At first I thought it was because of Anthropic’s servers lacking GPUs.
But it is now happening to all models.
It’s especially bad when it takes 4-5 minutes to respond and it makes a mistake by using the wrong syntax or because of some incorrect assumption.
Even when it worked at low latency it sometimes took a couple responses for Cursor to get the solution to a problem. Which was fine, because you could just try again if it misunderstood your prompt.
But now some things that used to take a couple minutes could easily take up 10-20 minutes.
Is there an explanation or a solution in sight?
I really enjoyed working with it (when it worked) and I’d love to know