Better Models Won't Fix AI Companions

I ran two small tests on AI companion behavior because I wanted to understand a question people keep circling around: Are AI companions bad because the model is not strong enough, or because the product treats relationships too flatly? The result made me think the second problem is underrated. This is not a product launch post. There is no demo link here. I am sharing the test because I think it points at a real design problem in AI companions. The short version: GPT-5.5 often wrote better...