Burg AIburg ai
Reference & SupportFeedback & Learning

Improving Signal

Tune reviews for better quality over time

Improving Signal

How Teams Can Tune Reviews Over Time

  1. Use custom instructions (Pro) — Add context about your project
  2. Adjust severity thresholds — Filter out minor issues
  3. Add file exclusions — Skip generated or irrelevant code
  4. Provide feedback — React to comments so we can improve

Common Mistakes Teams Make

Mistake 1: Too Many Files Included

Symptom: Reviews are slow and expensive

Fix: Add exclusion patterns for:

  • Test files (if you don't want test reviews)
  • Generated code
  • Vendored dependencies

Mistake 2: Custom Instructions Are Too Vague

Symptom: Reviews don't reflect your preferences

Fix: Be specific:

  • ❌ "Be stricter"
  • ✅ "Flag all async functions without error handling"

Mistake 3: Ignoring False Positives

Symptom: Same wrong comments keep appearing

Fix: React with 👎 so we can learn

Mistake 4: Not Setting Severity Thresholds

Symptom: Too many minor comments

Fix: Set minimum severity to "Major" in settings (Pro)

Best Practices

  • Start with defaults, then tune based on real feedback
  • Review your settings quarterly
  • Check which file patterns generate the most noise