" However, various strands of evidence indicate that a"

Strands of evidence? My best guess would be that:

0- this is ai generated slop

1- it's using that watermarking technique

2- it's obviously detectable and degrades quality

3- it's amplified when inferencing on its own content and generates slop

Oh look! Peirceian firstness for LLMs!