@blubeeme @bmeurer the case study in the peril of optimizing for the "right" benchmarks is fascinating