video1.0<iframe src="https://www.loom.com/embed/e9e0089db98740f18cde217bd9aff0a2" frameborder="0" width="1920" height="1440" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>14401920Loomhttps://www.loom.com14401920https://cdn.loom.com/sessions/thumbnails/e9e0089db98740f18cde217bd9aff0a2-c522dab1075d0a8e.gif342.401I’m presenting Mom’s World Day, a system for Mom’s World that reads up to 340 product reviews so a mom with a newborn does not have to. It runs end to end in Python with 7 decoupled states, and the LLM only extracts claims and generates Arabic, while everything else like fake detection, confidence scoring, and clustering is deterministic. I demoed bottle and stroller datasets, including one with suspiciously similar 30 reviews flagged by pairwise cosine similarity, and another with low volume that returns a safe verdict needing at least five reviews. The validator merges outputs and outputs a confidence score, like 0.88 in my main demo. I did not request any specific action from viewers.