Noam Brown argues traditional AI benchmarks fail to measure reasoning time

OpenAI research scientist Noam Brown discusses static benchmarks' inability to capture reasoning effort. He argues that thinking time should be factored into model evaluation.