Just thought I’d share, that there was a HUGE difference in Cursor reliability, stability, and behavior between 1.4.6, which I was on for a while, and 1.5.anything. I had a few issues with 1.4.6, and some were a bit annoying, but nothing truly ground me to a halt.
I was upgraded to 1.5.5 one morning after Cursor crashed (unknown why, it hung for a while, then vanished, then reappeared after a little while at the project selection screen). I had EXPLICITLY been rejecting updates to any 1.5.x version, as I’d been reading the forums and noted that people were reporting a wide range of issues with 1.5.x. The fact that I was upgraded, despite my express intent not to, after an arbitrary and random crash, is one frustrating thing. That really shouldn’t happen! IF a user chooses not to upgrade, they have a reason, and there should never be any circumstances under which an upgrade happens automatically.
Once I was on 1.5.x, I really started having problems. Just general unreliable behavior at first, but over time things became worse and worse. I upgraded to newer versions a couple of times, hoping that newer revisions would resolve issues. Instead, it seemed like each new version just introduced worse and worse problems. By the time I reached 1.5.9, which I was on for the better part of a day, the problems became so radically severe that I was literally UNABLE to use Cursor’s agent in any capacity. Even after downgrading to 1.5.7 and then 1.5.5 again, and even to 1.5.2 once, the severity of the issue did not change. Trying to issue a prompt in the agent, resulted in frequent hangs, from which I would sometimes hace to close and reopen Cursor or even reboot. Eventually, the issue became so devastatingly severe, that I would issue a prompt, get the bouncing elipses, and the agent/model would never do ANYTHING.
I tried to troubleshoot the issue, and would check the network diagnostics. Upon a fresh restart, the diagnostics worked fine. I could also use tab completions and most of the IDE, without issue. HOWEVER, the moment I would issue a prompt, then everything changed. The ide became unstable, would hang, freeze, when I would check network diagnostics, at first they would say there was a proxy on my network blocking the cursor requests (absolutely false), but after that, all attempts to run the diagnostics would result in either canceled requests, or test requests would be made that never completed. Eventually even quitting cursor wouldn’t resolve that issue, and it would persist even before using the agent. A full reboot was required to get back to a working state, however, that would only last until the first usage of the agent, then I would be back to network issues. It was like usage of the agent was blowing up the network stack in the app somehow.
I finally decided to get off 1.5.x entirely, and went back to 1.4.6. At first, it seemed as though the same problems had followed me all the way back to 1.4! I was encountering similar network diagnostic issues. After about an hour trying to get 1.4 working, out of the blue, like someone had flipped a switch, the connectivity issues vanished, and 1.4.6 started working again. I was still able to use Grok Code, so I just moved forward with my work.
I reverted to 1.4.6 on Tuesday. I’ve been using it since, and I have not had ANY of the issues I was encountering with 1.5, not since that sudden switch flip. The agent works, its been quite reliable. Occasionally the terminals in the agent will start stalling. I discovered that if I kill that terminal instance (open it then trash it), and instruct the agent to try again, then ti will create a new terminal instance and it will work fine for a while again. So there are some terminal issues, far less severe than they were with 1.5.x. I’ve been doing a lot of planning lately, working with MCPs to update third party services with stories, documentation, and other things. So I’ve had less terminal usage. Throughout all of that, the agent has been totally stable.
The frustrating thing is, I literally lost three solid days of work, because of the disaster that was Cursor 1.5.x. I worked through the entire holiday weekend, to try and catch up from other lost time, due in large part but not entirely to Cursor issues, from prior weeks. I lost ALL THREE days. Because 1.5 had such SEVERE issues.
This gave me great concern, about the paradigm that the Cursor team is following, with regards to how they release their product. This kind of devastating loss of time and progress, these kinds of debilitating bugs…they cannot happen when you have professionals spending $200/mo to use your tool. As a paying pro, I DEPEND on Cursor. As such, it MUST maintain its reliability, stability, correct behavior, it must function properly.
