o4 mini displays BIZARRE distance between my question and its reflection. That being said, through its careful iteration and implementation, it does very well. I wonder if this is a clear indicator of Cursor prompting and agentic handling shining very strongly (I say this as someone who has brutally criticized this company), because o4 DOES NOT start very close. It’s a fascinating thing to see because it is doing very well at the “needle in a haystack” changes you need. I use “Ask” to evaluate all possibilities, then select a few more likely, and have the Agent sort of review it and edit afterwards. It seems to be a strong workflow once your project reaches critical mass. If you slow it down and ask it questions, or to find you “why this is happening when xyz CSS from shadcn is conflicting with a third party package of some sort etc”, and break it into two questions, even though the agentic handling iterates, it still helps. Thanks Cursor for improvements. I’d still really like to see some more alert functionality in the IDE regarding usage based costs, like a meter etc of your “allocated spending”, but overall, I’ve been having much more success with Cursor lately, and I felt necessary to return here and say good job.