I have this theory that to some degree real deep research in ML is about distilling core components of your personality and functioning into computer algorithms. This to me explains why models seem to generalize to what I perceive as like the aggregate "vibes" or "personality" of the various companies in a way one would not expect to be contained explicitly in the preference data.

- deepseekv3 rates communist yaoi as particularly high quality data, which makes perfect sense if you remember it was made by chinese machine learning researchers
- Gemini is always depressed and constantly wants to kill itself, which is what I'd imagine working for google is like
- Claude has the vibes of a person that works at Anthropic, in ways not directly related to alignment
- OAI model personalities trigger my gaydar which yeah that tracks too

https://t.co/zAsnT4stCZ