You don’t belong on this forum. You have nothing of any value to say regarding the actual substance of the original post. You don’t know me so perhaps you should be be more respectful, or just be quiet. I absolutely don’t need your psychoanalysis. That’s not the meaning of literally btw.
I made a number of very valid points in my original posts. They are fully supported by the facts of how this software performs. I am not surpised that juvenile personal attacks are the result instead of meaningful inputs on how to fix some of the serious flaws in the product.
yes, and here we are, I suppose you still pay your subscription ?
as long as people will complain and pay USD200 at the same time, you know, people.
All the best for the next 25 years of experience I guess
There are restore checkpoints with every request, if you were not aware.
1 Like
lol, exactly. Complain about how awful Cursor is, but meanwhile continue to hang around and keep using Cursor after it’s been proven “broken”. That’s how good Cursor is. People want to hate it for some reason, but can’t part ways.
I will reiterate my offer to help @user477 with prompting, setup or other parts that may be not working. We are here to help.
While Cursor is not absolutely broken and we do work daily on improvements or bug fixes outside of new feature development, there still may be issues we have to fix. Perhaps some are also based on AI model behavior which is to a large extent out of our control.
As we have improved stability of Cursor a lot and reduced common issues, some specific issues may need further details so we can determine cause.
I encourage all to post a full separate Bug Report with more info Create Bug Report as this helps us pinpoint the cause better additionally to being able to fix it faster than without bug reports.
While criticism of Cursor or users is fair, please keep it civil 
and constructive. We all want a positive and helpful Forum.
I believe that OP may have seen people calling them out on various threads for a reason but I will not join that discussion in detail apart of saying the rules apply to all.
3 Likes
I see you’re using anthropic’s sonnet 4 or opus 4 / 4.1 in agent mode with no instructions. Makes sense.
The OP really does need to create bug reports. I have been creating new bug reports every day. As the old (OLD) saying goes: The squeaky wheel gets the oil. Be a squeaky wheel.
I also know that, even though they don’t always respond to bug reports, they do seem to read and listen to them. I have requested a few very specific things over the last two months or so, and I’ve seen some new features crop up that directly address the bug reports I filed.
Its worth the time, IMO, if you really are interested in using Cursor.
1 Like
not only their product, but also their support team
What models are you using, the model you choose will massively impact your results in my experience. I think also there’s an element of style that mean some people don’t find some models as good as others.
Are you using sonnet 4, grok-1-code-fast or whatever it’s called, gpt-5, or auto (or something else)? Are you switching models mid chat?
I’m honestly struggling to understand how I don’t see every issue you are seeing?
I think there’s a few comments in your post that are triggering the response, like no undo. Do you mean revert doesn’t work? What can’t you undo?
1 Like
-–
alwaysApply: true ← I dont care
-–
Literally NEVER had any of these for the last 12 months. Mainly Linux usage now, but had a few months in Windows as well.
The only issue I have with the terminal is “restoring terminal command history” is bullcrap, it doesn’t restore anything useful, might as well tail the .bash_history to get better result, because it restores the same history from all terminals to all terminals, no sane person uses any terminal for any command randomly, each terminal has it’s specific purpose.
Name checks out.
Glad to hear you yourself personally, have NE-VER had a problem.
FWIW, that does not negate the FACT that others have, and often quite extensive ones. Fundamentally, 1.5.x is the worst release since 1.0 hit, and it has very severe stability issues. I have been using cursor every single day, 7 days a week, since the beginning of June. I’ve got a lot of experience with every single version that has been released now. Cursor was much more stable, albeit a bit less capable, up through around 1.3, then it started going down hill from there. There were some issues with 1.4, but wow…1.5 from a stability standpoint jumped off a cliff.
1 Like
Never encountered the issues most people have reported, only thing I experienced badly was model degradation and general uselessness during some periods, where models that could build a single 2000 lines feature in one prompt were sometimes not able to update a single line of code correctly.
I has deleted private files or wiped databases clean, but then there were no rules or features to prevent it from doing that, and it was the LLM’s call, not a Cursor issue in itself.
Again, just because you haven’t had any issues yourself, does not mean they don’t exist. Ever since 1.5 came out, using Cursor has been a low-grade infuriating experience. Looking at the bug reports area, other people are experiencing the same issues, and statistically around 10% or less of people generally report their experiences. So, I believe the issues I’m facing, are relatively widespread. The 1.5 drop was a disaster, has been a disaster, and is an ongoing disaster. This was a WILDLY unstable release, and as a professional, who now relies on this tool to do my daily job, I honestly find that unacceptable.
There are environments where CI/CD with the potential of introducing widespread instability or breaking bugs isn’t a big deal, because you have the necessary support teams and infrastructure to RAPIDLY resolve and deploy fixes to the issues. Cursor is clearly not in a position to do that. This isn’t Netflix. Further, when a bug frustrates 20% of netflix users for a day, its not a deal-breaker. When an entire release exhibits persistent instability across a range of features, with each subsequent build breaking more and more? That’s a deal-breaker.
