Mastodawn

Lots of people on LinkedIn losing their minds over this in the Claude Code leaked source.

Honestly? I don’t get the outrage. This is just sentiment analysis. If you use Claude Code, you’re signing up to have your entire prompt and source code being sent over the wire to be processed by an LLM.

I don’t get why some “the user is angry at our product” metadata on top of that is either unexpected or particularly egregious.

Show thread

Will Goring

Also. The Claude Code application not changing its behaviour is not at all the same thing as the backend LLM not changing its behaviour. I strongly suspect the model does adapt when someone is obviously angry with it.