Fabien Roger comments on Notes from a mini-replication of the alignment faking paper

[Fabien Roger]

I am pretty new to this but I understood that this is automatically taken care of for you if you use the same prompts repeatedly with the API

True for OpenAI and Gemini, not true for Anthropic.