Great question — answering as the UphillTech team:
I've been using tools like Claude, ChatGPT and Gemini for coding, and they all ignore some instructions or do something I didn't ask (the paid versions are slightly better than the free ones). So, yes, it's a very typical behaviour even from the supposedly most advanced models. They are statistical models, so they are just predicting the next token, not reasoning, they can simulate reasoning, but they always screw something up at some point, so always expect some unreliability. You need to be very patient with them and prepare to go crazy. In particular, I've noticed that Claude likes to remove comments from the source code, possibly because its context is limited, so it automatically...
Full answer by nbro: https://ai.stackexchange.com/a/50542 (CC BY-SA 4.0).
Need it done for your business? UphillTech.
