Related ongoing thread:
OpenAI's new reasoning technique alarms AI safety experts - https://news.ycombinator.com/item?id=49552395