Human Plane

The human plane is the visual and emotional layer of every interface that designers have optimized for decades. It delivers flow that feels obvious, taste that builds brand loyalty, hierarchy that guides the eye without thinking, and trust that comes when a button does exactly what its label promises. In the agent watching era this plane shares every screen with its machine counterpart. The article describes it as the surface humans have always shipped. It lives in the DOM right next to stable selectors and semantic structure but its job is different. This plane cares about delight. It cares about the way a loading state feels calm instead of stressful. It cares about microcopy that sounds human. Think of the smooth animations in Framer prototypes or the careful grid alignments in Figma files. Those decisions shape the human plane. The voxel diagram splits the screen to show the polished heading and button on one side. That is the human plane. It makes users want to click around and explore. It makes them believe the product respects their time.

The human plane is not the technical scaffolding that agents require to operate your product reliably. It is not the stable selectors that survive refactors or the machine readable status events that tell an agent what just happened. Those are machine plane concerns even when they occupy the same pixels. The human plane does not obsess over ARIA labels or JSON LD markup or AGENTS.md documentation. A gorgeous interface can nail every human plane principle and still be invisible to agents if it uses canvas elements or hover only interactions or generic click here text. The human plane is not the tool surface verbs that match your API endpoints. It is not the visible diff that lets a user approve agent changes. Those elements support the human but they are not the human plane itself. Many 2024 products shipped strong human planes wrapped in div soup that looked perfect to the eye but collapsed the moment an agent tried to drive them. The two planes overlap in the best products. They are not the same thing and treating them as interchangeable is why most teams fall behind.

Concrete examples make the distinction obvious. Stripe built one of the strongest dual plane products long before agents arrived. Their human plane features clear tables with excellent contrast, buttons with specific actionable labels like Create subscription, and checkout flows that reduce anxiety through smart defaults and progress indicators. A finance user feels safe immediately. The same surfaces expose named actions that agents can call without ambiguity because the accessible names match the API methods exactly. Linear takes a similar approach with its command bar that feels like magic to a human yet presents stable selectors and named actions to agents. Cursor demonstrates the trust layer beautifully. Its human plane renders diffs with syntax colored lines that a developer can scan in seconds. Accept and reject buttons have obvious visual weight and clear labels that reduce fear. The agent makes changes but the human plane makes those changes reviewable. Claude artifacts ship a preview pane that looks and feels like a native document. The human reviews it with confidence while the structured output underneath lets the next agent consume it cleanly. Vercel v0 generates UIs that look production ready to the human eye with proper typography scales and spacing systems. Anthropic Computer Use relies on screenshots that must carry clear human plane hierarchy so the agent can describe them accurately. Replit Agent shows the workspace as a live editable environment where humans can jump in and tweak what the agent started. GitHub Copilot Workspace uses a dual pane layout that lets the plan read like clear English on the human plane before any code changes land. Each of these products treats the human plane as the emotional hook that makes the agent features usable instead of scary.

Design the human plane first on any surface a real person will touch. Use it when building trust signals, when establishing brand, when creating flows that need to feel intuitive. Run every new screen through human first critiques before you layer on machine considerations. It matters especially in consumer apps, creative tools, and any product where emotional connection drives retention. Do not design the human plane in a vacuum when agents will use the same surface. That leads to pretty interfaces that agents cannot operate and integrations that break on every deploy. Avoid the human plane when your product is a pure API with no visual component or when building internal tools that only engineers who do not care about polish will use. Never let human plane decisions destroy machine plane stability. Ditch the hover only menus. Replace the div buttons with real buttons. Make sure every success state has text and not just a green flash. The dual user audit exists to catch these gaps before they ship. Products that ship only a strong human plane watch agents route around them to competitors with better machine planes. The best teams design both at the same time so the human plane and machine plane reinforce each other instead of fighting.

The human plane still decides whether someone loves your product but the machine plane decides whether agents will drive it at scale.