48 Laws of Claude
1
TL;DR
- Utilize Claude's new models (Opus 5.5, Sonnet 5.5, Haiku 5.5) based on task complexity, with Haiku for cost-effective repetitive tasks and Opus for demanding problems.
- Integrate multiple Claude models to create efficient workflows, assigning tasks like sorting, responding, and complex problem-solving to the appropriate model.
- Configure Claude's reasoning effort and use Projects for organized workspaces, setting specific instructions and training the AI on your writing style.
- Provide examples instead of lengthy explanations, ask Claude to question your opinions, and always supply original information for tasks.
- Instruct Claude to ask clarifying questions, request finished results rather than answers, and define what a good outcome looks like.
- Leverage Claude Code for programming tasks, using features like /effort, Plan Mode, and subagents for efficient development and error checking.
- Connect Claude to your existing apps (Gmail, Google Drive, Slack) for seamless data integration and automated workflows.
- Utilize scheduled tasks and plugins to automate recurring jobs and extend Claude's functionality.
- Implement prompt caching and the Batch API for cost efficiency with the Claude API.
- Maintain control by setting boundaries for Claude, requiring approval for significant actions like sending emails, deleting data, or spending money.