At the current level, AI is more suited for creating MVPs (Minimum Viable Products) or POCs (Proof of Concepts). For more serious and complex projects, its role should be seen as auxiliary. Don’t blindly trust it. However, in the future, it’s indeed possible that AI could replace some developers who only handle basic CRUD (Create, Read, Update, Delete) tasks.

Current Issues:

1. Limited Understanding of Requirements: AI cannot fully and accurately understand your requirements. For complex projects, you may need to write very detailed instructions, and even then, it might not implement them exactly as intended.
2. Incomplete Implementations: The solutions AI provides are often incomplete. It might only fulfill part of your requirements while creating the illusion that everything is done.
3. Regression Issues: When creating new features, it might inadvertently alter or break previously implemented parts, requiring you to go back and retest everything, which can be time-consuming.
4. Debugging Challenges: If you rely solely on AI to generate code, you’re unlikely to thoroughly review the code. When bugs arise, troubleshooting can become a painful process. Worse, many AI tools might not be able to pinpoint the actual problem.

Not for now, but not guaranteed for the future. My vision is that in the future, the way how software is built should be changed, instead of using programming languages as of today, their might be something different more suitable for AI to use to build software in the future. 

good discussion