I don't think this applies, isn't this just the researcher using ChatGPT Codex on their machine?
It seems like a lot of the recent "breakthroughs" come down to this: spending millions of dollars in compute to solve problems that essentially amount to recreational math problems