was thinking about what happens when agents become the default interface for everything. in theory you just text your agent and it books the uber, orders the food, whatever. but some things need to be visual, humans evolved for visual interfaces, and a chat box is still just text in a prettier wrapper. the screenshot processor thing is a good example, it used to be an app where you could see and sort and search through pictures, and now it's all through the agent which is functionally capable but loses that visual browsing layer.
so maybe the future is micro apps, where the agent generates a custom ui on demand for whatever you need. a calorie tracker, a ride interface, whatever. but the problem is you can't build an agentic version of uber because you need their api, their driver network, their infrastructure. and even if you could, uber cares what ui you use because their ad revenue depends on you seeing their app. apis don't have ads built into them. what if they did? what if an api could bundle an ad into its response so the company still gets paid when you use their service through a third-party micro app. feels like that should exist.
it probably won't work that way though. developers would revolt if an api returned a forced ad widget, and most of these companies already make money on transactions not impressions. the real shift is that transaction fees replace ad revenue as the dominant model when nobody's looking at the app anymore. companies won't voluntarily give up dark patterns either, they'll fight by restricting api access, embedding dark patterns in the data response itself, or just refusing to open up. the micro app future doesn't happen because companies choose it, it happens because enough of them are forced into it by regulation or competition. the agent layer is the next platform war, and whoever owns it owns the user.