Plate
Plate is a GUI for the AI-native device that's coming.
The assumption underneath: people will own an AGI-level personal helper the way they own an iPhone or Mac today. This helper becomes the main entry point to the digital world — doing what personal devices do now, but with AI as the thinker and executor, reasoning about what the user wants and arranging the experience to deliver it.
Plate is the surface where that arrangement becomes visible.
The grammar built for the manual era — windows, menus, inboxes — wasn't designed for this. Some of it carries forward. Some of it shouldn't. Plate is a thesis about which patterns survive, which retire, and which still need to be invented.
What Plate is, and isn't
Plate isn't competing with productivity apps or chatbots. It's working at a different layer — where the device itself meets the person.
Plate isn't the AI. The AI is the device's brain; Plate is the surface intelligence comes through. This separation matters. Plate's value isn't in being the smartest model — it's in being the place where intelligence becomes usable, recognizable, and trustworthy.
Plate is hardware-agnostic. As input and output diversify — voice, AR, eventually neural — Plate's grammar persists. The bet is that interaction is defined by its primitives (intent, materialization, persistence), not by keyboards or screens.
Plate is measured against a future state. Today's work demonstrates direction. The product matures as AI matures and as AI-native devices become real. This is a long-horizon bet, not a quarterly one.
What Plate is made of
Plate has three layers.
Interaction grammar — the rules and patterns for engagement. Curatorial by nature: keeps what still works in today's GUI, retires what doesn't, invents what the AI-native context demands.
Visual vocabulary — the elements those patterns are built from. Familiar primitives skinned for Plate, plus new ones invented when the grammar requires them.
Space — the bounded environment where interactions occur, accumulate, persist, and fade. Not assumed to be 2D. Its shape follows the hardware Plate runs on.
Plate is interaction-led, not component-led: the grammar is the thesis, the vocabulary serves it, the space holds it. The AI composes within these three — it doesn't bypass the grammar or invent vocabulary at runtime.
Principles
Working principles, used to decide what Plate is and isn't. They'll evolve.
The AI can automate a lot of work that's manual today — quietly or visibly. But decisions with consequence belong to the user. Plate keeps those moments recognizable and reversible.
Plate keeps patterns that fit how humans think and recall. It retires patterns that exist because of past technological limits and no longer earn their place.
Every surface, motion, and notification spends some of the user's attention. Plate treats that spend as real cost, and only spends when the value is clear. What "clear" means is judged in context, not by a universal threshold.
Content surfaced without an explicit user request is presented as available, not pending. The user can take it, ignore it, or never notice it. Anything that requires a dismissal has already overstepped.
The user's work persists, accumulates, and can be returned to. Plate makes that context visible and navigable on the surface, giving the work a place to live across time.
The user should be able to tell what was done, why, and how to change or undo it — when it matters. Plate makes the relevant AI actions digestible, recoverable, and actionable on the surface.
A working document. Revised in public as the thinking sharpens.