I know this post is more about the committing on LLM "fixes", but find the other reasons more interesting.
Similar to the date & time library there are a couple of other things that look easy at a first glance, but get complicated very quickly, because it has so many special cases:
lexicographic sorting (different languages sort things differently)
Postal address formatting (different standards in different countries, with many different context sensitive rules)
How can you not? Like there are specific tells as the other user pointed out, but aside from that so much AI slop just has this absurd uncanny valley quality that Im actually kinda blown away that other people don't immediately see it.
I too, tried using regex to strip html tags. This was in php with hardly any knowledge of oop.
A university education really improved me there, since they trained you in Java to a high degree within 20 weeks (alongside stuff like UML). PHP is a fine starter language, but as applications grow, a more native oop approach with build-time validation, really made me a better developer.
Tried to do a proper analysis of a bug in homologation that's preventing our future customer to test the product, while N+3 and N+4 demand twice-daily updates through video call with ten other managers on it. Two weeks of proper madness, that was.
Stopped the meetings for one fucking day and we had it fixed before 5pm.
Tried to understand what options were available on a JS package, which mangled the options before passing them on to another JS package, which mangled the options before passing them onto another JS package, ...
At my last job, there was no planning of work/projects. Like, there was a general plan of "We need feature X by Q3 and here's what it should do", but nothing about breaking work down into smaller units or prioritizing different tasks.
The manager would drop an email: "Hey, can you do ...." and that was it. Now it's another thing to throw down the waterfall. Big surprise, the same bastard would harp about how the project was underperforming!
I had to parse an html table with a column of printed python dictionaries, the dicts wrapped a mixed delimiter list of SQL statements of various types.
It’s truly an abomination.
I’ve never had to use pythons AST module to parse data before…
Hum, i made a minimal RSS and Atom reader shell script based on sed and awk regex. I'm mostly fine, you just need to pick the relevant bits and ignore the rest, remove all newlines beforehand, about 10 regex strings in a case switch. And feed it to iconv, f*king • things. But html is, where the nightmare starts.