“The case for more ambitious language model evals” by Jozdien | LessWrong (30+ Karma) | Podwise