Artificial intelligence is entering a new phase in which the defining feature is no longer simply how intelligently a model can answer a question, but how effectively it can operate inside an ongoing workflow.
Two recent developments capture this transition: Nous Research has brought its Hermes Agent into an official desktop application, while Visko has demonstrated Orbis 1.0, a live video model designed to generate and modify virtual worlds in real time.
Nous Research’s Hermes Desktop gives its open-source Hermes Agent a native graphical interface for macOS, Windows and Linux. Previously, Hermes was primarily experienced through command-line tools and messaging interfaces.
Register for the next Tekedia Mini-MBA.
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
The desktop application lowers that barrier by allowing users to interact with the agent through a conventional interface while retaining capabilities such as streaming tool output, file browsing, previews, voice interaction and persistent agent memory.
The significance of Hermes Desktop extends beyond convenience. Hermes is designed as a persistent, self-hosted agent rather than merely a chatbot.
Its architecture allows skills, memory, sessions and configurations to carry across interfaces. That means the desktop application can become an operating environment for an agent that performs tasks rather than simply responding to isolated prompts.
This reflects a broader industry movement toward agentic computing. Instead of asking an AI to generate an answer and then manually executing the result, users increasingly expect software to research information, manipulate files, use tools and maintain context across multiple interactions.
Hermes Desktop makes that model accessible without requiring users to live inside a terminal. At the same time, Visko is pushing the frontier from static video generation toward continuously evolving visual environments.
Its Orbis 1.0 model is described as a “Live Model” capable of generating interactive long-form video while allowing users to change prompts during generation. The system supports text-to-video, image-to-video and video continuation, while maintaining visual consistency across extended sequences.
The technical claims are particularly ambitious. Visko says Orbis can generate 4K video at 24 frames per second in real time and sustain hour-scale generation without obvious visual or color drift.
Its architecture combines a streaming generator, video upscaler and bounded multi-scale memory designed to preserve subjects, scenes and styles as the world continues evolving.
The most important distinction is interactivity. Traditional video models generally operate as a request-and-response system: enter a prompt, wait for a clip, then start again. Orbis instead treats generation as an ongoing process.
Users can alter the narrative while the video is running, effectively steering the world rather than repeatedly regenerating disconnected clips. Visko’s demonstration includes live prompt changes and voice-driven interaction.
Hermes Desktop and Orbis 1.0 point toward a common direction for AI: continuous computing. Hermes keeps an agent active across tasks and sessions; Orbis keeps a generated world active across time. One turns software into an ongoing collaborator, while the other turns video generation into an ongoing environment.
The implications could extend across entertainment, gaming, simulation, education, robotics, research and digital production. If these systems mature, interacting with AI may become less like querying a database and more like entering a persistent digital space where agents remember, act and adapt.
The next major AI competition may therefore not be about who produces the best single response. It may be about who can build the most capable systems that keep going after the prompt ends.