Digital Agent outperforms o1 by 15% – trained with new RL-variant similar to R1arxiv.org11 points·let_tim_cook_··0 commentsOpen articleSaveView on HN