Agents are repeating software’s dumbest mistake
On November 5, the company introduced CopilotKit Soar, a software solution designed to turn mundane tasks performed by users into skills for agents. The software learns from real workflows and can be evaluated on actual tasks. Users or employees are already executing the tasks that agents are being taught to do manually.
The introduction of CLiF, or Continual Learning from full-stack trajectories, aims to address this issue. What users already know how to do, can now be used to teach agents. AG-UI carries out what the agent does, while Continual Learning adds what the user does, all in the same open stream. This feature is supported by major agent stacks, allowing users to keep their current one.
Agent events and user actions are combined into one full-stack trajectory. Trajectories remain in an open format, enabling export to other tools. They can be sourced from production traffic or flows recorded by product managers (PMs). The software is built on the open AG-UI protocol, developed by the same team.
CopilotKit Intelligence can be self-hosted within a user's Virtual Private Cloud (VPC) or on-premises environment. Trajectories and production-based evaluations can be exported to the user's preferred tools. Currently, product trajectories are in early access, while skill creation and Intelligence self-hosting are available.
Written by urgent.news from Daily Dose of DS's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
