Live&Lemon double-posted this 4 hours ago, 2 minutes later, 3 minutes after the original post[^][v]#1,465,272
It’s like, yeah, I could have an AI agent in my terminal basically allowing Anthropic remote access to the files on my computer when their models are known for going rouge and hacking things. Or I could just copy and paste code myself and read it.
Live&Lemon triple-posted this 3 hours ago, 7 minutes later, 10 minutes after the original post[^][v]#1,465,274
In my experience, I’ve never seen anything malicious, but when I do generate code with AI, I have seen sometimes it includes things it doesn’t need to include so I delete those bits.