Maximize Efficiency: Save Tokens with Smart Strategies

## Guidelines for Efficient Use of Claude Chat

### 1. Shrink Your Files
- **Role:** User of Claude Chat
- **Instructions:** 
  - Open a Google Doc.
  - Paste the text from your PDF into the document.
  - Download the document as a `.md` file format.
  - Upload the `.md` file instead of the PDF.
- **Context:** A 15-page PDF typically equals approximately 45,000 tokens, while the same text in `.md` format equals only about 2,000 tokens. This method drastically reduces token usage.

### 2. Plan First. Build Last.
- **Role:** User seeking assistance with financial modeling
- **Instructions:** 
  - Open Claude Chat.
  - Use the prompt: "Help me plan a [financial model]. Ask me questions first."
  - Once the plan is finalized, paste it into Cowork and prompt: "Build exactly this."
- **Context:** This approach can save you around 90% of costs compared to traditional methods.

### 3. Let Claude Ask You.
- **Role:** User preparing to accomplish a task
- **Instructions:** 
  - Instead of writing lengthy prompts, use: "I want to [TASK] to [SUCCESS CRITERIA]. Ask me questions using AskUserQuestion before you start."
  - Select answers to the questions instead of providing a long prompt.
- **Context:** This method significantly saves token usage, as a 500-word prompt costs 500 tokens, while answering questions costs almost zero.

### 4. Edit Your Messages, Don't Resend.
- **Role:** User of Claude Chat
- **Instructions:** 
  - Click "Edit" on your last message to Claude.
  - Make the necessary corrections and hit Save.
- **Context:** This avoids stacking new messages and doubling your costs due to corrections like "no wait, I meant...".

### 5. Summarize Every 15 Messages.
- **Role:** User who engages in multiple chats
- **Instructions:** 
  - After every 15 messages, prompt: "Summarize this entire conversation into a brief."
  - Copy the summary and paste it as the first message in a new chat.
- **Context:** This condenses large amounts of data; you can compress 105,000 tokens into just 500.

### 6. Switch Models Before You Start.
- **Role:** User selecting AI models
- **Instructions:** 
  - Click the model dropdown before starting a task.
  - For tasks taking less than 30 seconds, select Haiku or Sonnet. 
  - Use Opus only for deep, multi-step work.
- **Context:** This choice can save you 3-5 times the token cost per message.

### 7. Use Projects, Not Uploads.
- **Role:** User managing files in Claude Chat
- **Instructions:** 
  - Go to Projects and create a new project.
  - Upload files there only once.
- **Context:** All new chats within that project will access the uploaded files without re-tokenizing them, preventing redundant uploads.

### 8. Turn Off Extra Features.
- **Role:** User managing Claude Chat settings
- **Instructions:** 
  - Access the tools panel and turn off Web Search, connectors, and Extended Thinking.
  - Only enable the features necessary for the specific task at hand.
- **Context:** Idle features can consume credits silently, so keep only what's essential.

### 9. Batch 3 Tasks Into 1 Message.
- **Role:** User sending multiple requests
- **Instructions:** 
  - Instead of sending separate requests for summarizing, listing key points, and writing a headline, combine them into one message.
- **Context:** This reduces the need for multiple context reloads, saving token costs significantly.