"Chat is an outdated interface" – designers building with AI

What are the alternatives?

Intent and AI agents — What happens when you give AI full context and agency? A real world example is tab autocomplete. Extrapolate this and imagine a general purpose agent. Weekend activities based on your calendar, a followup email based on a phone call.

Generative UI — Surfacing relevant UI on demand. Why have to navigate to settings, instead ask to change card details and see relevant UI as needed.

Physical environment as an input, and output — General visual input with Humane, or using physical objects as the control (Dynamicland), and projection as the interface.

Fully generative output — Long term, is there a need for structured interfaces? Boldest of all the concepts, but the generative Minecraft Oasis demo proves it's possible using video models. Regenerating the output in the optimal way every time.

Voice as an input/output — Perhaps the most obvious, but if everyone is so inclined to speak, why do we text more than call? Should I have to speak to change a song or select the pen tool? Instead, it feels like voice should be a secondary input, offering additional context.

@dylanlp_