Skip to main content

Module 1.1: What Claude Is and How It Thinks

Understanding Your AI Collaborator

Before you can effectively cowork with Claude, you need to understand what it is and — just as importantly — what it is not. This lesson gives you the mental model that separates productive Claude users from those who get frustrated and give up.

What Claude Actually Is

Claude is a large language model (LLM) built by Anthropic. It processes text (and images) as input and generates text as output. But thinking of it as a "text generator" misses the point. Claude is better understood as a reasoning engine that:

  • Understands context: It reads everything you provide — instructions, documents, code, conversation history — and uses it all to inform its responses
  • Follows instructions: It can follow complex, multi-step instructions with nuance and judgment
  • Reasons through problems: It can analyze, compare, synthesize, and draw conclusions
  • Generates structured output: It can produce code, JSON, HTML, markdown, or any text format
  • Adapts to feedback: It adjusts its approach based on your corrections within a conversation

The Context Window

The context window is the most important concept for working with Claude. It is the total amount of text Claude can "see" at once — your messages, its responses, any documents you share, and system instructions.

Think of it like a desk. A larger desk lets you spread out more documents. Claude's desk is large (up to 200K tokens for most models, 1M for some), but it is not infinite. When the conversation gets too long, older context gets compressed or dropped.

Practical implications:

  • Front-load important context: Put your most important instructions and information early in the conversation
  • Be specific: Instead of vague references ("that thing we discussed"), restate key points
  • Use new conversations for new topics: A fresh conversation gives Claude a clean desk

What Claude Is Not

Understanding limitations prevents frustration:

  • Not a search engine: Claude does not browse the internet in real time (unless given tools to do so). Its knowledge has a training cutoff.
  • Not infallible: Claude can be confidently wrong. Always verify critical facts, especially numbers, dates, and technical specifications.
  • Not persistent: Each conversation starts fresh. Claude does not remember previous conversations unless you give it context.
  • Not a human: It does not have feelings, preferences, or experiences. Anthropomorphizing it leads to poor collaboration patterns.

The Right Mental Model

The most productive way to think about Claude is as a brilliant, knowledgeable colleague who:

  • Just joined your team today (no prior context)
  • Can read and process information extremely fast
  • Has broad knowledge but may not know your specific codebase, company, or processes
  • Works best with clear instructions and specific context
  • Will ask for clarification if you let it, but might guess if you do not

Key Takeaways

  • Claude is a reasoning engine, not just a text generator.
  • The context window is your shared workspace — manage it deliberately.
  • Claude is not a search engine, not infallible, and not persistent across conversations.
  • Think of Claude as a brilliant new team member who needs clear context to do their best work.

In the next lesson, we will cover the art of prompting — how to communicate with Claude for maximum effectiveness.