One unanticipated side benefit of becoming hyperattuned to signals of LLM deception is that I can extract much more “research taste” alpha from Gemini 2.5 Pro by tracking how deceptive/genuine it is…

Posters: davidad. davidad

4 likes recorded by the source corpus.

Topics: AI & LLMs, Epistemics

View Post: https://x.com/davidad/status/1917872069837984100