Post #2167861

After having worked half a day (my colleague spent even more time) trying to configure some reasonable secure settings for #claude we are surprised how buggy, broken in general shitty their permissions model and sandboxing is. It took us half a day to come up with a better working bubblewrap wrapper than they provided. You wouldn't guess that #Anthropic has billions at their disposel. At least it isn't my money. #llm