AI Fabrications, Context, and LLMs

The scientist who helped invent core building blocks of modern AI just walked away from Meta - with a blunt verdict: Large Language Models (LLMs) are a dead end.   In this episode of Real Talk, Sam Holcman unpacks why LLMs are fundamentally flawed as a foundation for serious enterprise decision-making - and walks through a live example of confident, credible, and completely wrong AI output.   The takeaway is simple and uncomfortable: the answer is not “more prompts” or “more tokens.” The answer is to stop blindly feeding the generic LLM machine and start building your own Enterprise Augmented Information™ (EAI) - an architected, governed data backbone that combines processing power with actual context and understanding, on your terms, for your enterprise.