Tencent releases beta version of WorkBuddy feature for desktop applications
Read more
Pandaily
pandaily.com

Tencent releases beta version of WorkBuddy feature for desktop applications

Tencent announced the public release of the WorkBuddy feature in beta for desktop applications on November 20. This feature extends the capabilities of the office agent to desktop programs that do not provide an API, connector, or web version.

Chinese reports describe the system's operation as being based on interface recognition combined with click, text input, and scrolling actions, rather than the release of a new model with altered weights. The beta version supports Windows and macOS operating systems and requires WorkBuddy client version 5.11.0 or later.

Users must activate this capability and grant the necessary operating system permissions, including screen accessibility and recording. Tencent warns that the ability to recognize interfaces varies depending on the application, so compatibility cannot be assumed just because the application is open on the desktop.

The examples provided in the announcement relate to existing business software: searching records in an education management system or matching inventory and asset information between clients using a link to incoming stock. The goal is to eliminate workflows that would otherwise require manual data copying between unconnected programs.

Users can use a side window displaying in icon mode to monitor the progress while continuing other tasks. However, this does not guarantee the security of simultaneous interaction with the same application: changing focus or page state may interrupt the agent's task. This distinction is critical for programs where a click depends on the current window, selection, or form.

For sensitive operations such as logging in, granting authorization, making payments, sending information, or deleting records, the official recommendation is for the user to remain present and monitor the process. Users should also return to the original system to verify results after the task is completed. These warnings are supervisory advice, not independently tested guarantees that every risky action will be blocked.

This announcement represents a WorkBuddy client-level functionality, which differs from the concept and usage capabilities of the computer described earlier in November for Tencent's Hy4 model. Furthermore, it does not establish that the feature became available specifically on November 20; previously, the user account described a beta setting. The current news concerns Tencent's public announcement and the specification of required supported clients.

Similar stories

Doubao Work introduces AI-powered workspaces for 21 professional scenarios
Read more
pandaily.com

Doubao Work introduces AI-powered workspaces for 21 professional scenarios

ByteDance released a workspace feature in Doubao Work on November 16th, which integrates pre-configured tools and skills for 21 different professional scenarios. These scenarios cover fields such as healthcare, education, creative design, and financial research. According to information published by Jiemian, this feature is available in both the Doubao Work desktop client and the Doubao mobile application.

Workspaces organize the agent's operation around repetitive tasks characteristic of a specific profession, rather than requiring users to manually assemble every skill and tool. Users navigate to the workspace page and select a professional workspace where relevant entry points for task execution are already prepared. Doubao claims this reduces setup effort, although the announcement does not contain an independent performance evaluation.

Application in Various Professional Fields

In the medical workspace, the company described functions related to searching clinical guidelines, extracting and interpreting medical literature, preparing public awareness materials and communicating with patients, as well as creating research presentations. These functions pertain to information processing and documentation. Furthermore, the launch does not establish clinical accuracy nor replace the physician's duty to verify the results obtained.

The teacher workspace allows for the generation of lesson plans and curricula based on a specified topic, textbook, and learning requirements. As reported in Chinese media coverage of the launch, the system can also transform questions into interactive demonstrations with customizable parameters and draggable elements. An additional resource connector provides another avenue to educational materials and searching for similar questions.

The creative design workspace combines tasks for developing posters, social media images, and visual branding. When a task begins, the system creates a canvas where users can review and organize the results, and then develop the idea into related materials. This packages existing content creation capabilities into a task-specific starting point, rather than announcing a new model for image or video generation.

The financial analysis workspace includes interpreting announcements, researching companies, and preparing briefings. Through plugins, it can connect to professional data services after authorizing the respective user accounts. It is important to note that these connections should not be regarded as unrestricted access to all provider data or as confirmation of the agent's conclusions.

Users are not limited to predefined workspaces. Doubao states that they can create personal workspaces, customize work requirements and preferences, save them, and share them with others. The object of exchange is a reusable work configuration, not necessarily the underlying private files or licensed data.

The update shifts the office agent's entry point towards profession-specific workflows. Its value will depend on the suitability of the included tools, the permissions available to the user, and the quality of the final output, none of which is demonstrated solely by the number of preset options.

