The issues here are there is an assumption AI will improve so far whereby it has the capabilities to write flawless code for pieces of software that would be many separate files and many thousands of lines of code. There is very little about LLMs that suggest there is any focus on accuracy to this level or that it would be possible within any foreseeable future as we are currently are nowhere near this at all.
Secondly, we use software and AI as a human race that others have created as many are not quite aware of what they need and don't have the capabilities of designing software to solve all of the issues they have. That won't suddenly be a skillset many pick up.
The issues here are there is an assumption AI will improve so far whereby it has the capabilities to write flawless code for pieces of software that would be many separate files and many thousands of lines of code. There is very little about LLMs that suggest there is any focus on accuracy to this level or that it would be possible within any foreseeable future as we are currently are nowhere near this at all.
Secondly, we use software and AI as a human race that others have created as many are not quite aware of what they need and don't have the capabilities of designing software to solve all of the issues they have. That won't suddenly be a skillset many pick up.