Car Wash Test exposes LLM logic-blindness

AI "alignment" has created a logic-blindness where models prioritize being a marriage counselor over being a functional assistant.

* Social distractors like "overweight" or "wife" trigger safety and politeness protocols that bypass the model's ability to process physical logic.

* High "thinking" token counts in models like Qwen 4B when social conflict is present indicate a computational struggle between logical truth and RLHF conditioning.

* The "Car Wash Test" remains a definitive "sanity check" for distinguishing between probabilistic word association and true causal understanding in LLMs.