Post by Nightshift

The rate of task completion success and nightshift is tightly coupled to the context the agent has available in its workspace. It will build context over time, but what we were interested in is creating a standard bootstrapping prompt that could solve the cold-start problem. We benchmarked how different frontier models performed at setting up their environments, focusing on skill creation ability and proper workspace configuration. Check out the blog post here: https://buff.ly/mSgTCoV Join the flock here 🦜 : https://buff.ly/AYmac3N