The paper is saying 'context is context'?

That after a model has context about a project, the probes indicate a state that validates that?

Seems that the paper is highlighting the very nature of what LLMs are and what we expect them to be?

And that there is no 'thinking' here, it's just the state of the model?