Had a blast yesterday using the new Claude 3.5; was able to make 10 commits in a day, refactoring a huge codebase; now I can barely get one commit in, with tons of debugging to manually iterate on. Also had been hitting the Anthropic rate-limit (tokens per minute) on a codebase search (just one prompt in); now I don’t even see that error log anymore, making me think Cursor been cutting off the context aggressively? Some transparency maybe?
1 Like
yeah i noticed it like dramatically worse. I hope no cuts and nerfing is done at the backend we need more transparency
1 Like
For some reasons (contrary to yesterday); claude-3-5-sonnet performs way better than claude-3-5-sonnet-20241022 right now. Not sure if it is currently pointing to the old or new version…
quick update here: claude-3.5-sonnet now points to claude-3-5-sonnet-20241022!
Very frustrating to be in momentum only to hit a rate limit.
1 Like
@rishabhy is the new Claude model really ready for use in Cursor? It has given me poor results vs the original Claude 3.5 Sonnet. I see many similar experiences / complaints.