video1.0<iframe src="https://www.loom.com/embed/d2054d21017a437fb1793e93cfff0c63" frameborder="0" width="1728" height="1296" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>12961728Loomhttps://www.loom.com12961728https://cdn.loom.com/sessions/thumbnails/d2054d21017a437fb1793e93cfff0c63-beeaa8910cd2ee32.gif261.290666This Loom explains a chatbot approach that verifies answers to reduce hallucinations. The speaker argues that typical chatbots sound confident even when wrong, so the system determines what mistakes the question could lead to, checks the answer immediately, and if it cannot verify, it says so rather than guessing. They demonstrate step by step verification on a question related to The Thousand Autumns, including handling cases where two books share the same title, and they test a fake name to confirm it does not blindly accept information. They also show examples including a subnet mask question where prior instructions remain consistent and mention slower responses and an included final repo report with comparisons to a standard chatbot and typical RAG.