This is not a social media site, or some other online product, where little bugs here and there, don’t have any meaningful impact. Most users of social media are not going to LOSE MONEY or CRITICAL TIME because they can’t see the correct follower count, or number of likes, or something like that. Cursor, is an ENTIRELY DIFFERENT CATEGORY of product:
Cursor is a PROFESSIONAL TOOL, used BY PROFESSIONALS, for PROFESSIONAL WORK.
My experience with Cursor 1.5.x has GREATLY affected my outlook on upgrading to future versions. I know 1.6 is coming. I know it is SUPPOSED to bring some highly requested improvements, including specifically by me, such as a redesigned terminal integration system for all platforms (as I understand it). That is WELCOME, because I previously had DEBILITATING issues with the terminal while working on another project on Windows 11+WSL2.
HOWEVER, I can’t help but be extremely wary and trepidatious about the next version, because 1.5 was horrendously, HORRIFICALLY BAD. That was a horrible release, and I hope the Cursor team takes this to heart: It really shook my trust in the product. I’ve been able to do a lot of good work with Cursor, stuff often in just hours, that would have taken me days to do manually myself. There IS value there…but that value, can RAPIDLY VANISH when issues, especially the kind that get progressively worse and worse to the point where they literally HALT your work, like I experienced with 1.5.x (particularly 1.5.9…it was when I upgraded to that version, which was OFFERED by the IDE itself, that things REALLY took a nose dive and all hell broke loose).
I am really hoping that starting with 1.6, that deeper, broader, more stringent testing is performed. That the community of PAYING PROFESSIONALS, will not be used as the primary beta test audience for the product, while they are having to perform their DAILY WORK. I honestly do not think a classic social-media web site/app type of release paradigm will work with this product. We, your paying users, need stability, reliability, and a general form of consistency, because when Cursor fails, we fail, our work falters and fails, and the companies we work for, pay more, and its not just the monthly cost for Cursor…the cost of failures here is FAR greater than just $200/mo. I lost a lot of time, but THREE SOLID DAYS was horrific. Thousands of dollars, as not only did I lose the time initially, I had to spend more time recovering, which required I push out other tasks into the future, which is costing even more time. HORRIFIC!
Please consider your release paradigm, Cursor team. Please don’t treat your paying users like beta testers. Please don’t treat us like mobs and throngs of social media users. Every feature matters. Every quirk, every hang, every time we have to fiddle and fuss…its costing us money. This is a different arena, and testing & verification of each build for stability, reliability, consistency, is very important. I don’t know if a flip-flop approach to versioning might be helpful. Use odd-versions for “interim/unstable”, and even-versions for “reliable/stable”? For sure, 1.5 felt like an interim series of builds intended for those interested in playing with the latest features, but without ANY guarantee of stability, while 1.4 seemed overall more stable. Even 1.4, though, introduced new issues that were never resolved, that 1.3 didn’t seem to have. Which itself, did the same thing over 1.2…
Perhaps 1.6, could start a new flip-flop cycle, introducing a known-stable, known-reliable build that those of us who critically demand stability in order to be able to continue performing our jobs on a daily basis at the velocity we must, can STAY on, while 1.7 could be kept out of the “forced automatic upgrade” cycle, and then 1.8 could be another known-stable, known-reliable release?
In any case. The difference between 1.4 and 1.5, now that I’ve gone back to 1.4, is significant. Tuesday through now, I have not had any of the issues I was suffering on 1.5. That has finally allowed me to just hunker down and get some work done. I’m still behind, but now I have a hope that I might at least be able to get abreast of the curve again. Really hoping, NEEDING, 1.6 to be as stable, reliable, robust, durable as possible. At the VERY least, make sure we can choose to upgrade, don’t force any upgrades on us, a I just cannot handle another stint like last week and particularly last weekend, with so many issues, failures, unreliable behavior, hangs, etc.