Based on the last news of Developers that are YouTubers, we think about what they describe and therefor we can try to understand how and why. Is easy to code 20.000 lines per day using AI, and naturally, is easy to review 20.0000 lines of code using AI. But the main question that comes into my mind is: Does AI create a dedicated environment to test the code it is reviewing? Because, you need to test the code and testing is how you know that the code works, even if you can review the code to understand how the "problem" was solved and to see if there is some extra lines that do not need to exist in that particular piece of code.
So I ask: Does AI create a dedicated environment to test the code which is reviewing?
In fact I would like to know that, since at the end, it is all that matters. But to understand if a function should not be in one piece of code, you need to understand all other pieces and the hole application structure, while in my interpretation, AI can review code samples without including them within all other pieces of code that the final application have. Since I don't know, I ask.
The complexity of building an application is Big, because is not only the functionality but the "Pointer" that makes functionality, being well applyed into a specific logic structure that depends on many other conditions that are not the Application itself, but Users, Bots, Exploits, etc... And with that, we have systems like OpenBSD that have those kind of "Philosophies" that say to developers: "Find the Bug before the Hacker find it". With that, we understand that most of code bugs are in fact the result of developers who did not spend enought Time to think about what they were doing.
Fast and Good, Few Can, Since Most Will fail and will produce many Mistakes
You can listen this talk about how useful AI can be, if you "Guide AI" to do things correctly.