It ain’t that bad, even though I get a new version of Cursor 1-2x a day lol.
It worked in that release but is not working again in today’s version (Version: 1.5.6).
1 Like
That’s been a bug for a long time. It’s 90% fixed for me with some changes to the powershell config file
1 Like
Found this in an edge case - fix coming today
2 Likes
idk that putting your email is allowed. But also people can just respond to you here. Not to mention people should respond to you here. If someone gives you information that works it would be helpful to others with your issue to see it as well
What model are you using? Have you tried smaller tasks to prove it can actually do those as expected and try to narrow down where it is getting derailed and confused. Your claim is that Cursor somehow benefits by helping people make projects they can’t ship, is unlikely. I think trying to give a very clear and narrow task and seeing if Cursor will just straight up refuse would be a good test. I am very doubtful that Cursor is preventing you intentionally from completing your project. They have very little to gain from that. They would at least have the AI finish your tasks but just do it for 10x as many tokens. To not complete the task at all makes Cursor useless.
Aaaaaaand we are back to the gear (1.5.7):
Free credits now. This looks like my grok-code-fast usage.
When the performance issues of Cursor app on Windows will be solved? For about two weeks app performance is poor. It’s lagging even when I type something on keyboard. Versions older than approx. 2 weeks didn’t have that problem. Now when you start the app it’s almost OK, but the longer you work, the less responsive and laggy it gets. At some point full restart is the only option. What has changed? I’m not seeing any breaktrought functionalities added to justify such poor performance. I’m on high-end PC with 128 GB RAM, NVME drive and 16 CPU cores.
Having coding app lagging so much feels like operating a typewriter, not a modern system
+1 on wanting a rollback. I’ve had nothing but issues since updating. Constant model failing to do simple code updates, or undoing its own changes then going “oops let me redo those”, or just hanging up and not being able to. on top of that the “ding” is driving me insane.
1 Like
Have you tried deleting all your old chats? I periodically delete everything older than two weeks, seems to fix all performance issues.
I got updated to 1.5.5, and now manually updated to 1.5.7
So far good experience on MacOS.
The context% display is super useful!
gpt-5-fast-high is crushing it.
The system notifications are a bit silly right now (constant dinging). Could get awesome once we could configure more (NOT after every file edit). Default should be:
- expect input
- all done (which is the same if you think about it)
2 Likes
I noticed that gpt5-high (and other models too) got much dumber than it was before. It doesn’t follow perfect structure. Sometimes does service inside of another service. Import libs in functions. Builds schemas inside routers. I believe it’s an issue of new prompts from cursor
Do you have sounds for notifications turned on in system preferences?
1 Like
I’m noticing with every update the agent with small task is getting more and more redundant (I’m using Auto) in responses and less consistent is this the aim for cursor vs previous use cases were somewhat intuitive?
The action selector became completely unpredictable in version 1.5.
I expect it to always have focus based on its appearance. But what happens instead: I insert a new line into the editor by pressing Enter.
Also, if you closed this selector by accident, you can’t re-run the command. Instead, you have to ask the agent to output the command again (which wastes tokens).
When I instruct it to “begin implementation,” instead of actually generating the file, it sometimes starts endlessly listing code samples and then proudly announces, “Implementation complete!”
Yes. Dings on every edited file, done, or asking sth. I just want a reliable done / ask thing.
The normal “finished” Ding does not work for me, half of the time.
Now I’m very sure. New cursor prompts got much worse. It does less tasks in one usage, it does unnecessary work, it makes wrong structure.
ALL of models work worse. idk what was in update.
1 Like
What version? Can you update to 1.5.7?
1 Like