Reconstructing the benchmark behind Luc Julia's 64% LLM reliability claimgithub.com5 points·matthieu_bl··0 commentsOpen articleSaveView on HN