this isnt about the progress of FHE, this is about models hiding reasoning traces (i.e. "thinking") from their paying customers because they dont want them knowing how question A got to answer B.