I’m completely unable to convince the assistant there is a bug in the code it gave me, namely in collision detection and when I ask it to make test code it just “fakes” the intended result.
Anyone else faced this and is there any way to get the assistant to “rollback” its opinion or write reliable test code?
In this case I used to have working code but the assistant overwrote it.