Dan Selsam is one of the most brilliant AI researchers I have ever met. Just a few months ago he was very sceptical about model capability trajectories, and felt that radically new paradigms were needed for progress. You can disagree with him but everyone needs to read this