Cursor is super inneficient

Feedback
1

Two things:

  1. What’s the point of having mid-response “Think” blocks, if you remove the previous Think block from the response? This leaves the model repeating OVER AND OVER the exact same CoT. Incredibly inneficient.

  2. The models, instead of applying one large edit on the whole file, they go and change one little thing at a time, and WAIT FOR THE SLOW APPLY MODEL to finish every little edit before continuing. Solution: let the models generate all their edits, and then apply everything at once.

2

  1. models need to evolve more in terms of reasoning for this to be addressed, imo. this is not a cursor issue..

  2. that causes crashes. using max they do large edits. this a token usage based problem.

1 Like