Demo, all content is generated
Question

Is it me or did Cursor's Auto model get worse this week?

Open · 482 views · asked by sunita_d · edited

Same project, same kind of prompts, but this week the code is noticeably worse. Forgets imports, ignores my rules. Is there a way to know which model Auto picks? Should I just pick one myself?

What I’ve tried

Switched to a specific model for a day, felt better, but that uses my requests faster.

Comment
What kind of tasks? UI or backend? bea_quinn · edited
Mostly UI with forms. sunita_d · edited

3 answers

bea_quinn · edited

Auto can pick different models depending on load, so yes, output can vary day to day. For work where quality matters I pick a specific model and accept the cost; for renaming and small stuff Auto is fine.

Comment
sergio_ruiz · edited

Before blaming the model, check your side: did your project or chat get bigger? Long chats and a growing rules folder make any model worse. A new chat plus a precise prompt often "fixes" a bad week.

Comment
My rules folder grew to 9 files this month... I'll clean up and see. sunita_d · edited
ingrid_h · edited

Keep a small set of prompts you use a lot and compare output over time. Without that, 'worse this week' is really hard to tell from a harder task this week.

Comment