Use Case #5: Browser QA Loop

I shipped a feature. It worked in dev. It broke in production. Classic.

Instead of debugging blind, I let Claude test it visually.

The Prompt

"There are problems with this explainer. The voice sucks, arrows don't work, animations aren't showing. Use your Chrome extension to test it and see for yourself and fix and improve until it is amazing."

What Claude Did

Phase 1: Visual Audit - Claude took a screenshot, analyzed it, and identified: text cut off at top and bottom (viewport issue), navigation arrows not responding, animations stuck on first frame, wrong voice playing for narration.

Phase 2: Issue Triage - It prioritized: viewport issues (blocking), navigation (breaking), animations (degraded experience), voice (polish).

Phase 3: Fix → Deploy → Verify - For each issue: trace the code causing it, implement fix, commit and push, wait for Netlify deploy, take new screenshot, verify fix worked. The voice issue required regenerating audio files. Claude identified the right voice, updated the generation script, ran it, and replaced the files.

Phase 4: Regression Check - After all fixes, Claude clicked through the entire experience: every slide transition, every navigation path, every animation trigger. Found one more issue: slide 7 had a timing problem. Fixed it. Re-verified.

The Before/After

Before: I would have tested in dev, deployed, gotten bug reports, fixed one thing, deployed again, created new bug, repeat.

After: One session, all issues found and fixed, verified on production.

What Made This Powerful

Visual confirmation beats logs. "The button works" in code doesn't mean the button works for users. Seeing it work is knowing it works.

Iterative verification. Each fix got verified before moving on. No fix chains that break each other.

User perspective. Claude tested like a user would - clicking things, looking at the screen. Not like a developer - checking function calls.

Try It Yourself

Copy this to run a visual QA loop on your own app:

I need you to test my deployed application visually. Here's the setup:

**Application:**
1. What's the URL to test?
2. What features should I test? (list main user flows, or "test everything")
3. What should I look for? (broken UI / missing data / visual bugs / all)
4. Should I fix issues as I find them? (yes / no, just report)

I'll use the Chrome extension to:
- Navigate through each feature
- Take screenshots and analyze what I see
- Identify visual and functional issues
- Prioritize by severity
- Fix each issue (if authorized)
- Verify fixes on the live site
- Run a final regression check

What's the URL?

This was the BizBrain OS explainer page, January 2026.