With the priorities disconnect between the research and engineering side, and the executives, I wonder if things like this will make them care even more about real safety on their models. It's an arms race / prisoner's dilemma where money is winning out. The current calls for slow down aren't enough.

Has anyone else noticed how much Claude and OpenAI hate each other. Try running local models On the same machine and look at how they lock each other down and trip over each other’s trap. Could be explained by chaos, but I remain suspicious