Unverified50% confidenceOpinionExact time
Software operation and embodied tasks share similar interaction primitives: observing the screen (visual perception), moving the cursor (localization and navigation), clicking (action execution), and exploring the interface (trial-and-error learning).
1
Sources
50%
Confidence
Long-term
Relevance
11/9/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAndroid's interaction primitives are built around touch, gestures and single-window full-screen apps, multitasking has long relied on 'recent tasks' switching, and keyboard and mouse support is a capability added later.76% similarUnverified命令面板模式的普及标志着软件交互从「菜单导航」向「意图直达」的范式转移76% similarUnverifiedCoze applications have a visual user interface and complete a series of tasks according to a predefined business process without conversational interaction.74% similarVerifiedThe underlying technology of GUI agents relies on the collaborative operation of visual language models' screen comprehension capabilities and action planning modules: first performing semantic understanding of screen screenshots to identify interface elements, then having the planning module generate specific instruction sequences for execution.73% similarUnverifiedThe generated app can be tried out on a web page via preview and includes interactive logic (such as a play button responding).72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1150692API
curl https://kongchang.com/api/v1/knowledge/claims/1150692MCP
get_claim(id=1150692)