I have not ugraded to version 0.47 due to the disastrous experience that i had with version 0.46. I am still on a much older version for stability reasons. For those who have upgraded to version 0.47, can you please detail your experience. Has most of the bugs been fixed? Is the quality of output and context awareness much better ? Thanks as i look forward to your responses
I would love to know as well! Is this bug fixed: [BUG] Terminal tool gets stuck after exit code ?
i used the 0.47 version a little and had a terrible experience of the agent changing part of my code i did not ask for.
it changed my ui as if it did not understand my request that had nothing to do with that ui. i wonder if it is a problem that came with 0.46 and 0.47 or if it occurred on the 0.45 versiopns too.
Im asking because i was considering paying for the service in the company.
Personal experience: yes things improved a lot in 0.47, but I never had issues with performance or crashes in previous versions. My issues were outages of models, some tools etc…
Some other issues that were related to model handling in earlier versions were improved in 0.47 as Cursor internal parts were adjusted for better supporting specially the new models like 3.7…
However i have also improved on my side the prompting, understanding of model behavior and why things go wrong. Shorter but more focused chats helped, less context or shorter files helped (stuffing code in large files i not a good project structure).
Note that depending what your issue was initially and what caused it it may be better but nobody can guarantee it.
E.g. hallucinations when you have a long chat still happen, that isnt fixed by app installation as it depends on the content. Specially if you ask corrections and then eventually that trips up the AI later in same chat session.
remember to use new chat with summary to improve responses also don’t ask for huge refractors make small tasks for the ai, you will have less deletions. make a checklist of things you need to complete and give it to the ai 1 task at a time huge changes usually break the ai a lot
Yes I noted that using the new chat every time gets work done easier longer chats become an issue after sometime
I just pop-out the terminal if it’s stalled or not starting and then Cursor continues as if it never existed and tries again or does something different.
I’ve used the new system and had a few issues with the 0.47 release, so I decided to roll it back after testing. The main problem was the retry hard limit. Once the max retry for tool calls was reached, I couldn’t go forward.
In the past, I’ve been able to keep pushing it with a continue prompt and it would work, but in 0.47, I couldn’t get it to go past this. To me, this is a breaking feature, and I ended up rolling back to 0.46.
Yes, it works, but you have to sit and do it every time, this is very annoying
I can confirm: since i updated (i didn’t want to) to version 0.47.5, it keeps on crashing every 10mn or so… I used to work on the 0.45.17 and I never has that crash where everything closes and I just have the option to reopen… Very annoying, a big downgrade
Is there an archive to retrieve the older versions ?