The main 'fighting' that I've experienced is it being "confidently" incorrect about something I'm very knowledgeable about, so I have to constantly check what it's doing anyway. I've also had to keep it from creating unnecessary functions and variables, from leaving empty comments (as in, empty of substance or actual context, leaving comments just to leave comments). It also constantly makes up problems to "solve". It seems to have only gotten worse and worse with each subsequent model release. It wastes so many tokens doing things that don't matter or are otherwise incorrect.

I honestly feel less productive in my job with it for 80-90% of the time, but the company makes me use the slop bot. It works for the basics, but any time I add complexity, I begin having to fight against the machine. I could see how I could feel more productive, it can iterate through code very fast. But most of it has something wrong, is bloated, or is just completely irrelevant. I ultimately just end up feeling like I wasted an entire day to get a subpar and soulless result.

Idk, I don't use it for work but in the limited interactions I've had with ai it's required arguing with to get it to actually give me correct results. Most of my experiences with it are when it's shoved into places it doesn't belong. Customer support, search engines, contact forms etc