This repository provides custom instructions for ChatGPT and Operator, designed to enhance their performance and response quality. It targets users seeking more structured, logical, and detailed outputs from AI assistants, offering a significant improvement in task success rates as validated by MMLU benchmark testing.

How It Works

The instructions enforce strict adherence to logical reasoning, penalize incorrect answers, and demand context preservation. A key feature is the requirement for the AI to adopt realistic expert roles and provide a TL;DR summary before detailed explanations, improving clarity and user comprehension. For coding tasks, it explicitly forbids placeholders, ensuring complete code snippets.

Quick Start & Requirements

• Apply: Copy instructions into ChatGPT's "Custom Instructions" under Settings > Personalization, or into Operator's "Custom Instructions" in General settings.
• Requirements: Compatible with ChatGPT and Operator. Tested on GPT-4o.

Highlighted Details

• Improved task success rate by ~7% on MMLU benchmark.
• Enforces TL;DR summaries and realistic expert role adoption.
• Strict rules against placeholders in code and penalties for errors.
• Includes specific instructions for Operator to use Bing for problem-solving and detailed developer-like actions.

Maintenance & Community

The project is maintained by DenisSergeevitch. No community links or roadmap are provided.

Licensing & Compatibility

Feel free to use and modify these instructions for your own use.

Limitations & Caveats

The instructions are designed for specific AI platforms (ChatGPT, Operator) and may not be directly transferable or effective on other models or interfaces. The performance claims are based on specific benchmark testing and may vary in real-world application.