With billions/trillions of dollars floating around, is it hard to imagine benchmarks could be biased?
I think it's safe to assume everything AI related is heavily biased until proven otherwise. Just like in pharma.
you didnt answer my question. Why would cognition be biased towards making anthropic look good?
People game benchmarks for fake internet points to get their favorite web framework to the top of the list. I'm pretty sure they will do it for billions of dollars.