L8 - The Lab output differ from results in Video

In video (Evaluation Part 1) (time 9:47), for user request
customer_msg_3 = f"““tell me about the smartx pro phone and the fotosnap camera, the dslr one. Also, what TVs do you have?””", the output is JSON and some text, that shouldn’t be in the output. When I try to run this cell, the output is only JSON.
Also, when in the video (time 16:51) the evaluation run trough 10 examples, the output differ from what I have as a result (by the way, in my case, the
Fraction correct out of 10: 0.8333333333333334
(3 incorrect, two of them have 0 score, and one 0.33333…)).

Why is that happened? Is it connected with the ChatGPT updates or something else?