All of these attacks that use a coding harness are just {language} & chatbot. Anything 'scary' about this is just PR from companies that fetishize 'existential risk of AI'.

undefined

Some discussion: https://news.ycombinator.com/item?id=49672281