Let Claude think comprehensively before responding!
A super quick reminder: Thinking claude is not aimed for benchmarks or huge leaps in math or something, since those are pre-determined by the base model (new Claude-3.5 Sonnet). I only want to explore how further we could reach with Claude's "deep mindset". That said, when using it in your daily tasks, you will find Claude's inner monolog (thinking process) very very fun and interesting.
This project consists of two main components:
thinking-claude/ ├── extension/ │ ├── .vscode/ │ ├── chrome/ │ ├── firefox/ │ └── changelog.md ├── model_instructions/ │ ├── changelog.md │ ├── v3.5-20241113.md │ ├── v4-20241118.md │ └── v4-lite-20241118.md ├── LICENSE └── README.md The project is organized into two main directories:
extension/
: Contains browser extension implementationsmodel_instructions/
: Contains thinking protocols for different versionsEach directory maintains its own changelog for version tracking.
The thinking protocol instructs Claude to follow a natural, thorough thought process before providing responses.
The browser extension enhances the Claude interface by making the thinking process more manageable:
Quick Install: Install directly from the Chrome Web Store
Manual Installation:
git clone https://github.com/yourusername/thinking-claude.git
chrome://extensions/
extension/chrome
folderabout:debugging#/runtime/this-firefox
extension/firefox/manifest.json
filemodel_instructions
folderOnce installed, the extension automatically:
Contributions are welcome! Feel free to:
MIT License - feel free to use and modify as needed.
Special thanks to Claude! We build this whole extension together!