It will still often tell you that it can do things that it can’t or vice versa, it’s like it’s guessing whether it might be able to do things or not based on what humans told it that it could do in the training process rather than actually being aware of its own capabilities