Again it’s more the way you use it, if you tell the agent to go out and create a large program by itself. That’s an issue. You’re just creating a ton of code without agency, leaving design decisions up to the AI is pretty terrible.
But if you know how to code, and you can guide it to make small metered changes based on your knowledge of coding and do it much faster than normal.
The main issue comes when junior developers aren’t getting the proper training on how to solve problems or coding issues. Or even how to properly architect things. Or for open source projects, not having the knowledge of the code base to be able to make changes or fixes to issues because of the spaghetti code they let the agent create.
Trust me when I say, every single developer at my very large company uses AI to code, it’s very good and very useful. But it’s just how you use it that determines its usefulness and longevity.
If you’re saying you disagree with me that AI is really good at making code, it’s obvious that you don’t code or that you’re using the tools incorrectly.