Why This Matters

If you cannot explain what a tool is, you cannot use it deliberately. Most Claude frustrations come from one of two misconceptions: treating it as a search engine ("it should know the exact fact I want") or treating it as an oracle ("it said this, so it must be correct"). A precise mental model prevents both.

Concept Explained

Claude is an AI assistant built by Anthropic. The word "assistant" matters: a good human assistant would help you draft, summarize, analyze, code, and brainstorm — but they would never replace your judgment about what to ship. Claude plays exactly that role, except it never tires, is always available, and is trained on a massive corpus of text and code. Technically, Claude is a large language model (LLM) — currently the Claude 4 family, with Opus 4 as the most capable, Sonnet 4 as the balanced default, and Haiku as the fast/cheap option.

How It Works

A Claude interaction has four phases: (1) you write a prompt; (2) Claude's service assembles context — your message, any files you attached, any system prompt, any prior messages in the conversation; (3) the model generates a response token by token based on patterns learned during training; (4) the response is streamed back to you. That loop repeats for every message. The most important shift in mindset: Claude does not retrieve answers from a database. It generates them, token by token, based on probabilities. That is its superpower (can produce text that has never existed before) and its risk (can produce confident-sounding nonsense — called hallucinations).

Step-by-Step Tutorial

1. Recognize the three primary surfaces you will touch daily

For most beginners, three surfaces cover 90% of daily use: claude.ai (the web chat interface), Projects (knowledge containers), and Artifacts (interactive shareable content). The other surfaces — Anthropic API, Claude Code CLI, MCP servers — become important later.

2. Have your first conversation

Go to claude.ai, sign in, and type a simple prompt like "Explain what an API is in 3 sentences." Read the response. Notice the tone — Claude tends to be thoughtful and nuanced, not breezy.

3. Try a creative task

Ask Claude to write a short poem about your hometown, or to draft an email apologizing for a missed meeting. Notice how Claude's tone shifts to match the task.

4. Try a reasoning task

Ask Claude a multi-step question: "If I have 3 apples and give 1 to a friend who already had 2, how many apples do we have between us?" Verify Claude reasons correctly (4 — you have 2, friend has 2).

5. Form your 60-second explanation

Write down what you would tell a colleague who asked "what is Claude?" Aim for three sentences: (1) what it is (AI assistant by Anthropic); (2) where it lives (claude.ai web, API, Claude Code); (3) what it's good and bad at (good at generating text/code from context, bad at being correct without verification).

Real-World Example

Imagine you're a marketing manager asked to write a blog post about a new product feature. You have the feature spec but no time to draft. You paste the spec into Claude and ask: "Draft a 600-word blog post for a technical audience explaining this feature, with a clear intro, 3 sections, and a call to action." Claude produces a structured draft in 30 seconds. You read it, edit the opening to match your brand voice, fix one technical inaccuracy, and ship. The 2 hours you saved go to thinking about distribution strategy — which is exactly what a good assistant frees you to do.

Example Prompts / Commands / Code

# Simple explanation
Explain what a JWT is in 3 sentences for a junior developer.

# Drafting
Draft a polite email to my team announcing a 1-week delay in our launch.
The delay is because we found a critical security bug. Apologize and
reassure them we have a fix plan. 150 words max.

# Reasoning
I have 12 hours of work and 8 hours until deadline.
Tasks: A (4h, must-do), B (3h, important), C (5h, nice-to-have).
What should I drop or defer? Explain your reasoning.

Common Mistakes

• Treating Claude as a search engine — it doesn't retrieve facts; it generates them.
• Treating Claude as an oracle — confident tone does not equal correctness.
• Pasting huge documents without context — Claude dilutes attention on noise.
• Expecting Claude to know your team's conventions without telling it.
• Pasting secrets (API keys, passwords) into prompts — they leave your machine.
• Asking vague prompts and being disappointed by vague answers.

Best Practices

• Always read what Claude produces. Treat every response as a draft from a smart but unverified colleague.
• Keep official docs open in a browser tab. Cross-check unfamiliar APIs.
• Use Claude for the shape of content first; verify the substance after.
• Build the habit of asking "what could go wrong with this?" both to Claude and yourself.
• Set your account preferences thoughtfully — privacy matters.
• Use Projects to organize context for recurring work.

Troubleshooting

Practical Exercise

Frequently Asked Questions

Further Reading

• Claude product page
• Anthropic — About
• Claude documentation
• Claude plans

Official References

• About Claude
• Claude product page
• Claude documentation
• Anthropic blog

Related lessons: BE-02, BE-03, BE-04