Failure taxonomy v0 and the first end-to-end baseline

Baseline lab: replay 20 recorded support traces through three binary checks at $0, then rank the failure modes by trace count. The first scorecard reads 16 of 20 (80.0%) with one check and one mode; the pinned baseline reads 10 of 20 (50.0%) and names four modes, led by refund_past_window at 4 of the 10 failures. Fixing the top three is projected at 19 of 20 (95.0%) — fixing the three the team happened to notice stops at 18 of 20 (90.0%).

This Evals Engineering lesson is part of a paid plan. Sign in with a plan that includes this course to read it.