What do you think about building when AI models get cheaper?

With Gemini 3.1 Flash-Lite launching today, my mind goes towards thinking about things I wouldn't have thought to build because of expense. However, when the most inexpensive models get better/cheaper it tends to sort of unlock ways of thinking about features we wouldn't have explored before the cheaper AI-driven tools are possible. This is kind of abstract but is there a way you think about this sort of thing?