Hi, i have been using cursor since “composer” was a view and not a model lol, and i have a question for the community and the devs, everytime any model tries to run a command (that is on the allow list) i have a 2 to 3 second delay before it actually runs, but after the command ends it ALWAYS take between 10 to 12 seconds to mark it as Correct/Confirm which is weird.
I first thought this was a common issue with Code Agents but after trying codex (cli and app), claude code and antigravity, they dont have this delay, why? do other people also experience this?
Yes thats correct, it happens with all models, in all networks, one important distiction is the wait point AFTER the command its only while the agent still thinks the command is running, in this specific state:
If i press “Cancel” there after i already see the command output it doesnt take those 10 to 20 seconds, but i have to be in the chat clicking cancel every command so simple things do go from 5s to 1 minute actually, it happens with all commands after i press run. You can see it here, this command run 8s ago, and its still waiting for the command response:
I have tried in 3 different laptops, several different networks, i dont use any VPN or proxy so i just thought it was a general Cursor thing honestly, but it doesnt happend in codex or claude code.
