I think 99% of what less wrong says is completely out to lunch. I think 100% of large language model and vision model safety has just made the world less fun. now what.