A whole four working days to adjust the model in between preprint release and the version of ChatGPT I'm using, then! Do you think that's plausible? I certainly don't.

Or simply the model was improved between the author's test and the release of the paper.

BTW the time stamp of the model is easily falsifiable.

We are talking about a billion dollar business opportunity so expect foul play all along.