. The next lessons are all about giving it what it needs.

Screens change, habits last

Names, screens and features change quickly. This course describes ideas that stay the same: how to ask, how to check and how to stay safe. If a button has moved, look for the same idea elsewhere on the screen.

Open Claude and ask it to explain one thing you find confusing at the level of a twelve-year-old. Then ask it to explain the same thing for an expert. Compare the two replies. This shows you straight away how much the way you ask changes what you get.

Practice for this lesson

Check what you learned

Five quick questions. Pick an answer and you will see straight away whether it is right and why.

What is Claude?

  1. A search engine
  2. An AI assistant you talk to in ordinary language
  3. A spreadsheet
  4. A web browser

Claude is an AI assistant that reads your message and writes a reply in ordinary language.

What should you do with Claude’s answers on important matters?

  1. Trust them completely
  2. Ignore them
  3. Delete them
  4. Check them

Claude can be wrong, sometimes confidently, so important information needs checking.

Where can you use Claude?

  1. Web, desktop and mobile apps
  2. Only on a phone
  3. Only in a library
  4. Only in email

Anthropic lists the web, a desktop app for Mac and Windows, and mobile apps.

What is a token?

  1. A password
  2. A kind of file
  3. A small piece of text a model reads
  4. A payment

Models split text into small pieces called tokens, and limits are measured in them.

Why did the two token guesses differ more for long text?

  1. The text changed
  2. Claude refused
  3. The tool was slow
  4. They are rough rules of thumb

The two rules are only estimates, and the gap grows with length.

Summary

  • Claude is an AI assistant that writes replies in ordinary language, and it can be wrong.
  • You can use it on the web, on desktop and on mobile with one account.
  • Tokens are small pieces of text, and limits are measured in them.
  • Give it the context it needs, and check what matters.

Next lesson

Lesson 2 walks through your first real conversations.

Questions

Is Claude free to use?

There is a free plan, and paid plans with more features and higher limits. Plans, prices and limits change and differ by country, so check Anthropic's current plans page.

Can Claude do my homework for me?

It can help you understand and write better, but handing in work you did not do may break your school's rules. Check your institution's policy on AI help.

Is Claude always right?

No. It can produce confident mistakes. Anthropic's own help pages advise checking important information.

Progress is saved in this browser.