Small decisions. Interesting possibilities.Submit contentSubmit

Measuring progress in agent traces

Got early access to @typesafeai Jev and spent a few hours building on it. The problem I keep hitting: agent traces tell you which tool ran, how long it took, what it returned but never whether the agent is actually getting anywhere. An agent editing, testing and reverting the same file six times looks perfectly…

Added to Jevfast

How it uses Jev

Jev evaluates each agent step for task alignment, progress, repetition, or signs of being stuck. JevScope plots these assessments over time.

App workflow

  1. Observe agent steps

    JevScope watches the actions in an AI agent's work trace.

  2. Assess what each step means

    Jev evaluates alignment, progress, repetition, and whether the agent appears stuck.

  3. Plot the assessments

    JevScope displays the per-step signal over time.

Why it is interesting

It adds an interpretation of trajectory to ordinary tool logs, making repeated or stalled work visible as a changing signal across steps.