Cursor rocks.
Shouldn’t we asap develop a radically new AI coding best practices that fully leverages AI capabilities? If AI handles 99% of coding, code needn’t follow human-centric limitations. Next: What about training LLMs on synthetically generated AI-focused coding best practices?
A few ideas:
Monolithic Code Structure
- Even having full codebase context the use of larger, more comprehensive files instead of breaking code into many small files might lead to more cohesive code.
AI-Optimized Naming Conventions
- This would provide more context for the AI
Dynamic Code Restructuring
- Allow the AI to restructure code dynamically based on the current task.
Test-Centric Development with Feedback Loop
Implement a feedback loop where:
- Human defines test cases and acceptance criteria.
- AI generates code to pass the tests.
- If tests fail, AI analyzes failure, adjusts code, and retries.
- Process repeats until all tests pass or maximum iteration limit is reached.
AI-Friendly Version Control
…