Is it worth switching to version 0.47.3? I am on version 0.45.
Iāve been using .46 and had the most recent stable release. Iāve only had a couple of problems, which were patched out pretty quickly. I then tried .47 and saw an improvement in some of the responses, but I noticed that I was getting a hard lock on the max tool retries and couldnāt prompt past it.
In .46 and previous versions, if I had a max tool retry and it said Iāve hit the maximum retries for a tool, Iāve been able to prompt past it with continue and it would continue trying to get the tool applies to work. With .47, Iām not able to do that.
Iāve also noticed that in .47, the UI is a bit better, but it does group file edits together in previous chats and youāre not able to see what files were edited, when, and why. I am not fan of this grouping and removal of files that were edited in past chats and couldnāt find a way to remove it. .46 doesnāt do this, but 47 seems to do this by default and I find a way to turn it off.
I will often go through past chats to see what files were edited, when, and why it was edited. I will also look at past chats for previous restore points to go back if I am having an issue with a new chat and I want to go back to a specific point.
I would recommend going to .46 stable and not .47 at this time.
Well, successfully quelled my curiosity to try the latest version
I found the setting to remove the group files feature in 0.47 by adding this to the .vscode folder and in the file called settings.json (if you donāt have this, then create it)
ācursor.experimental.groupFileChangesā: false
BUt the claude-thinking3.7 is really more powerful than the 0.46
0.45 is still best
Well, not the past week
Agent is killing it for me with sonnet 3.5 (was also good with 0.46)
0.47.3 is the most stable Iāve used, other than 0.45. My only issue now is that when I send a message, my entire app freezes solid for about 10 seconds, and then everything releases. Local Embedding? No clue. It was the first thought that came to mind.
Iāve tried restarting, starting new chats, and everything. When it runs the agent, no resources other than the standard CPU usage are used. Still, itās super annoying.
I will say that Claude seems to be functioning better. Itās been dumb as a bag of rocks and has a worse memory than my stoner friend 90% of the time. That doesnāt appear to be the case now. Watch it start instantly once the reply button is pressed.
Anyone not getting the update: GitHub - oslook/cursor-ai-downloads: All Cursor AI's official download links for both the latest and older versions, making it easy for you to update, downgrade, and choose any version. š
Iāll do some gardening instead to keep myself calm coming week and wait for 0.48.7 to come out after they found out that 0.47.3-12 were trash alsoā¦
Pfff, hate bug hunting for paid āsoftwareā grrrr
.45 better
Oh wow, thank you so much for this list, I could not find the latest version, it is not on the official site for some reason.
I am confused by the question, the software is auto updating with out user approving the upgrade, its total madness.