Meet Claude
What Claude is, how it works, and what makes it different from other AI assistants. Your first step to becoming a Claude power user.
The assistant that reads the whole document
Picture this: you paste a 50-page contract into an AI assistant and ask "What are the three riskiest clauses for my company?" Most AI tools choke, hallucinate, or give you a generic answer. Claude reads every page, identifies the specific clauses, explains why they're risky, and suggests alternative language — all in under a minute.
That's Claude. Built by Anthropic, it's an AI assistant designed to be helpful, harmless, and honest. But what does that actually mean in practice, and why should you care?
What Claude actually is
Claude is a large language model (LLM) created by Anthropic. Like all LLMs, it predicts the most likely next token (word or word-piece) in a sequence. But Anthropic trained Claude with a specific focus on being safe, steerable, and genuinely useful — not just impressive.
Here's what sets Claude apart:
| Strength | What it means for you |
|---|---|
| Long context window | Paste entire documents, codebases, or datasets — Claude processes up to 200K tokens at once |
| Careful reasoning | Claude shows its work and is less likely to make confident-sounding mistakes |
| Instruction following | Give detailed instructions and Claude sticks to them consistently |
| Honesty about limits | Claude tells you when it's uncertain rather than making something up |
| Safety by design | Built with Constitutional AI — trained to be helpful AND avoid harm |
✗ Without AI
- ✗Confidently makes up facts
- ✗Ignores nuance in your question
- ✗Generic one-size-fits-all answers
- ✗Struggles with long documents
✓ With AI
- ✓Says 'I'm not sure' when uncertain
- ✓Asks clarifying questions when needed
- ✓Follows your specific instructions precisely
- ✓Handles 200K tokens — entire books
There Are No Dumb Questions
Who is Anthropic?
Anthropic is an AI safety company founded in 2021 by Dario Amodei, Daniela Amodei, and other former members of OpenAI. Their mission is to build AI systems that are safe, beneficial, and understandable. Claude is their flagship product.
Is Claude free?
Claude has a free tier at claude.ai with usage limits. Claude Pro ($20/month) gives you more usage and access to the latest models. Claude Team and Enterprise plans are available for organizations.
How is Claude different from ChatGPT?
Both are powerful LLMs, but they differ in training approach and philosophy. Claude uses Constitutional AI (a set of principles it was trained to follow), tends to be more careful about accuracy, and excels at long-document analysis and nuanced instruction following. The best advice: try both and see which fits your workflow.
The Claude model family
Claude isn't just one model — it's a family of models optimized for different needs:
Claude Opus — The most capable model. Best for complex analysis, nuanced writing, and tasks that require deep reasoning. Slower but more thorough.
Claude Sonnet — The balanced model. Fast, capable, and cost-effective. Great for everyday tasks, coding, and most professional work.
Claude Haiku — The speed model. Ultra-fast responses for simple tasks, classification, and high-volume use cases. Ideal when you need quick answers.
Think of it like hiring: Opus is the senior consultant you bring in for the hardest problems. Sonnet is the reliable full-time employee who handles everything well. Haiku is the intern who's incredibly fast at straightforward tasks.
Where you can use Claude
Claude lives in multiple places — pick the one that fits your workflow:
- claude.ai — The web interface. Sign up and start chatting. Best for general use, document analysis, and writing tasks.
- Claude Code — A command-line tool for developers. Write, debug, and refactor code directly in your terminal. We'll cover this in depth in Module 3.
- Claude API — For developers building AI into their own products. Pay per token, integrate anywhere.
- Mobile apps — Claude on your phone for quick tasks on the go.
- Team & Enterprise — Shared workspaces with admin controls, higher limits, and data privacy guarantees.
What Claude can (and can't) do
Let's be honest about capabilities:
| Claude excels at | Claude struggles with |
|---|---|
| Writing, editing, and summarizing any text | Real-time information (no live internet access) |
| Analyzing documents, data, and code | Precise mathematical calculations |
| Explaining complex concepts simply | Remembering previous conversations |
| Following detailed multi-step instructions | Tasks requiring visual understanding of images with tiny text |
| Coding in dozens of programming languages | Predicting the future or giving financial advice |
| Translating between languages | Accessing your files, email, or calendar directly |
Find your Claude use case
25 XPConstitutional AI: why Claude behaves the way it does
Most AI models are trained purely on "what sounds right." Claude is trained on something more: a constitution. This is a set of principles — like "be helpful but avoid causing harm" and "be honest about uncertainty" — that guide Claude's behavior.
Massive text dataset from books, websites, and code
Neural network learns language patterns and knowledge
Model is trained to follow principles — be helpful, harmless, and honest
The final assistant you interact with, shaped by its constitution
This means:
- Claude won't help with genuinely harmful requests — but it will help with legitimate edge cases like security research
- Claude admits uncertainty rather than guessing
- Claude follows your instructions but pushes back when something seems off
- Claude tries to be maximally helpful within ethical boundaries
There Are No Dumb Questions
Does Constitutional AI make Claude less useful?
No — it makes Claude more useful. An AI that confidently makes things up or helps with harmful requests isn't actually helpful. Claude's guardrails mean you can trust its output more, not less.
Can I override Claude's safety guidelines?
You can't bypass core safety training, but Claude is highly steerable. You can adjust tone, style, detail level, and approach. For professional contexts like security testing or medical research, Claude understands nuance and will engage appropriately when you explain the legitimate context.
Your first Claude conversation
Head to claude.ai and try this right now:
Step 1: Sign up for a free account (email or Google sign-in)
Step 2: Start a new conversation
Step 3: Try this prompt: "I'm a [your job title] at a [your company type]. What are the three most impactful ways I could use you in my daily work? Be specific to my role."
Step 4: Follow up: "Let's try the first one right now. Walk me through it step by step."
Have your first real conversation
50 XPKnowledge Check
1.What training methodology makes Claude different from most other AI models?
2.Which Claude model should you use for most everyday professional tasks?
3.What is Claude's context window size?
4.When Claude is uncertain about a fact, what does it typically do?