I am honestly wondering if there is somehow a bug in my cursor account? I seem to have a plethora of problems that many people do not. Even other team members at my company. I have had this long, progressive degradation in Cursor’s functionality for some time now. Over two months. I think it was around 1.3.x, early, that I remember things last working well (except for the terminal.) Since 1.3, there has been a continuous and progressive loss of functionality. I can’t use most of the built in @ commands anymore, can’t use custom / commands, model behavior seem to degrade after the first week or so of using them, and most models seem to have NONE of the capabilities that Claude has generally had (not sure if that is expected, but I can’t seem to attach images to context for any model except Claude, can’t do web searches with any model but Claude, and Claude doesn’t seem to do that well anymore.) I also can’t use @Web links, the model instead tells me that it cannot read web pages, however I thought the EXPRESS PURPOSE of @Web ws to have the agent/model read that specific web page.
The past week in particular, it also appears as though my rather extensive set of rules, is basically not followed at all, by any model. Things that used to be extremely reliable, such as httpie level cli testing of my apis, which run on a specific port, were suddenly being tested at port 3000, which is not what I use. I have rules that explain this, and in the past it always worked fine. Lately, though, the agent always uses port 3000. Then of course, fails, then reports that the server isn’t running. I insist it is, and it then fumbles around trying to figure out why it can’t connect. It simply doesn’t seem to understand its used the wrong port…its not even one ofthe things it tests, it will flail about aimlessly checking everything EXCEPT the port! Darn thing won’t even look at my local environment variables to learn the proper port. Which it used to do by default, I didn’t even need a rule originally.
What causes this level of degradation? I am honestly quite baffled… Cursor was AMAZING a few months ago. What in the world happened? Why does the whole thing seem to struggle so much now?
At this point, all of the functionality that I chose Cursor for, seems defunct, and yet, some of my other teammates are able to use most of this functionality (@Docs doesn’t seem to work for most of them, and I think that most of the @ command issues were a known issue?) Given that, I am wondering if there is some kind of cursor account level issue here? I can’t otherwise account for why I myself, have so many issues with Cursor, while team mates seem to be having very few if any problems.
Anyway, I thought I’d mention it, as, my problems are so extensive and so systemic now, the benefits that Cursor provided early on for me, are almost gone. Basic agent functionality still works, and the model can still code, although it seems like many of the models (most of the ones I use regularly: Claude, Grok, GPT-5) have lost most of their capability, smarts, efficacy? The model still being able to code does help accelerate things a bit, but using Cursor has become a huge chore now, due to all of the broken functionality.
I am pretty sure I am not the only person who experiencing these issues. I wonder if there are account-level problems that are at the root of these progressively degrading issues? I don’t know how much of Cursor’s functionality is remote, vs. local, but it does seem like there is a fair amount of remote functionality (beyond just the communication with the model itself, at least.)
Well, I hope you guys can track down why these issues plague me, and others. It has had such a devastating impact to your product and its usability, that I am now very seriously considering alternatives, as Cursor has become a very frustrating experience, instead of the amazing experience it started out as months ago.
