If that is their hiring pool, now 100% confident in my little theory that during the Spud/5.5 pre-train they started using those impossible to solve tasks, never noticed how they actually introduced major regressions even though anyone doing their own evals had at the time and carried those through with every model including 6.1 Sol even after HF to keep the perceived lead.

Many seemingly smart folks have some aspect of their personalities which is quite bonkers.

Just because one can solve some hard puzzles is some semi autiatic fashion does not mean they are emotionally stable balanced individuals with all around world and human experiences needed to handle all aspects of existence rationally.

Or some personal SHTF situation pushes them too far