A runnable experiment in generative UI: type a search, get a full-page image, tap where you want to go next, get another full image—moving through the picture instead of digging through menus or link lists.
If you haven’t bumped into generative UI in this sense yet: the screen is generated on the fly when you type or tap—not the same fixed chrome with slightly smarter text.
This isn’t claiming to be the final shape of anything, but it’s fun to use. The line I keep coming back to is moving through the picture instead of treating the web like a directory listing.
Opening search → full-page image → tap to branch → next image. Behind that: web snippets, a scene compile step, image generation, and vision on the tap—all wired together, not a slide mock.
Tradeoff, said plainly: link lists still win for exact answers and provenance. This shape shines when “show me” beats “tell me in paragraphs.”
MIT — Copyright (c) 2026 Manu Pareek.