context window

The context window is the AI's working memory. It's the total number of tokens the model can process in a single conversation — everything you've said, everything it's said back, and everything in the system prompt. When you exceed it, the AI starts 'forgetting' earlier parts of the conversation.

Claude offers 200K tokens (~150K words). GPT-4 offers 128K tokens. These sound huge, but they fill up fast when you're pasting codebases, documents, or long conversation histories. Understanding context windows explains why AI conversations 'drift' after a while — the model literally can't see what you discussed earlier.

Practical context window management is a core AI operator skill: knowing when to summarize, when to start a new conversation, and how to structure prompts to use context efficiently.

When to Use It

When planning prompts, debugging conversation drift, or deciding how much context to include in an AI interaction.

Try This Prompt

Why It Matters

Context window management is the difference between AI that remembers and AI that forgets. Master it and your AI interactions become dramatically more effective.

Memory Trick

It's literally a window — you can only see what fits in the frame. Scroll too far and the beginning disappears.

Example Prompts

Common Misuses

• ×Thinking a bigger context window means better AI — quality often degrades with very long contexts
• ×Not realizing the response counts against the context window too
• ×Assuming the AI 'remembers' previous conversations — each conversation starts fresh

Related Power Words