Find bugs in one homepage.
A deliberately imperfect TestBucks homepage. Try screenshots, console logs, and network evidence with a copyable OpenTest.ai bug-report prompt.
Open Lesson 1 →Pause and Practice
Examples to play with during the course. Capture evidence and try it with your AI coding agent.
Compare pages and dedupe issues.
Capture Home, Menu, and Stores. Attach the screenshots and bug list, then ask your agent to consolidate shared defects.
Open screenshot and deduplication practice →Crawl TestBucks and inspect the captures.
Collect evidence from the three linked pages, check it, and dedupe findings. No functional checkout testing.
Open crawl practice →One agent. Six tasks.
A worked example using isolated browser workers, bounded tasks, and a combined evidence report.
Open the parallel-worker example →Resize text and review meaning.
Compare broken and reflowing layouts, inspect alt text, and draft evidence-backed reports.
Open accessibility practice →Who ever tested the human?
Compare ten minutes of your testing with ten minutes of AI testing.
Test yourself against AI →Turn legacy scripts into intent tests.
Download Playwright or Selenium examples. Copy the source and conversion prompt into your AI coding agent, then review the OpenTest.ai intent output.
Open legacy conversion practice →Make the app more testable.
Copy the full TestBucks HTML with a prompt to add stable data IDs, useful diagnostics, and repeatable synthetic data controls.
Open testability practice →Record once, replay intelligently.
Two prompts: perform a public Starbucks browsing flow and save its OpenTest.ai intent, then replay that intent against the current interface.
Open the two-prompt lesson →Same homepage. Three perspectives.
Ask AI to review the Starbucks.com homepage as a rushed commuter, keyboard-only user, and first-time customer. Compare qualitative feedback and star ratings.
Open persona-feedback practice →Find risks in the source code.
Copy a checkout sample and ask AI to identify validation gaps, unsafe retries, and dependency risks—or use your own approved project source.
Open source-risk practice →Test the boundaries. Question the rule.
Try a last-name form, copy its source with a boundary-test prompt, and distinguish implemented limits from intended behavior.
Open form-boundary practice →Inject a fault. Expose it. Verify recovery.
Ask your agent to create a local practice replica, inject a mock payment timeout, drive the error dialog, and check retries for duplicate charges.
Open payment-timeout practice →Four customers. Four different orders.
One prompt coordinates isolated browser workers, different synthetic customers, and separate orders—then combines their results.
Run parallel customer journeys →Turn results into a beautiful report.
Copy a pass/fail grid and ask AI for a custom TestBux report with branding, subtle animation, and an executive summary.
Create a quality report →Combine risks, results, and missing evidence.
Create a branded release recommendation with ranked risks, untested areas, evidence gaps, and the next tests to run.
Build a release-confidence report →Explore the working coffee shop
Try customization, cart updates, discounts, and simulated checkout.
Open TestBucks →Synthetic training examples. No AI service is called by these pages.