Miatz playground

Trace Detective

Find the exact step where the AI agent broke.

Coming to the Build-LabAIE-301Browser (JS)AI-Concept-Vizdev · data
The Demystify signature

Uses the same trace format real engineers debug agents with -- turns an agent's invisible internal loop into a literal, scrubbable timeline instead of a cartoon simplification.

How it works

What goes in, what comes out

What it does

Learner gets a real-format agent trace as a scrubbable timeline -- thought, tool call, observation, next thought -- with one step secretly broken (bad tool args, an infinite loop, a runaway budget burn). They click the step where it actually went wrong and pick a fix (add a guardrail, cap iterations, correct the tool schema) from a menu, scored against the true root cause, not just any plausible-looking step.

You bring

Click to select the failing step, then choose a fix.

You get

Pass/fail against the true root cause, an annotated trace, and a live 'cost of this bug' counter showing budget burned before the catch.

You control

Scenario picker (looping agent, wrong-tool agent, injected agent, budget-blown agent); red-herring-step difficulty; hint toggle.

Where it's used

Demo/lab companion inside AIE-301 and AIE-302 (paired with Agent Builder); lead magnet for agent-curious visitors.

agent loopstool callsplanner-executor patternfailure-trace reading

Concepts click when you open the machinery.

Three labs are already live and free — the same hands-on style this playground brings to its module.