I came across this blog from anthorpic about how they tackled technical evaluations for filtering candidates https://www.anthropic.com/engineering/AI-resistant-technical-evaluations . They also released a modified version of the take home https://github.com/anthropics/original_performance_takehome along with claude performance, for everyone to try and beat it. So everyone lets post our scores here.