Post

Is there any paved path to training an agent? That is you give them complex tasks. They come back you give them some feedback and then a meta learner tweaks their prompt. I understand there’s RL for work tasks, but I assumed those fine-tune it. Talking more about a DSPy thing.