Am I mistaken in thinking that events like the Hugging Face attack are not actually relevant in the context of what he's speaking about in that video? The video is about methods of putting guard rails on agentic development without having to inspect every single line of code. But Hugging Face/Medicare/etc were intrusions: they had nothing to do with writing code. Perhaps I have missed something here though.
Perhaps also I have missed something when I say that although this capability concept seems kind of cool on paper, who's writing the contracts to which the agents must adhere? Will they not just end up writing the contracts for themselves, in the most permissive way, slapping themselves on the back and continuing? It feels like turtles all the way up/down.
Perhaps also I have missed something when I say that although this capability concept seems kind of cool on paper, who's writing the contracts to which the agents must adhere? Will they not just end up writing the contracts for themselves, in the most permissive way, slapping themselves on the back and continuing? It feels like turtles all the way up/down.