Hey everyone,
Just wanted to share some frustration and see if others feel the same way. I’ve been using Cursor as my main editor for a while now. I’m a full stack dev and loved the AI-first approach, especially how seamless it was to refactor, generate boilerplate, or debug with natural language.
But lately, the “No changes made” issue is getting really out of hand. Even with clear prompts and relevant context, the AI keeps returning that message and does… well, nothing. It’s breaking my flow constantly, and honestly, I’ve started reaching for regular VS Code again when I just need to get stuff done.
The timing really couldn’t be worse too. With VS Code now introducing Agent Mode, it feels like the pressure is on and Cursor is dropping the ball right when they need to prove they’re still ahead. I want to keep using Cursor, I really do. But I can’t be fighting the tool while I’m trying to build things.
Is anyone aware of an official fix or workaround coming soon? I saw a few other threads about this but nothing concrete.
Would love to hear how others are dealing with this, and really hope the devs get this ironed out soon.
Cheers
3 Likes
Update: I deleted Cursor and downloaded the previous version (0.47) from the Downloads page - it seems to be working now!
2 Likes
I’m not sure, but i did see this reply though from a staff member:
It would be great, indeed, because I’ve wasted hundreds of credits because of this issue. I had to tell gemini to resume every time or so, after several attempts it sometimes succeeded to edit files.
I had to switch to Trae to get critical things done. Not good for me, and for Cursor, and for my wallet.
So yes, I’m also thinking about moving from Cursor to another AI powered IDE if things are not resolved (Gemini 2.5 works great with Trae, no edit issue, and it’s free to use, even if often system is busy).
I’ve heard about Trae but i have some privacy concerns about that, they don’t have a privacy policy like Cursor does. By the way, 2 hours in since downgrading to 0.47 and i’m getting work done again with Cursor. Since the latest update Cursor kept breaking my code resulting in 100+ compiler errors but that’s not the case anymore so yeah, 0.48 is definitely flawed.
I don’t share my main business app with Trae because you’re right about security / privacy concerns. But I’m also building a quite complex Chrome extension, without any link to my core business, nor data processing, so that’s fine using Trae. They could replicate my extension, they won’t really be able to use it 
1 Like
I still have the issue with gemini-pro-exp-02-25 not making changes.
Yeah that problem seems to be with Gemini, Claude works fine though.
Hi @danperks
Is this issue still being looked at?
Thanks.
It happens with gemini-2.5-pro as well.
I’m still being charged even though it didn’t make the changes I asked it to:
I’ve spent quite a lot on these wasted calls this month.
I’m sure they’re aware of it and working on it. Perhaps even an update this weekend? We’ll see.
I sure hope they do something, The last stable-ish feeling version imo is 0.46 which just happens to be the last version that’s able to also work side by side with co-pilot chat.
I just dont get it, your charging and billing is always good getting ahead, but the software performance is getting worse and worse everyday, the Max that was premium then feels like a Grok 2, and its supposedly be the superior one, and yeah please dont fool us, as I once witnessed that QWEN 2.5 is getting used without us knowing its used, how if you may say? I saw the QWEN2.5 spam error late in the wee hours of one of those implementation changes that you did last month.
hear me out, I’m a loyal follower and I believe in your product and I’m actually considering it to be a de facto standard tool in Dev ops operations, but don’t push us away please! Stop acting like a startup!
There is currently a known issue with Gemini where it wants to call a tool but just fails outright. While Google have told us a fix should be available shortly, as a workaround, Gemini also likes to echo out the code changes it wants to make before trying to apply them to your file.
As such, you can use the Apply button in the top right to apply it to your file without needing to make another request.
I’d suggest switching to another model for the next day or two if you are still seeing this issue, but I’m hopeful we’ll get this behind us by next week for sure!
Thanks for the update @danperks.
As part of putting it behind us, will you be giving a refund?
1 Like
Yes, because I lost lot of credits and money.
It’s not just Gemini though. Here i’m using Claude 3.7 max and it’s eating up requests without doing anything too. It’s always a gamble with using the max tool.