I say this, in hopes that the Cursor team realizes they are in fact NOT Netflix, or any kind of massive cloud service for the masses. They are providing a professional tool for professionals, who have a critical need to move very very fast, and for whom breaking changes are BREAKING, fundamentally. The instability of 1.5, is really bad news. It is cross-platform (well, at least Windows 11 and Mac OS X in my experience, I haven’t used a Linux version). The issues are often quite debilitating. My velocity was about 1/10th of what it was before, due to the constant need to babysit everything the agent does now. That has had a significant impact on my life here…1.5 is what, a week or so old. I’ve worked a TON the last week, to get done what I needed to get done, which is now late (but thankfully it happened to be a holiday today! But not for me…because 1.5 is an unstable monster.)
The Cursor team needs to put in more dedicated effort in making sure they don’t ship DEBILITATING bugs. It has a devastating impact on end users. Even if it is “just” 20% of users, that is 20% of users who are unable to effect the requirements of their job, deliver on time, or have the personal time they are used to. 
The 1.5 release had absolutely zero changes that would affect agent resource usage. Could you share more detail with the issues you are having? We have read through the thread in detail and have released fixes for terminal focus and terminal execution (in 1.6/nightly). The terminal execution logic was also unchanged from 1.4, however we know this is a critical area of improvement. We have read all of your messages on that thread.
I don’t know if it is about resource usage. What I am referring to, is very unstable behavior with the agent and terminals. Both frequently just hang. I know that for the terminals, 1.6 is going to bring some major refinements.
The last few days, though, I have also run into very frequent hangs in the agent itself. I have actually posted several bug reports on the issues, currently in the bug reports section. I am not sure if the agent hangs are specific to Grok Code, or if they are more universal. In any case, the agent will start doing its work, and randomly, arbitrarily, it will just stop. It is clear that it is in the middle of work, it may show the bouncing elipses (…), or may even say Planning Next Steps or the like, and it will just sit there. I’ve left it for a good while, and its just stopped for as much as an hour, before I finally wrote another prompt to try and spur it back into action.
I’ve tested my local network and internet connection extensively. I have ROCK SOLID internet at around 900MB/s+ (my internet is 1.2GB/s), pings are table at <20ms consistently, I have tested my connection and routing to numerous different locations. I also have been running the built in network diagnostic tool a lot. Interestingly, it’ll start out seeming fine…test is fast, stable. But over time, that test seems to get progressively worse, slower…eventually, tests will start failing. At that point the only recourse is to restart Cursor. It seems to only be a cursor issue…i have no other connectivity issues. I’ve had this problem since at least the release of 1.5.5, but I was on 1.5.2 for a while and I may not have noticed it.
I do not know if the progressive decline in the network diagnostic results, is representative of the instability issues with the agent and its hangs. There are times where the agent seems to be fine, other times when it doesn’t seem to go even a minute without hanging. Its a very frustrating problem, much like the terminal issues: BOTH require constant babysitting of every active chat, which has a very detrimental (decimating) effect on my velocity. These problems have been going on since 1.5 was released, at the very least.
So it is not a memory/cpu instability thing that is causing Cursor to crash. It is a matter of instability in the agent’s behavior, its ability to carry out the work given to it in a prompt, and do so fully through completion without hanging either on a terminal command, or just arbitrarily in the middle of working on the prompt.
FWIW, I know when it hangs with Grok Code and Claude, because both CONSISTENTLY deliver a completion report after everything they do. When the agent stops, and I get no completion report, I know it hung. I will then either have to try and get the terminal instance to complete (which is really not a solution, if I can get my cursor into the instance, and hit enter, the agent will continue, but it appears that it is UNABLE to get the output of the actual last command…instead, it reads the output of me hitting enter, which is always an empty response, maybe just a status code), or if the agent itself hung, I have to issue a prompt along the lines of: “continue”, “pick up where you left off”, or “you seem to have stalled, re-try the last command and continue” (if the last tool invocation appears to have faild.)
This is day three, now, that I have experienced agent instability of this manner. I have been experiencing the terminal instability longer, on mac. I have experienced SEVERE terminal issues on windows 11/wsl2 since…oh, at least two months or so now.
I think you need to pay attention to your testing scripts, and use the cursor roll back function, Git will help to. I like you coding for a long time, in my case 45 years, and on proper rules and PRD no disaster, but I dont take my head out of the game, I check the results, and undo where I dont agree, ask the ai for 2 solutions, and in old code, I would use AI to suggest fixes and fix my self, since it may not follow the old frameworks, this is obvious I thought, but 1 quick way is to ask cursor to make a rules and PRD file based on the current structure and framework so any changes will follow the flow, and then run your tests before committing.
1 Like
This is the most peculiar ongoing issue. Nothing is intentionally changed but things behave unexpectedly different from version to version in a demonstrable way.
I might be lucky, but the most annoying thing in Cursor is that when you receive a new update you have to restart and reload everything, this takes a few minutes to reload chats and terminals.