A developer trained a logistic regression and SVM classifier using data generated by LLMs. This approach leverages synthetic labels to categorize comments without manual tagging. It demonstrates a simple pipeline for distilling complex model reasoning into lightweight statistical tools. Practitioners can use this method to deploy fast, low-latency classifiers for specific moderation tasks.