The dude probably ChatGPT-ed the entire thing. I might as well ask those tools directly.
1) "Crediting" the source models explicitly.
2) Generating away or cleaning up afterwards: incorrect spelling, hands with unnatural fingers, non-sensical devices/symbols etc.