Currently, I am coding a project with Node.js and PHP, including multiple plans and features. It seems that Cursor is reading through the files in my project. Is my current token usage reasonable, and is there anything I should optimize?
One thing that works for me is not switching models mid-chat. Even though it can be very helpful, it makes all cached tokens be redirected to the new model, so you will be charged for that, and the agent needs to get context all over again.
Keeping conversations short also helps, and obviously, knowing which model to use in each case!
1 Like