90%
12 hours ago
- #AI coding
- #software engineering
- #future of development
- AI is currently writing 90% of the code for the author's infrastructure project, with predictions that AI will write nearly all code within 12 months.
- The author emphasizes the importance of maintaining high standards, reviewing every line of AI-generated code, and being responsible for the final product.
- Key practices include using raw SQL instead of ORMs, OpenAPI-first development, and leveraging AI tools like Claude and Codex for debugging and code review.
- AI-generated code can be problematic if not carefully managed, leading to issues like poor system architecture, unnecessary dependencies, and hidden errors.
- The author highlights benefits such as faster research and implementation, easier refactoring, and improved infrastructure setup with AI assistance.
- Despite AI's role, strong engineering skills are essential to avoid brittle systems, security holes, and scalability issues.