Anthropic has introduced the 'Record a skill' feature within Claude Cowork for the Claude desktop application. Users can now teach the model new skills simply by recording their screen and commenting on their actions via voice, allowing the system to generate ready-to-use automation scenarios.
What Happened
The Claude desktop application now features learning through demonstration. A user records a video of their screen accompanied by audio commentary, after which the system analyzes the visual sequence and sound to create an automation script that can be re-run. The feature is available to users on Pro, Max, and Team plans.
Context
This technology implements the concept of multimodal Learning from Demonstration (LfD). Instead of writing complex code or detailed text prompts, interaction shifts to observing human actions, where the AI analyzes time-series visual data and audio to form structured instructions.
Why It Matters for the Industry
For the industry, this is a significant step toward creating full-fledged Action Agents — operational agents capable of performing actions within interfaces. The transition from text-based control to multimodal input radically simplifies the development of automation agents and shifts the paradigm of AI system creation, transforming Claude from a chatbot into a task-execution tool.
Why It Matters for Users
Regular users gain a powerful No-code AI Automation tool, allowing them to instantly automate routine tasks in desktop applications without programming skills. This significantly reduces the cognitive load required to create complex scenarios and increases productivity in daily work.
What Is Not Yet Known / Limitations
At this stage, the tool is a closed proprietary solution without an open API, which limits its integration into industrial production pipelines. Additionally, experts point to potential risks regarding security and data control when recording user actions.
Sources
Author
Look at AI, Editorial Team