Cursor v1.5 - Release Discussions

We’re looking at improving this now. Thank you

1 Like

I just updated to
Version: 1.5.2
and the bug where I have to scroll the chat all the way down each time is still not fixed.
What’s going on with this?
Also, maybe you could add a small arrow that allows easy scrolling down
that would ease the problem, in case for some reason you can’t actually fix the bug.

Sorry to hear this - we had thought this was fixed. This is when switching chats on windows?

1 Like

Yes.

Interesting—while I’m talking to you now, I see that some of the chats do stay fine.
I’ll run a few tests and update you.

v1.5.0 ~ v1.5.3 unresponsive then crash, use cmake-tools build project

Something really strange happened to me.
For a few minutes it worked fine—on chats that weren’t moving, meaning chats where I had already scrolled to the end, I didn’t need to scroll down again.
Only if the agent was working did it jump back to the top.
Then another update appeared.
I clicked update, and it installed.
And then the problem came back—it got as bad as before again.
But the Cursor version number didn’t change.
A weird and complicated story.
And now I keep getting this error all the time too.
@andrewh @condor @tmm1

@andrewh @condor @tmm1
Now I’m here, and this problem has gotten much worse—like someone fixed it and then broke it again, and along with it also broke the update system.

Version: 1.5.3 (user setup)
VSCode Version: 1.99.3
Commit: aece02022292184bb34fa0c899a97928489b7b70
Date: 2025-08-22T04:48:34.365Z
Electron: 34.5.8
Chromium: 132.0.6834.210
Node.js: 20.19.1
V8: 13.2.152.41-electron.0
OS: Windows_NT x64 10.0.19045

that is the latest - 1.5.3 - is it just the chat issue? we are releasing a couple of updates only to beta users to resolve some issues reported in this thread

1 Like

I’m not on beta, and I still got the update!
Also, for a moment the problem was better in the previous version—definitely better than it is now.
And now it’s completely bad again.

Now every time I switch tabs, it jumps back either to the beginning of the conversation or to some random place in the middle.
What’s really a shame now is that, like I said, there was a version in between where it was a bit better.

Got it. Between these versions, nothing related to chat scoop changed, but we are looking at it.

1 Like

I can promise you—there were five minutes when, switching from chat to chat, as long as I didn’t run the agent, the chat stayed at the bottom!
Scroll all the way down.
Now, if I switch to another tab, it jumps to the top or to the middle.
But not that it really matters :grin:


Thanks for removing keep all button, we definitely not used it before.

Hi, it seems v1.5 has set the max timeout of an MCP to 1 minute, can we please bump that (or add a setting)?
I have MCPs that often take a couple of minutes to complete and these are now timing out at 1 minute which they weren’t previously.

I’ve turned off the new Al-based Terminal Completion Detection setting but that didn’t seem to help.

Thanks for you hard work Cursor team.

The errors that I have are not related to the UI.

They redesign the UI in all kinds of ways, but leave a bunch of UI bugs and at the same time trigger a bunch of non-UI bugs

CleanShot 2025-08-21 at 10.20.15@2x

CleanShot 2025-08-21 at 10.20.15@2x1060×354 32.2 KB

First, my Cursor opens a workspace containing two projects (frontend and backend). I defined a project rule in the backend and set it to trigger “always.”

I noticed that in version 1.5, as shown in the image, a project rule was automatically introduced. However, it doesn’t seem to determine whether to reference it based on the file or project I’m currently editing, nor can it be manually canceled.

When I edit the frontend project, this rule, which should only apply to the backend, is also pulled into the context. I hope this can be improved.

How do I see these stats? They do not appear I guess because I’m in MAX mode? I still want to see the spend, would love this.

GPT-5 yesterday slowed down to a crawl, at about 1 token/sec or even slower. At the same time GPT-5 in warp worked fast. Just an observation.

I just got the update to 1.5.3 and still having the same issue, am I the only one with this issue?