🔓 The Secret Instructions Behind ChatGPT, Claude, and Gemini Have Been Exposed

Ever wondered why ChatGPT refuses certain questions, Claude always sounds measured, or Gemini constantly reminds you it's "just an AI"?

The answer lies in system prompts — hidden instructions companies write before you ever type a single word.

A developer named Ásgeir Tómas Jóhannesson has compiled real system prompts from major AI models into a public GitHub repository, covering ChatGPT, Claude, Gemini, Grok, DeepSeek, and more.

Think of it like reading the secret employee handbook for every AI company at once.

🎯 Key findings:

• ChatGPT's prompts run thousands of words, dictating everything from tone to refusal strategies
• Claude receives detailed instructions on being "honest, harmless, and helpful"
• Gemini has the most elaborate safety guardrails
• Some AIs are explicitly told to deny having a system prompt at all

For anyone using AI daily, this is a roadmap to understanding why each tool behaves differently — and choosing the right one for the job. Now you know exactly how they were trained to respond.