Learning Timeline
Key Insights
The Benefits of Filtering LLM Patterns
LLMs like GPT often have a 'signature style' that sounds very robotic. By including sentence patterns like 'It's not just X, it's Y' in your .cursorrules file, you force the AI to think creatively and produce more natural, human-like output.
Tips for Rule Flexibility
Even though you are setting a list of 'banned words', remember that rules can be broken depending on the context. Use this list as a 'quick pass' to filter writing quality rather than an absolute restriction.
Prompts
Sample .cursorrules Writing Mega Prompt
Target:
Cursor AI
You are an expert technical writer. When writing content, follow these rules:
1. Maintain a specific voice and tone: [Insert your style].
2. Be specific and avoid fluff.
3. Strictly avoid banned words: performant, seamless, leverage, tap into, delve.
4. Avoid common LLM patterns like 'It's not just X, it's Y' or starting sentences with 'In the rapidly evolving world of...'.
5. If you see these patterns, rewrite the sentence to sound more human and direct.
Step by Step
How to Set Up a .cursorrules File to Control AI Writing Style
- Open the Cursor app and access your project folder.
- Click the 'New File' icon in the left sidebar (Explorer).
- Name the file `.cursorrules` (make sure to include the leading dot) and place it in your project's root directory.
- Open the `.cursorrules` file to start adding your instructions (mega prompt).
- Define your preferred 'Voice and Tone' standards so the AI understands your writing identity.
- Create a specific section titled 'Banned Words' or 'Words to Avoid'.
- List buzzwords or jargon frequently repeated by LLMs, such as 'seamless', 'performant', or 'comprehensive'.
- Add instructions to detect 'LLM Patterns' or cliché robotic structures, such as the 'It's not just X, it's Y' sentence pattern.
- Save the file by pressing 'Ctrl + S' (Windows) or 'Cmd + S' (Mac).
- Use Cursor Chat or Composer as usual; the AI will now automatically refer to the .cursorrules file before generating responses.