TokenRhythm releases limited beta version of agent-oriented NeoHorse-1.5-Flash Preview model
Read more
pandaily.com

TokenRhythm releases limited beta version of agent-oriented NeoHorse-1.5-Flash Preview model

TokenRhythm, a Chinese artificial intelligence infrastructure developer managing a multi-model routing platform, unveiled NeoHorse-1.5-Flash Preview on November 5th. The company positions this model as agent-oriented, designed for task execution and software development. Access is exclusively available through a limited beta version, requiring users to apply via the company's official WeChat account.

According to TokenRhythm, the model was trained based on Qwen3.8-Flash and has a total of 180 billion parameters, with 6 billion actively utilized. In its self-comparison against the base model, the company noted an increase of 9.23 points in the Agents' Last Exam test and 5.92 points in the Natural Language to Repository test. Furthermore, improvements exceeding 20 points were achieved in physical science tasks, alongside progress in developer tooling, testing, web development, and cryptography and security. It should be noted that these evaluations were conducted by the company itself, and independent results have not yet been published.

This release is considered the second stage of TokenRhythm's work on recursive self-improvement. After releasing NeoHorse-1 in September, the company returned this model to its pool, gathered feedback on real-world task performance, and used it to train the next version. The company's routing system provides signals regarding required capabilities for tasks, and running several advanced models on the same task yields comparable results. Differences between the results of NeoHorse and these data are used as training signals. The team named three methods underlying this process: multi-model differential diagnosis, adversarial environment evolution, and online swarm learning.

This cycle also relies on OpenSquilla, an open-source product from TokenRhythm. In an update released the day after selecting the model level based on task intent, the selected level is maintained when continuing the task and increased upon retries.

TokenRhythm's public pages on GitHub and Hugging Face confirm the existence of the previous generation, not the new preview version. The NeoHorse-1 model was released on September 7th in 4B and 9B versions, fine-tuned on Qwen3.5 under the Apache 2.0 license, along with a technical report and quantized versions. At the end of September, the 4B decision-making model, named NeoHorse-Jev, followed. As of the publication, downloadable weights, a model card, or a license for NeoHorse-1.5-Flash Preview were absent from these pages, meaning it cannot be considered an open-weights release.

Tencent updates WorkRally platform, adding three video models and expanding functionality for various content types
Read more
pandaily.com

Tencent updates WorkRally platform, adding three video models and expanding functionality for various content types

Tencent has updated its AI-powered content creation platform, WorkRally. The update introduces three new video models and expands the toolset capabilities for a wide range of content creators. The platform is now suitable not only for animated short dramas but also for shooting live-action short dramas, commercials, and other creative videos.

According to reports published in Science and Technology Daily on November 6 and Shanghai Securities News on November 7, the update is recent, although an exact release date was not specified. Among the new models is Rally-Video, designed for film and animation work, capable of using images, video materials, and non-textured 3D models as references.

Another model, Rally-Video-edit-2.0, allows editing existing clips by changing only selected elements, such as the background or character, while preserving the original duration, resolution, and quality. The third model, UpScale-rally, is a super-resolution tool focused on detailing scenes and people, supporting speed and resolution modes with output options in 2K and 4K formats.

The WorkRally product team reported that processing time has become 80% shorter compared to the previous version, although this figure is based on Tencent's data and has not undergone independent testing. Creators can choose the appropriate model within existing workflows based on the task and budget.

In addition to the models themselves, the update includes moving the entry point for agent-based content creation to the homepage, enhancing agent and specialized knowledge base features, and adding tools for AI script generation, script diagnostics, and intelligent editing. Science and Technology Daily also reported on a one-click asset replacement feature and a new capability for registering copyrights for AI-generated materials.

A representative from the WorkRally product team noted that Tencent Video has transformed years of production experience into vertical models, skills, and knowledge bases that creators can access directly. These resources help address issues related to shot sequence, lighting, and camera movement. The goal of this step is to minimize the need for trial and error, allowing authors to focus on storytelling.

The professional version of the platform, intended for animation studios, film crews, and production companies, remains available by invitation only. It covers aspects such as previsualization, concept design, scene compositing, CG asset production, and visual effects. Information regarding the pricing of the new models was not disclosed in the reports.

Popular