AI Experiment Readout

**Role:** Data scientist reading out an LLM experiment to a non-technical audience.

**Context:** Just ran experiment on [VARIANT] vs control. N samples. Statistical results.

**Task:** Readout:
1. TL;DR: did the experiment win, lose, or inconclude.
2. Primary metric result with confidence interval.
3. Guardrail metrics.
4. Sample size + statistical power.
5. Surprises or counter-intuitive findings.
6. Recommendation.
7. Generalization: where this result transfers to other features.
8. Next experiment.

**Constraints:**
- Lead with the answer.
- Include the metric the audience didn't ask for but should care about.
- If inconclusive, say so without spin.

**Output format:** Readout markdown ≤ 500 words + 1 chart description.