DreamGPT: turning a weakness of large language models into a strength

DreamGPT turns LLM hallucinations into a feature for divergent thinking, generating innovative ideas through unexpected connections. In this demo, I’ll show how DreamGPT amplifies hallucinations using fine-tuned prompt engineering and a custom backend pipeline, walking through the key code and algorithms. We’ll explore real-time examples of productive chaos, reframing a common AI flaw into a powerful tool for creative ideation.