lifelonglearner comments on Reducing Agents: When abstractions break