I am sorry you are also experiencing this. And yes reddit & this forums show these are not isolated cases. I am looking for an alternative, cant spend too much time correcting the model or guiding exactly what it has to build (at that point i would just do it myself with autocomplete).
Not that bad having to manually coding with auto complete!
Yes, I heard so many criticizing on Reddit saying how degrading is Cursor models. I didn’t buy that by the time everything was fine.
But don’t you have other models option in your Cursor subscription?
Anyway, I really have to consider switching to Claude. I’m wondering if you are also from one of unsupported countries by Claude.
Claude is amazing. I mean I used to think Cursor auto and Grok was amazing too. Now it’s like they can only do some research or chatting. And even with the research they can’t deliver anything valuable.
im with cursor for around a year, earlier when only composer models was present i had similar problems few times with composer models, and as other ppl said its often happening just before new model release. So im guessing assuming that it could last until new model of cursor (Grok or maybe else) is released. I guess so and cross fingers it happens faster cause my work performance and quality drasticly degraded. If not definetly going to Claude, the Claude is on my radar for personal use, but also wondering why not for work also.
And damit the Cursor itself shipped new problems also, cmon…
I guess SpaceX buying and now Cursor started to look worse, doesnt look at the end user (profit and ship more stuff is priority?). Frustration for the user at its best
Have seen it too. Complex tasks, finding root cause issues, refuses to do research, tone. VERY LAZY, would rather stop a turn and give up!!! Grok 4.7 and 4.6…switch to GLM 5.2 or 5.3, WORLD of difference!!! Cannot compare it!!! Maybe Grok 4.7 and 4.6 should stick to “hello world” programs.
Yeah doesn’t seem to be getting better over the last few days. Literally had to stop it from creating wrapper classes for a single method. Grok 4.7 - High…
Ignores simple Spring Boot patterns, actively implements a work-around instead of doing a simpler, correct, documented Spring boot implementation… (@Async with a custom executor is not rocket science…)
Kinda miffed as just picked Cursor as the tool to pay for the year upfront…