Logic getting progressively worse?

I’ve been using Cursor for a few months and recently, especially starting this week the logic of the AI is just a lot worse than it was weeks, months ago.
Has anyone else noticed this? What models are you using?