AI for QA / Quiz
AI in CI: Useful Patterns, Safe Boundaries
Where an LLM step belongs in a pipeline (summaries, triage hints) and where it must never hold the gate.
- Difficulty
- Medium
- Format
- Quiz
- Points
- 150
- Estimate
- 7 min
// MISSION BRIEF
Your Mission
Bolting an LLM into CI can save real time or quietly wreck your signal. This quiz covers advisory patterns and the hard line: do not make an uncalibrated free-form model the sole authority over pass or fail. A validated model grader can participate when its risks are measured and bounded.
// FIRST CONTACT
Battle teaser
A sound use of an LLM step in CI is:
- AMaking an uncalibrated free-form model the sole authority that decides whether the build passes
- BHaving the model rewrite failing tests to pass
- CSummarizing a failed run and suggesting likely causes, as an advisory annotation on top of the deterministic result
- DReplacing assertions with 'ask the model if it looks right'
Answers, scoring, hints, and the full battle stay sealed.
// SKILL TAGS
ai-workflowsci-cdguardrailsai-for-qa