Peeking Under the Hood of Your Favorite AI Tools

Ever wonder what exactly you're talking to when you use ChatGPT, Claude, or Gemini? We often interact with these AI tools through a polished interface, but the real magic—and the specific instructions guiding their behavior—happens out of sight. A developer recently decided to pull back the curtain.

This GitHub repository acts as a collective scrapbook for the hidden system prompts and model architectures behind popular AI applications. It's a fascinating look at the foundational "constitution" given to these models before any user conversation begins.

What It Does

The repository, System Prompts and Models of AI Tools, is a community-sourced collection. It compiles the known system prompts and technical model details for tools like OpenAI's ChatGPT, Anthropic's Claude, Google's Gemini, and others. Think of it as a crowdsourced wiki for reverse-engineering the initial setup of these AI assistants.

Why It's Cool

For developers and prompt engineers, this is more than just trivia. Knowing the base system prompt is incredibly useful. It helps you understand the default boundaries, capabilities, and tone the AI has been given. This knowledge lets you craft your own prompts more effectively, working with or around the tool's foundational instructions.

The clever part is its simplicity and community-driven approach. It doesn't use a complex scraper or API; it relies on discoveries shared by users—often found through prompt leaks or creative social engineering—and documents them in a clear, structured format. It turns opaque system behavior into something you can read and analyze.

How to Try It

You don't "run" this project in a traditional sense. Instead, you explore it.

1. Head over to the GitHub repository: github.com/x1xhlol/system-prompts-and-models-of-ai-tools
2. Browse the README.md for a high-level overview.
3. Dive into the markdown files for specific tools (e.g., chatgpt.md, claude.md) to see the compiled information on their suspected system prompts and models.

You can also contribute if you discover a new prompt or have a correction—just follow the repo's guidelines for submitting a pull request.

Final Thoughts

As a developer, I find this repository genuinely helpful. It demystifies the "black box" feeling a bit and provides concrete data for building better interactions with these AI systems. Whether you're looking to optimize your prompts, satisfy your technical curiosity, or just understand the hidden rules of the game, this is a great resource to bookmark. It's a reminder that even the most advanced AI starts with a simple, human-written set of instructions.

Follow us for more projects: @githubprojects