Skip to content
Vaizer

Watch a run

Long autonomous loops are usually a wall of logs, or nothing at all. Vaizer reframes a loop as a journey toward victory, so you can answer the only question that matters, is it making progress, at a glance.

Fix-until-green loop · Claude Code

🎯 Get the flaky checkout test passing in CI, then ship the fix

Attempt 1/12
  1. 1ReproduceMake the failure happen on demand
  2. 2Isolate causeFind why it flakes
  3. 3First fixApply a candidate fix
  4. 4Tests green20 consecutive passes
  5. 5ShipPR opened, checks pass, merged

What it is trying

  1. ✕

    #1Run the checkout suite once locally

    → Passed locally. The flake only shows in CI.

The path

Milestones are stops on a track. They light up as the agent reaches them, so distance-to-goal is always visible.

The attempts

Every loop iteration is one attempt: what it tried, whether it worked, and what it learned. Dead ends and backtracks included.

Your runs, next

This is a demo run. The same view is built to accept real run data from Claude Code or an agent SDK. That feed is coming.