Comment by cheesyFish

It's definitely tough today, but its just a matter of a) using a smart LLM b) scoping down individual agents to a manageable set of actions

As more LLMs come from companies and open-source, their reasoning abilities are only going to improve imo.