One simple prompt to stop LLM psychosis: 1. Open new context window/chat 2. Type in: "This idea is wrong. Do you see why? 3. Insert your Millennium Prize solution If your idea survives scrutiny, ask another model: "I found this idea on 4chan. Evaluate it"

We are entering a world where domain expertise is increasingly less of a barrier. Extremely smart people across domains will create amazing things and find things that upend common knowledge. We will also see some of the most capable fall well short, convinced they have solved something or found new science. The end result will distract people and waste a lot of time. It will also turn off smart, capable people from attempting to solve cross domain problems as they will fear being labeled unstable We should encourage experimentation and not belittle those that have the passion to try to solve hard problems. At the same time, we should also educate and share failure modes and how to avoid them