GPT5 performance

It’s nothing new or special in my experience so far in developing my heavy full stack application and I’m really disappointed. I am forced to use Cursor on Windows and so I spend my time fighting with it as it doesn’t understand how to navigate a shell environment even though I’ve added many iterations of rules as well as context stuffing it. Syntax errors everywhere and on top of it SSH is brutal. I was hoping GPT5 would help but it seems worse for me in many ways. Is this common? I hope it is just my n00bness. I was hoping to run Cursor CLI inside Cursor IDE but it doesn’t work on Windows and so I’m thinking I may need to bring in Claude Code for now. When the prices go back to full for GPT I don’t plan on using it at all and instead using one of the cheaper openrouter variants. Those benchmarks OpenAI bragged of seem COOKED :wink: