Context limits are real. Learn to work within them or expand them creatively.

(Spoiler: "Just paste the entire codebase" doesn't work. Context limits are the AI equivalent of "too many tabs open". We've all been there. 🗂️)

Context Window Sizes

• GPT-4 Turbo: 128K tokens (~96K words)
• Claude 3.5: 200K tokens
• Gemini Pro: 1M tokens (largest!)
• GitHub Copilot: ~8K tokens (IDE context)

Strategies When Context is Too Small

• 1. Chunking: Break large files into sections, process separately
• 2. Summarization: Summarize early conversation, keep summary as context
• 3. RAG (Retrieval): Use vector DB to fetch only relevant parts
• 4. Switch Models: Use Claude/Gemini for large context needs

Best Practices

• ✓ Include only relevant files in prompts
• ✓ Use @-mentions (Cursor) to control context explicitly
• ✓ Summarize previous conversation when approaching limit
• ✓ For multi-file edits: focus AI on one file at a time