Most AI tools are black boxes or dense math. LLM Unboxed makes Large Language Models mechanics visually intuitive through zero-jargon animations. With no setup, sign-ups, or API keys required, it animates tokenization, vector mapping, self-attention, and memory step-by-step using beginner-friendly concepts. Uniquely, it includes a visual Topic Drift scenario showing how models hallucinate in real time as attention blips stray into wrong node clusters—giving learners true visual intuition.