Notice the loophole: there’s no qualification of how much problem context the AI started from. Most of the problem -> code “work” would still be done by a human in that situation — even if technically 50% of the code is “AI generated” [because the human did all the hard work of generating the context necessary for those tokens, including the preceding tokens of code].
As the saying goes… lies, damned lies, and statistics.
But yeah this could probably be maintained by like five people.
A lot of my problems show up on lists like “25 falsehoods programmers believe about addresses”. These were things that were maybe acceptable in 1999, because they didn’t know better, but only having a single street line is a problem.
I could rant, but half of it would be sales directors who have never used Salesforce shooting us in the foot.
I haven't used an implementation of SFDC that I've liked yet.
Which, surprisingly, is a really necessary product and quite useful.