There's a Benchmark Test That Measures AI 'Bullshit'—Most Models Fail | Coinfeather