"Both arrive at the same conclusion. Models don’t hallucinate because they’re broken. They hallucinate because the systems we’ve built around them reward confident output and punish honest uncertainty."

https://iain.so/why-ai-models-hallucinate

Why AI Models Hallucinate

In September 2025, OpenAI published a paper that said something the AI industry already suspected but hadn’t quite articulated. The paper...

Iain Harper's Blog