Comment by woah

Jev's bet is that if it works well enough for random use cases that nobody complains, then management won't feel a need to develop a benchmark/eval, and they won't need to employ all those data science guys.