There are cases where it is too easy to choose the opposite action from what we are trying to choose. I’ve encountered two situations:
-
The [Skip] | [Run] layout changes the moment you hit [Run]; Suddenly the Skip button shifts over where the Run was. So if you accidentally double-bounce when clicking Run, you inadvertently end up issuing a Run → Skip pair. Then of course, the agent is confused, and you have to ask it to regenerate the command it was last doing so you can run it again. Instead of moving things around, just replace the [Run] button with a running indicator and leave the [Skip] where it originally was.
-
While reviewing files, clicking the [Accept] button on each file sometimes can result in the next file Reject [Accept] pair getting positioned slightly differently (or some other timing issue causing placement changes), and then an attempt to accept becomes an unfortunate Reject. Going back through with the AI to reapply the rejected changes is difficult; it essentially means repeating all the work which was recently done :(.
-
Even worse! If it is busy thinking for a long time, and you suspect it is stuck, you can click [Cancel]. Instantly when you do, the [Revert all] button moves to where you just clicked. So then, if you double-bounce, you end up ALSO reverting all your changes.
So in summary, please make it so that clicking in the same (x,y) on the screen clicks the same choice each time instead of sometimes the opposite choice.