Were the agents ever tasked with algorithm improvements? Post just says he didn't find any ("report essentially no algorithm advancements"). These LLMs are usef

Were the agents ever tasked with algorithm improvements? Post just says he didn't find any ("report essentially no algorithm advancements"). These LLMs are useful for optimization tasks where they can attempt a change and then measure performance boosts, so just wondering.