An advanced guide to AI product evals and what to automate

The post says nearly half of 25 product manager openings it shared the previous week asked for experience writing evals. It cites company-reported gains from investing in evals, including Ramp taking automatic receipt-collection accuracy from 35% to 83%. It also introduces an advanced follow-up to “Building eval systems that improve your AI product” by Hamel Husain and Shreya Shankar.