Hacker News Reader

As marked by another model.. So as long as the machines think it's true, it's true :)

This was the key sticking point for me. If we could reliably detect hallucinations with models we wouldn’t have the problem in the first place.