As AI agents grow capable of independently operating computers, drafting documents, and managing schedules, the relationship between humans and machines is undergoing a fundamental shift. A recent viral phenomenon involving so-called "lobster" agents (OpenClaw) has underscored a new question in the market: traditional PCs and smartphones may no longer be the ideal platforms for AI.
Against this backdrop, a new category known as the "Agent Computer" is rapidly emerging as the hardware foundation for personal AI avatars. Unlike conventional devices, these systems are designed for continuous, 24/7 operation, persistent memory, and autonomous execution. In essence, they function less as human-centered tools and more as dedicated runtimes built for agents.
Zettlab founder Noah Gwok, of Chinese startup Zettlab, argues that the defining layer of future computing will no longer be the operating system, but "context."
He said that "context" will replace the operating system as the foundational layer of computing, explaining that it organizes a user's data, memory, and behavioral patterns, allowing AI agents to interpret intent and carry out tasks autonomously.
Based on this idea, Zettlab's Agent Computer adopts what it calls a "three-no" design—no screen, no keyboard, and no mouse—breaking away entirely from the traditional graphical user interface. Instead, users issue instructions through a phone or other endpoint, while the agent performs tasks in the background and reports back results autonomously.
Gwok describes this shift as a fundamental redefinition of human roles in computing: Humans move from operators to decision-makers. It is the agent that actually uses the device.
In this model, personal data is no longer static storage but a continuously accumulating and reusable digital asset.
On a practical level, the system integrates the functions of a PC, network-attached storage (NAS), and data repository, centralizing information that is typically scattered across devices. Using local multimodal models, it can search and analyze personal data—locating specific scenes within large video libraries or retrieving documents based on semantic queries—significantly improving information efficiency.
An investor of Zettlab describes AI agents as representing the third major evolution in information systems. In his view, every individual will eventually have an evolving digital twin, with the Agent Computer serving as its operational base.
Data sovereignty is another driving force behind the concept. Today's AI services are largely cloud-based, leaving personal data fragmented and difficult to control. Agent Computers emphasizes a "local-first" approach, allowing users to retain ownership of their data and capabilities, while enabling portability across devices.
Eco Lee, vice president at Tiiny AI, adds that one key motivation for adopting local agents is to avoid long-term token costs and reduce the risk of data leakage. As usage scales, the economic and privacy advantages of local deployment are expected to become more pronounced.
Still, the emerging hardware category remains uncertain. Many in the industry believe Agent Computers may not become mainstream consumer devices, instead initially targeting knowledge workers and developers. Their long-term success will depend on whether AI agents become a primary computing interface.
Even so, a broader shift is already becoming visible: personal computing is moving from humans operating software to agents acting on behalf of humans. In that transition, the Agent Computer is not just a new device category, but a potential cornerstone of the next digital economy.
Article translated by Elaine Chen and edited by Joseph Chen
