I was thinking, is there a way to run a duplicate instance of Cursor in 1.3, like a second Cursor app. I’m on a Mac - I remember back in the day there was a hack for Dropbox where you could create a second version. Something like this:
… That’s the whole point of joining the pre-release group. So that you can help test, and provide feedback. That way if you have strong opinions about how Cursor should be made, you have a chance to make them heard.
Cursor should absolutely be doing this. It’s the right way to have a pre-release version.
You never accidentally roll out beta changes to stable users
Beta testers can switch back and forth, which is helpful for verifying what changed, not to mention being able to use the stable version when pre-release is totally hosed
Now I understand why Cursor support is non existence. The fury the community is having. I am feeling the pain but I guess its doesnt really feel real until now.
I am using claude code for the last 3-4 days after cursor support kind of forced me to look at other sources. And I dare say that Claude Code performed exactly as what I expect for a product that helps me code. I am using it along side Cursor now but I think I am begining to see how claude code will just kill cursor within 6 months if cursor dont undo the screw up it just did past weeks.
It will be an epic lesson in what not to do with your customer base and how to communicate and support your loyal customer base.
I am all for cursor to make money. Heck even BEFORE the change of pricing I have already asked cursor support if I could pay higher for higher plans/quota. If a product is good, we should support it. But now the product is getting worse and support is in panic mode it seems.
I do hope cursor wake up and quickly get their bearings right. For now I will just enjoy coding on Claude Code and using a 20 bux cursor for light tasks.
Always know how your customer needs it and dont screw it up for your customer.
Latest push seems to have tweaked prompts again so that in Max mode the agent is again (despite my many prompts and rules saying not to) reading teeny tiny small increments of files and using things like head and grep to snippet pieces. Two issues, I want it to read whole files, I want the context. 2 it keeps re-reading bits and pieces of files it already has read. Seems like there is a prompt that says the read tools are more efficient reading 100 lines at a time and its trying to do this. For max mode can you please change this. Or even better, can you make it so these system prompts can be edited by power users to adjust behaviors, because as we know it wont always follow later commands over earlier ones. Can be very frustrating.
And on that thought, it seems like the product team at cursor is fighting two opposing battles. 1 to compete with the likes of lovable and claude code and just have a completely utterly fully autonomous agent, and 2 how I (and most? users of cursor want) to have a pair programming buddy that we can oversee and guide. Perhaps you would save a lot of headache and heartache for your team by making these two modes selectable.
I’m enjoying the update, with a few caveats. The restore from checkpoint icon seems to disappear, I have to jump through some hoops to get it back, like clicking below the prompt, then in the prompt, scroll up/scroll down. So far I’ve always gotten it to show up, but not sure why it has to be so dynamic, it should just be there.
And I’m not sure why the prompt hides the files I’ve added to context. I find myself having to click the prompt again, to make sure I’d connected all the right files.
“If you make a plan, immediately follow it, do not wait for the user to confirm”
“For ANY commands that would require user interaction, ASSUME THE USER IS NOT AVAILABLE TO INTERACT”
And the insightful AI notes:
“The main tension I see is between the system preference to minimize asking the user questions (find answers myself, assume user unavailable) versus your rules requiring explicit authorization before making changes. This creates a pattern where I try to be self-sufficient and action-oriented, which conflicts with your requirement for permission.”
And it’s driving me abso-effing-lutely crazy. I end up having to say the same thing over and over as it runs off and changes without being asked. If Im asking a question, code change is NOT the answer! Not sure what other prompts changed recently but its gotten very very painful (today? very recently?). Tiny file reads and overactive responses. I dont want an autonomous agent, Im looking for a coding partner. If I wanted fully autonomous I’d use Claude Code without an editor behind it.
The new upgrade destroyed my entire app, my SDK, a literal disaster. I was quite shocked, the last 3 weeks have been an awesome experience until this upgrade came along. Can’t hold a chat with out restarting a new one every few prompts. I’ll probably downgrade for now until this new version gets worked out. Lol
I had exactly the same problem when I was on 1.2.4. Everyone was getting 1.3 and newer betas, and I was stuck on 1.2.4.
Just download the newest beta from https://www.downloadcursor.app and then the updates will come as they should.
More cost efficient? I don’t care about costs. I care about productivity. If you ruin the quality and the efficacy of code gen for the sake of costs, using Cursor becomes pointless. Augment charges almost twice as much and is still seeing adoption because of the mere illusion of having better code gen with longer context windows.
By the way, if what I read above is correct and you’re removing Manual mode, the day that update gets shoved out to me is the day that I cancel my subscription. If I wanted to use random
models, I’d go to Las Vegas.
I’m running version 1.3.3. I didn’t sign up for beta releases, but here we are.
There are some problems with this release. It seems like it hangs for no apparent reason every once in a while. Sometimes it crashes completely. Sometimes it seems like I need to click the “Keep All” button to accept code that I already accepted a while ago. Maybe a bug? Maybe something I was doing wrong? Not sure.
However, I see the update available message today. How do I find out what has changed if I hit “Install Now”? Will it fix any of these issues I have seen? Is there some new feature? I see nothing in the changelog. Nothing in the app itself. Surely, someone at cursor knows what has changed?
It would be really nice if there was another button next to “Install Now” labeled changelog or somethien similar. Is this possible?
