# AI Code QA

**Using clip.qa with AI coding tools like Claude, Cursor, and Copilot.**

## The Problem

AI coding tools ship fast. But they ship bugs:

- 1.7x more bugs than human-written code
- Bugs in unexpected places
- Hard to catch because they're subtle

Traditional QA is too slow to keep up with vibe coding velocity.

## The Solution: Vibe QA

**Vibe QA closes the loop:**

1. **Code** — Write code with Claude, Cursor, or Copilot (or let the AI write it)
2. **Test** — Screen record the bug on your device
3. **Report** — clip.qa generates a structured bug report
4. **Fix** — Paste into Claude/Cursor, the AI writes the fix

## Workflow Example

### Step 1: Code with Claude

You ask Claude to build a payment form:

```
Build a payment form with name, card number, and CVC fields. 
Add validation and error messages.
```

Claude writes the code in 30 seconds.

### Step 2: Test on Device

You download the app and test the form:

- Tap the card number field
- Notice it doesn't validate the length
- Record this bug on clip.qa (30 seconds)

### Step 3: Generate Report

clip.qa analyzes the video:

```
**Bug: Card validation not working**

Steps to reproduce:
1. Open payment form
2. Enter invalid card number (less than 16 digits)
3. Tap submit

Expected: Form shows "Invalid card" error
Actual: Form accepts invalid card

Device: iPhone 15, iOS 18.1
App version: 1.0.0
```

### Step 4: Fix with Claude

Paste the report into Claude:

```
Fix this bug from my app:
[paste bug report]
```

Claude responds:

```
The problem is the card validation regex. Here's the fix:
[code]
```

You apply the fix. The bug is fixed in < 5 minutes.

## Benefits

- **Speed** — From bug to fix in < 5 minutes
- **No friction** — No manual reports, no back-and-forth
- **AI native** — Works with Claude, Cursor, Copilot
- **Accurate** — Video context makes the bug clear
- **Automatic** — AI generates structured reports

## Tools That Work

- **Claude** (web, Claude Desktop, MCP)
- **Cursor** (IDE integration)
- **GitHub Copilot** (VS Code, GitHub.dev)
- **Other LLMs** (Gemini, ChatGPT, local models)

## Best Practices

1. **Record while testing** — Don't wait to record later
2. **Narrate briefly** — "This field validation isn't working"
3. **Keep recordings short** — 30 seconds to 2 minutes
4. **Export immediately** — Don't batch; fix bugs as you find them
5. **Iterate fast** — Test again after the fix

## Get Started

[Download clip.qa](/apps) • [Learn Vibe QA](/vibe-qa) • [Read Guides](/guides)