This is what I have been working on and although the research focuses on literary creativity I strongly suspect that it also applies to inventiveness in general including the mathematical intuition that leads to new discoveries. The Goblin Realms: Topological Mapping of Human and AI Semantics The "Goblin Realms" framework identifies a fundamental topological divide between the way humans and artificial intelligence navigate the landscape of meaning. Using high-dimensional mapping, the research reveals that human creativity relies on a high "leap rate," where authors jump across semantic voids to connect distant concepts. In contrast, AI models perform an "algorithmic crawl," sticking to predictable, precomputed paths even when instructed to be original. This distinction is quantified through the "Goblin Criterion," which focuses on versatile, low-friction words that serve as hubs for these creative jumps. The study further demonstrates that reading difficulty is driven by these cognitive leaps rather than simple vocabulary rarity. Ultimately, neurobiological data confirms that while AI mirrors standard human logic, it fails to replicate the unique neural patterns associated with true divergent thinking. I still have more multilingual experiments to run with Large Language Models, then human experiments ranking results (if I can get funding), then I hope to find the holy grail, a harness for LLMs that lets them make genuinely creative leaps in the same way humans that do. And then formally write it all up, so the AI's can read it, because a lot of humans don't read anything these days. 😆