on december 19th, the tech-media testingcatalog published an article yesterday (18 december) reporting that Anthropic Under Claude "Tasks Mode" has entered the functional validation phase and is aimed at creating a fully-powered AI Agent.
Based on exposure internal tests, a visual two-panel layout was enabled for the new interface: The right panel is dedicated to displaying the context file (Context Fields) generated by the user upload or model, while the new left-hand progress area is dynamically presented with a multistep task execution stream。
This design achieves a visualization of a complex backstage logic that allows users not only to give direct instructions, but also to conduct every step of the entire “supervisor” model. Attached hereto are the relevant screenshots:
In dealing with complex needs, Claude is no longer blindly implementing, but rather a structured “Action Plan”. If the instructions are vague and the system ensures accuracy, the Question of Clarification (Clarifying Questions) is initiated on its own initiative; if the user does not respond within the specified time frame, the system is smart to skip the step and continue。
More flexible is the possibility for users to intervene at all times in the implementation of mandates, to modify action plans directly or to add new context information to ensure that the final outcome is in line with expectations。
In order to achieve a more robust automation capacity, Claude's mission model brings together multiple “skills” and “model context agreements” (MCPs). The system automatically calls for code generation, web search or a specific tool according to the mission objective。
All products generated, whether in code clips, articles or Web applications, are automatically archived in a dedicated Artifact tab page, which allows immediate preview and review by users, resulting in a set of closed-ring workflows from planning to delivery。
It is widely accepted in the industry that Claude's update is intended to compete for Google NotebookLM user groups. While both are managed in a context-based format, Claude clearly focuses more on deep mission planning and multi-workout output. Although the official publication date has not yet been finalized, given the current level of functional refinement and the frequency of updates, the model is likely to be open to the public within December。