Optimizing Computer Agents for Efficient Task Execution Tracking

·

With a Computer Agent, is it possible to have the agent keep track of how it accomplishes the requested task and then reference that each time it is called on to complete a task? Every time the Computer Agent starts trying to do its thing, it burns through a LOT of credits just trying to re-discover how to accomplish the same task it accomplished yesterday. Does that make sense?

👀

1

1 comment

· Sorted by newest

Alon J.
·
·
Hey Tom! Ahh, you mean sort of learn from it's previous steps so it doesn't make mistakes again? I don't believe this is possible, but a way around this would be super specific with your prompting. I know, not a great answer 😕

Loading comments

Alon J.
·
·
Hey Tom! Ahh, you mean sort of learn from it's previous steps so it doesn't make mistakes again? I don't believe this is possible, but a way around this would be super specific with your prompting. I know, not a great answer 😕