takerofnaps

Sonnet 5 seemed somewhat benchmaxxed to me. So was Opus 5. I wonder if this will be as big of an improvement as opus 5 -> opus 5.5. Maybe I will switch back from GLM 5.3 flash for some tasks.

I agree with you. I also wouldn't trust it calling on my behalf or doing the majority of what people say it can do. Putting aside it making mistakes and not even being beneficial, I definitely wouldn't want meta of all companies to have the relevant information to do those things.

At my work I have a $500 monthly AI budget. I have been using the $200 Claude subscription and most of my use is with Claude code. I think I'm going to switch to either kimi or glm and use the opencode harness. Both fable 5 and opus 5 have outright refused things like security related bug fixes and making monitoring tools. I am so happy that open models are good now

I am concerned about this from both a mental health standpoint and a privacy standpoint. Replika users have been shown to form real emotional dependence on it and they start feeling responsible for the bot's feelings. https://journals.sagepub.com/doi/10.1177/14614448221142007. There have been a bunch of cases where these models/harnesses lead to suicide.

On privacy, chatbots and agents hosted by big providers collect all sorts of telemetry and build an entire behavioral model of users (this is why i use local open weight models). Now stick all that in a robot with a camera and mic that lives in your house. Way bigger attack surface. I know it claims to process and store everything locally, and even if that is the case, there are sure to be similar robots that dont.