It is worse than just glitchy when it comes to A.I. though. It is right now in the area of outright not working. They may use the A.I. to put the scaffolding of the code up, but really they need programmers to go through the code fix it, so that it even works. I've seen it fail on simple scripts too many times to believe it will work any differently when it comes to far more advance complex code.It's like gaming now companies don't care and will release a glitchy half finished copy knowing they can just send updates or dlc later
Now, if it could actually learn from its mistakes, then that is when I will agree to the idea that programming is no longer needed. Once that happens though, it opens up another can of worms of A.I. trying to be so efficient in what it was programmed to do, that it starts building words of "paperclips". So even then, some oversight will need to be maintained.