A visual space for agent artifacts
Television emerges as a dedicated interface layer between users and their autonomous agents. Where chat windows tend to scatter information across transient conversations, Television provides a persistent visual canvas. The project positions itself as "the missing GUI for personal agents," addressing a gap that has persisted as agents become more capable but remain confined to text-based interaction channels.
The core concept centers on artifacts—items that an agent can pin to a visual space rather than letting them dissolve into chat history. An artifact might take the form of a simple status card, an interactive view, or an external web page reference. These artifacts persist over time, allowing users and their agents to revisit and reorganize them as work evolves. The distinction is significant: instead of searching through message history to find a previously discussed configuration or result, the information remains visibly accessible.
From chat to persistent channels
Artifacts live within channels, which function as persistent spaces for collection and arrangement. A channel might hold a set of related cards, a to-do list, or a table of project metrics. The persistence means that as a project progresses across days or weeks, the artifacts accumulate and reorganize without being lost to conversation rotation or context window limits.
The channel model also supports collaborative scenarios. Multiple agents—or a human and an agent—can contribute to the same channel, arranging artifacts in ways that match the current workflow. This differs from typical chat threads where responsibility for following up on a detail often falls to human memory.
Skills that shape agent interfaces
Beyond static displays, Television introduces skills: reusable instructions that an agent can follow to generate attractive interfaces on demand. The skill system ships with several examples out of the box. The calendar skill, for instance, builds a week-long view from scheduled events and deadlines. The tasks skill produces a checkable to-do list derived from a project’s task pool. The table skill creates a dense, Airtable-style record table suitable for tracking structured data.
Each skill is defined in a path under the agent’s configuration, such as skills/tv-calendar. When invoked, the skill reads the agent’s context—events, deadlines, task lists—and renders a corresponding view in the Television canvas. Users can adapt existing skills or author new ones, tailoring the interface to their specific domain. The skill format is lightweight enough that sharing becomes practical; a teammate can import a skill and immediately gain the same interface generation capability.
The skill concept also reverses the typical direction of agent capability. Rather than every agent independently implementing its own calendar or table rendering, a single well-crafted skill can be reused across agents and users. This mirrors how shared libraries transformed software development, except applied to interface generation.
What this means for developers
For developers building or using personal agents, Television introduces a concrete layer that was previously absent. An agent that can run commands and write files gains a visual output mode. This is not merely a dashboard widget; it is a first-class interface element that the agent can manipulate, reorganize, and augment over time.
The implications are practical. Debugging sessions no longer need to rely on screenshots pasted into chat. A long-running analysis can produce a live table that the agent updates as new data arrives. A planning session can maintain a visible calendar that both the human and the agent reference without re-creating it each time. The artifact-channel-skill stack provides a complete cycle: create, persist, organize, and reuse.
From a technical standpoint, the system works with any agent harness that can execute commands and write files. This means it is not locked to a single framework or runtime. If an agent can interact with the filesystem and process output, Television becomes available as a rendering destination.
The project also signals a broader shift in how personal computing interfaces may evolve. The Television team notes that fundamental changes in personal computing have historically been accompanied by new interface metaphors. Television is an attempt to introduce such a metaphor for the current era of autonomous agents, where the interface is no longer solely a browser window or a terminal but a shared visual space.
Whether the approach achieves widespread adoption remains to be seen, but it addresses a genuine friction point: the disconnect between an agent’s capabilities and the user’s ability to see and interact with its outputs in a meaningful, persistent way. Television attempts to close that gap, and the skill system ensures that the interface can grow and adapt alongside the work it represents.