← Pause and Practice

WORKED EXAMPLE · INTERACTIVE AI TESTS

One agent. Six isolated browser workers.

Give each worker a narrow task in the same TestBucks application. Compare their evidence—not just their pass/fail summaries.

Open the practice shop →
  1. Start with two workers if you have limited tooling or budget; expand to six only if supported.
  2. Use separate browser profiles or isolated contexts. This shop stores orders and cart state in localStorage: tabs in the same profile are not isolated.
  3. Copy the prompt below into your browser-capable AI coding agent. Let it confirm tooling and execution boundaries before running.
  4. Review each task’s observed result, evidence, and uncertainty. Consolidate duplicate findings.

What a useful result looks like

Six rows: worker, task, expected outcome, actual observation, evidence reference, and pass/fail/blocked. A blocked worker is not a pass. Ask for screenshots and relevant console/network evidence where available.

Example only: “Worker 2 — quantity update — expected subtotal doubles; actual result and screenshot required.” Do not substitute this example for an observed result.

Local versus cloud

VMs are optional; isolated browser contexts may be sufficient. Cloud workers cannot reach your localhost automatically. Use an approved deployed fixture or secure connection. Do not expose a debugging port publicly or provision paid resources without explicit approval. No agent or VM is launched by this page.